mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-05-29 06:31:32 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 14

remove Vercel AI Gateway and Vercel Sandbox (#33067)

Browse source 
* remove Vercel AI Gateway provider and Vercel Sandbox terminal backend

Both Vercel-hosted integrations are removed end-to-end. Users on the AI
Gateway should switch to OpenRouter or one of the other aggregators
(Nous Portal, Kilo Code). Users on the Vercel Sandbox backend should
switch to Docker, Modal, Daytona, or SSH.

What's removed:
- `plugins/model-providers/ai-gateway/` provider plugin
- `hermes_cli/vercel_auth.py` Vercel-Sandbox auth helper
- `tools/environments/vercel_sandbox.py` terminal backend
- `ai-gateway` provider wiring across auth, doctor, setup, models,
  config, status, providers, main, web_server, model_normalize, dump
- `vercel_sandbox` backend wiring across terminal_tool, file_tools,
  code_execution_tool, file_operations, approval, skills_tool,
  environments/local, credential_files, lazy_deps, prompt_builder,
  cli, gateway/run
- `AI_GATEWAY_BASE_URL` constant, `_AI_GATEWAY_HEADERS` auxiliary-client
  header set, run_agent base-URL header/reasoning special-cases
- `[vercel]` pyproject extra and `vercel`/`vercel-workers` from uv.lock
- env vars: `AI_GATEWAY_API_KEY`, `AI_GATEWAY_BASE_URL`, `VERCEL_TOKEN`,
  `VERCEL_PROJECT_ID`, `VERCEL_TEAM_ID`, `VERCEL_OIDC_TOKEN`,
  `TERMINAL_VERCEL_RUNTIME`
- Tests: deletes test_ai_gateway_models.py and
  test_vercel_sandbox_environment.py; scrubs references across 23
  surviving test files (no entire tests deleted unless they were
  dedicated to AI Gateway / Sandbox)
- Docs: provider tables, env-var reference, setup guides, security
  notes, tool config, terminal-backend tables — English plus zh-Hans
  i18n parity
- `hermes-agent` skill: provider table entry and remote-backend list

What stays (intentional):
- `popular-web-designs/templates/vercel.md` — CSS design reference,
  unrelated to Vercel-the-AI-product
- `x-vercel-id` in `stream_diag.py` headers — generic Vercel CDN
  response header, useful diag signal on any Vercel-hosted endpoint
- `vercel-labs/agent-browser` URL in browser config — lightpanda
  browser project, different OSS effort
- `userStories.json` historical contributor entry mentioning Vercel
  Sandbox — archive, not active docs

Validation:
- 1153 tests in the 22 targeted files pass (`scripts/run_tests.sh`)
- Full repo `py_compile` clean
- Live import of every touched module + invariant check (no
  `ai-gateway` in `PROVIDER_REGISTRY`, no `_AI_GATEWAY_HEADERS`, no
  `vercel_sandbox` in `_REMOTE_TERMINAL_BACKENDS`)

* test: convert profile-count check from change-detector to invariant

The hardcoded "== 34" assertion broke when ai-gateway was removed.
Per AGENTS.md change-detector-test guidance, assert the relationship
(registry count >= number of plugin dirs) instead of a literal count.
Counts shift when providers are added/removed; that's expected.
This commit is contained in:
Teknium 2026-05-27 00:43:32 -07:00 • committed by GitHub
parent cb38ce28cb
commit febc4cfec0
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
95 changed files with 111 additions and 3088 deletions
Download patch file Download diff file Expand all files Collapse all files

2
README.md
View file

@ -22,7 +22,7 @@ Use any model you want — [Nous Portal](https://portal.nousresearch.com), [Open
<tr><td><b>A closed learning loop</b></td><td>Agent-curated memory with periodic nudges. Autonomous skill creation after complex tasks. Skills self-improve during use. FTS5 session search with LLM summarization for cross-session recall. <a href="https://github.com/plastic-labs/honcho">Honcho</a> dialectic user modeling. Compatible with the <a href="https://agentskills.io">agentskills.io</a> open standard.</td></tr>
<tr><td><b>Scheduled automations</b></td><td>Built-in cron scheduler with delivery to any platform. Daily reports, nightly backups, weekly audits — all in natural language, running unattended.</td></tr>
<tr><td><b>Delegates and parallelizes</b></td><td>Spawn isolated subagents for parallel workstreams. Write Python scripts that call tools via RPC, collapsing multi-step pipelines into zero-context-cost turns.</td></tr>
<tr><td><b>Runs anywhere, not just your laptop</b></td><td>Seven terminal backends — local, Docker, SSH, Singularity, Modal, Daytona, and Vercel Sandbox. Daytona and Modal offer serverless persistence — your agent's environment hibernates when idle and wakes on demand, costing nearly nothing between sessions. Run it on a $5 VPS or a GPU cluster.</td></tr>
<tr><td><b>Runs anywhere, not just your laptop</b></td><td>Six terminal backends — local, Docker, SSH, Singularity, Modal, and Daytona. Daytona and Modal offer serverless persistence — your agent's environment hibernates when idle and wakes on demand, costing nearly nothing between sessions. Run it on a $5 VPS or a GPU cluster.</td></tr>
<tr><td><b>Research-ready</b></td><td>Batch trajectory generation, trajectory compression for training the next generation of tool-calling models.</td></tr>
</table>

4
agent/agent_init.py
View file

@ -736,8 +736,8 @@ def init_agent(
client_kwargs["default_headers"] = _codex_cloudflare_headers(api_key)
elif "default_headers" not in client_kwargs:
# Fall back to profile.default_headers for providers that
# declare custom headers (e.g. Vercel AI Gateway attribution,
# Kimi User-Agent on non-kimi.com endpoints).
# declare custom headers (e.g. Kimi User-Agent on non-kimi.com
# endpoints).
try:
from providers import get_provider_profile as _gpf
_ph = _gpf(agent.provider)

13
agent/auxiliary_client.py
View file

@ -269,7 +269,6 @@ _API_KEY_PROVIDER_AUX_MODELS_FALLBACK: Dict[str, str] = {
"minimax-oauth": "MiniMax-M2.7-highspeed",
"minimax-cn": "MiniMax-M2.7",
"anthropic": "claude-haiku-4-5-20251001",
"ai-gateway": "google/gemini-3-flash",
"opencode-zen": "gemini-3-flash",
"opencode-go": "glm-5",
"kilocode": "google/gemini-3-flash-preview",
@ -384,15 +383,6 @@ def build_nvidia_nim_headers(base_url: str | None) -> dict:
return {}
# Vercel AI Gateway app attribution headers. HTTP-Referer maps to
# referrerUrl and X-Title maps to appName in the gateway's analytics.
from hermes_cli import __version__ as _HERMES_VERSION
_AI_GATEWAY_HEADERS = {
"HTTP-Referer": "https://hermes-agent.nousresearch.com",
"X-Title": "Hermes Agent",
"User-Agent": f"HermesAgent/{_HERMES_VERSION}",
}
# Nous Portal extra_body for product attribution.
# Callers should pass this as extra_body in chat.completions.create()
@ -3609,8 +3599,7 @@ def resolve_provider_client(
else:
# Fall back to profile.default_headers for providers that declare
# client-level attribution headers on their profile (e.g. GMI
# User-Agent for traffic identification, Vercel AI Gateway
# Referer/Title for analytics).
# User-Agent for traffic identification).
try:
from providers import get_provider_profile as _gpf_main
_ph_main = _gpf_main(provider)

4
agent/model_metadata.py
View file

@ -47,7 +47,7 @@ def _resolve_requests_verify() -> bool | str:
_PROVIDER_PREFIXES: frozenset[str] = frozenset({
"openrouter", "nous", "openai-codex", "copilot", "copilot-acp",
"gemini", "ollama-cloud", "zai", "kimi-coding", "kimi-coding-cn", "stepfun", "minimax", "minimax-oauth", "minimax-cn", "anthropic", "deepseek",
"opencode-zen", "opencode-go", "ai-gateway", "kilocode", "alibaba", "novita",
"opencode-zen", "opencode-go", "kilocode", "alibaba", "novita",
"qwen-oauth",
"xiaomi",
"arcee",
@ -59,7 +59,7 @@ _PROVIDER_PREFIXES: frozenset[str] = frozenset({
"glm", "z-ai", "z.ai", "zhipu", "github", "github-copilot",
"github-models", "kimi", "moonshot", "kimi-cn", "moonshot-cn", "claude", "deep-seek",
"ollama",
"stepfun", "opencode", "zen", "go", "vercel", "kilo", "dashscope", "aliyun", "qwen",
"stepfun", "opencode", "zen", "go", "kilo", "dashscope", "aliyun", "qwen",
"mimo", "xiaomi-mimo",
"tencent", "tokenhub", "tencent-cloud", "tencentmaas",
"arcee-ai", "arceeai",

1
agent/models_dev.py
View file

@ -158,7 +158,6 @@ PROVIDER_TO_MODELS_DEV: Dict[str, str] = {
"alibaba": "alibaba",
"qwen-oauth": "alibaba",
"copilot": "github-copilot",
"ai-gateway": "vercel",
"opencode-zen": "opencode",
"opencode-go": "opencode-go",
"kilocode": "kilo",

5
agent/prompt_builder.py
View file

@ -610,7 +610,7 @@ WSL_ENVIRONMENT_HINT = (
# misleading — the agent should only see the machine it can actually touch.
_REMOTE_TERMINAL_BACKENDS = frozenset({
"docker", "singularity", "modal", "daytona", "ssh",
"vercel_sandbox", "managed_modal",
"managed_modal",
})
@ -624,7 +624,6 @@ _BACKEND_FALLBACK_DESCRIPTIONS: dict[str, str] = {
"modal": "a Modal sandbox (Linux)",
"managed_modal": "a managed Modal sandbox (Linux)",
"daytona": "a Daytona workspace (Linux)",
"vercel_sandbox": "a Vercel sandbox (Linux)",
"ssh": "a remote host reached over SSH (likely Linux)",
}
@ -738,7 +737,7 @@ def build_environment_hints() -> str:
and a Windows-only note that `terminal` shells out to bash, not
PowerShell).
- For **remote / sandbox** terminal backends (docker, singularity,
modal, daytona, ssh, vercel_sandbox): host info is **suppressed**
modal, daytona, ssh): host info is **suppressed**
because the agent's tools can't touch the host — only the backend
matters. A live probe inside the backend reports its OS, user, $HOME,
and cwd. Falls back to a static summary if the probe fails.

4
agent/usage_pricing.py
View file

@ -711,8 +711,8 @@ def normalize_usage(
output_tokens = _to_int(getattr(response_usage, "completion_tokens", 0))
details = getattr(response_usage, "prompt_tokens_details", None)
# Primary: OpenAI-style prompt_tokens_details. Fallback: Anthropic-style
# top-level fields that some OpenAI-compatible proxies (OpenRouter, Vercel
# AI Gateway, Cline) expose when routing Claude models — without this
# top-level fields that some OpenAI-compatible proxies (OpenRouter, Cline)
# expose when routing Claude models — without this
# fallback, cache writes are undercounted as 0 and cache reads can be
# missed when the proxy only surfaces them at the top level.
# Port of cline/cline#10266.

1
cli-config.yaml.example
View file

@ -29,7 +29,6 @@ model:
# "arcee" - Arcee AI Trinity models (requires: ARCEEAI_API_KEY)
# "ollama-cloud" - Ollama Cloud (requires: OLLAMA_API_KEY — https://ollama.com/settings)
# "kilocode" - KiloCode gateway (requires: KILOCODE_API_KEY)
# "ai-gateway" - Vercel AI Gateway (requires: AI_GATEWAY_API_KEY)
# "azure-foundry" - Microsoft Foundry / Azure OpenAI (API key or Entra ID)
# "lmstudio" - LM Studio local server (optional: LM_API_KEY, defaults to http://127.0.0.1:1234/v1)
#

3
cli.py
View file

@ -562,13 +562,12 @@ def load_cli_config() -> Dict[str, Any]:
"singularity_image": "TERMINAL_SINGULARITY_IMAGE",
"modal_image": "TERMINAL_MODAL_IMAGE",
"daytona_image": "TERMINAL_DAYTONA_IMAGE",
"vercel_runtime": "TERMINAL_VERCEL_RUNTIME",
# SSH config
"ssh_host": "TERMINAL_SSH_HOST",
"ssh_user": "TERMINAL_SSH_USER",
"ssh_port": "TERMINAL_SSH_PORT",
"ssh_key": "TERMINAL_SSH_KEY",
# Container resource config (docker, singularity, modal, daytona, vercel_sandbox -- ignored for local/ssh)
# Container resource config (docker, singularity, modal, daytona -- ignored for local/ssh)
"container_cpu": "TERMINAL_CONTAINER_CPU",
"container_memory": "TERMINAL_CONTAINER_MEMORY",
"container_disk": "TERMINAL_CONTAINER_DISK",

1
gateway/run.py
View file

@ -818,7 +818,6 @@ if _config_path.exists():
"singularity_image": "TERMINAL_SINGULARITY_IMAGE",
"modal_image": "TERMINAL_MODAL_IMAGE",
"daytona_image": "TERMINAL_DAYTONA_IMAGE",
"vercel_runtime": "TERMINAL_VERCEL_RUNTIME",
"ssh_host": "TERMINAL_SSH_HOST",
"ssh_user": "TERMINAL_SSH_USER",
"ssh_port": "TERMINAL_SSH_PORT",

9
hermes_cli/auth.py
View file

@ -379,14 +379,6 @@ PROVIDER_REGISTRY: Dict[str, ProviderConfig] = {
api_key_env_vars=("NVIDIA_API_KEY",),
base_url_env_var="NVIDIA_BASE_URL",
),
"ai-gateway": ProviderConfig(
id="ai-gateway",
name="Vercel AI Gateway",
auth_type="api_key",
inference_base_url="https://ai-gateway.vercel.sh/v1",
api_key_env_vars=("AI_GATEWAY_API_KEY",),
base_url_env_var="AI_GATEWAY_BASE_URL",
),
"opencode-zen": ProviderConfig(
id="opencode-zen",
name="OpenCode Zen",
@ -1440,7 +1432,6 @@ def resolve_provider(
"github": "copilot", "github-copilot": "copilot",
"github-models": "copilot", "github-model": "copilot",
"github-copilot-acp": "copilot-acp", "copilot-acp-agent": "copilot-acp",
"aigateway": "ai-gateway", "vercel": "ai-gateway", "vercel-ai-gateway": "ai-gateway",
"opencode": "opencode-zen", "zen": "opencode-zen",
"qwen-portal": "qwen-oauth", "qwen-cli": "qwen-oauth", "qwen-oauth": "qwen-oauth", "google-gemini-cli": "google-gemini-cli", "gemini-cli": "google-gemini-cli", "gemini-oauth": "google-gemini-cli",
"hf": "huggingface", "hugging-face": "huggingface", "huggingface-hub": "huggingface",

7
hermes_cli/config.py
View file

@ -712,8 +712,7 @@ DEFAULT_CONFIG = {
"singularity_image": "docker://nikolaik/python-nodejs:python3.11-nodejs20",
"modal_image": "nikolaik/python-nodejs:python3.11-nodejs20",
"daytona_image": "nikolaik/python-nodejs:python3.11-nodejs20",
"vercel_runtime": "node24",
# Container resource limits (docker, singularity, modal, daytona, vercel_sandbox — ignored for local/ssh)
# Container resource limits (docker, singularity, modal, daytona — ignored for local/ssh)
"container_cpu": 1,
"container_memory": 5120, # MB (default 5GB)
"container_disk": 51200, # MB (default 50GB)
@ -5239,9 +5238,6 @@ def show_config():
print(f" Daytona image: {terminal.get('daytona_image', 'nikolaik/python-nodejs:python3.11-nodejs20')}")
daytona_key = get_env_value('DAYTONA_API_KEY')
print(f" API key: {'configured' if daytona_key else '(not set)'}")
elif terminal.get('backend') == 'vercel_sandbox':
print(f" Vercel runtime: {terminal.get('vercel_runtime', 'node24')}")
print(f" Vercel auth: {'configured' if get_env_value('VERCEL_OIDC_TOKEN') or (get_env_value('VERCEL_TOKEN') and get_env_value('VERCEL_PROJECT_ID') and get_env_value('VERCEL_TEAM_ID')) else '(not set)'}")
elif terminal.get('backend') == 'ssh':
ssh_host = get_env_value('TERMINAL_SSH_HOST')
ssh_user = get_env_value('TERMINAL_SSH_USER')
@ -5438,7 +5434,6 @@ def set_config_value(key: str, value: str):
"terminal.singularity_image": "TERMINAL_SINGULARITY_IMAGE",
"terminal.modal_image": "TERMINAL_MODAL_IMAGE",
"terminal.daytona_image": "TERMINAL_DAYTONA_IMAGE",
"terminal.vercel_runtime": "TERMINAL_VERCEL_RUNTIME",
"terminal.docker_mount_cwd_to_workspace": "TERMINAL_DOCKER_MOUNT_CWD_TO_WORKSPACE",
"terminal.docker_run_as_host_user": "TERMINAL_DOCKER_RUN_AS_HOST_USER",
"terminal.docker_env": "TERMINAL_DOCKER_ENV",

68
hermes_cli/doctor.py
View file

@ -25,7 +25,6 @@ load_hermes_dotenv(hermes_home=_env_path.parent, project_env=PROJECT_ROOT / ".en
from hermes_cli.colors import Colors, color
from hermes_cli.models import _HERMES_USER_AGENT
from hermes_cli.vercel_auth import describe_vercel_auth
from hermes_constants import OPENROUTER_MODELS_URL
from utils import base_url_host_matches
@ -49,7 +48,6 @@ _PROVIDER_ENV_HINTS = (
"DEEPSEEK_API_KEY",
"DASHSCOPE_API_KEY",
"HF_TOKEN",
"AI_GATEWAY_API_KEY",
"OPENCODE_ZEN_API_KEY",
"OPENCODE_GO_API_KEY",
"XIAOMI_API_KEY",
@ -324,7 +322,6 @@ def _build_apikey_providers_list() -> list:
("MiniMax", ("MINIMAX_API_KEY",), "https://api.minimax.io/v1/models", "MINIMAX_BASE_URL", True),
# MiniMax CN: /v1 endpoint does NOT support /models (returns 404).
("MiniMax (China)", ("MINIMAX_CN_API_KEY",), "https://api.minimaxi.com/v1/models", "MINIMAX_CN_BASE_URL", False),
("Vercel AI Gateway", ("AI_GATEWAY_API_KEY",), "https://ai-gateway.vercel.sh/v1/models", "AI_GATEWAY_BASE_URL", True),
("Kilo Code", ("KILOCODE_API_KEY",), "https://api.kilo.ai/api/gateway/models", "KILOCODE_BASE_URL", True),
("OpenCode Zen", ("OPENCODE_ZEN_API_KEY",), "https://opencode.ai/zen/v1/models", "OPENCODE_ZEN_BASE_URL", True),
# OpenCode Go has no shared /models endpoint; skip the health check.
@ -340,7 +337,7 @@ def _build_apikey_providers_list() -> list:
"Arcee AI": "arcee", "GMI Cloud": "gmi", "DeepSeek": "deepseek",
"Hugging Face": "huggingface", "NVIDIA NIM": "nvidia",
"Alibaba/DashScope": "alibaba", "MiniMax": "minimax",
"MiniMax (China)": "minimax-cn", "Vercel AI Gateway": "ai-gateway",
"MiniMax (China)": "minimax-cn",
"Kilo Code": "kilocode", "OpenCode Zen": "opencode-zen",
"OpenCode Go": "opencode-go",
}
@ -690,7 +687,6 @@ def run_doctor(args):
"openrouter",
"custom",
"auto",
"ai-gateway",
"kilocode",
"opencode-zen",
"huggingface",
@ -1262,68 +1258,6 @@ def run_doctor(args):
issues,
)
# Vercel Sandbox (if using vercel_sandbox backend)
if terminal_env == "vercel_sandbox":
runtime = os.getenv("TERMINAL_VERCEL_RUNTIME", "node24").strip() or "node24"
from tools.terminal_tool import _SUPPORTED_VERCEL_RUNTIMES
if runtime in _SUPPORTED_VERCEL_RUNTIMES:
check_ok("Vercel runtime", f"({runtime})")
else:
supported = ", ".join(_SUPPORTED_VERCEL_RUNTIMES)
_fail_and_issue(
"Vercel runtime unsupported",
f"({runtime}; use {supported})",
f"Set TERMINAL_VERCEL_RUNTIME to one of: {supported}",
issues,
)
disk = os.getenv("TERMINAL_CONTAINER_DISK", "51200").strip()
if disk in {"", "0", "51200"}:
check_ok("Vercel disk setting", "(uses platform default)")
else:
_fail_and_issue(
"Vercel custom disk unsupported",
"(reset terminal.container_disk to 51200)",
"Vercel Sandbox does not support custom container_disk; use the shared default 51200",
issues,
)
if importlib.util.find_spec("vercel") is not None:
check_ok("vercel SDK", "(installed)")
else:
_fail_and_issue(
"vercel SDK not installed",
"(pip install 'hermes-agent[vercel]')",
"Install the Vercel optional dependency: pip install 'hermes-agent[vercel]'",
issues,
)
auth_status = describe_vercel_auth()
if auth_status.ok:
check_ok("Vercel auth", f"({auth_status.label})")
elif auth_status.label.startswith("partial"):
_fail_and_issue(
"Vercel auth incomplete",
f"({auth_status.label})",
"Set VERCEL_TOKEN, VERCEL_PROJECT_ID, and VERCEL_TEAM_ID together",
issues,
)
else:
_fail_and_issue(
"Vercel auth not configured",
f"({auth_status.label})",
"Configure Vercel Sandbox auth with VERCEL_TOKEN, VERCEL_PROJECT_ID, and VERCEL_TEAM_ID",
issues,
)
for line in auth_status.detail_lines:
check_info(f"Vercel auth {line}")
persistent = os.getenv("TERMINAL_CONTAINER_PERSISTENT", "true").lower() in {"1", "true", "yes", "on"}
if persistent:
check_info("Vercel persistence: snapshot filesystem only; live processes do not survive sandbox recreation")
else:
check_info("Vercel persistence: ephemeral filesystem")
# Node.js + agent-browser (for browser automation tools)
if _safe_which("node"):
check_ok("Node.js")

1
hermes_cli/dump.py
View file

@ -279,7 +279,6 @@ def run_dump(args):
("DASHSCOPE_API_KEY", "dashscope"),
("HF_TOKEN", "huggingface"),
("NVIDIA_API_KEY", "nvidia"),
("AI_GATEWAY_API_KEY", "ai_gateway"),
("OPENCODE_ZEN_API_KEY", "opencode_zen"),
("OPENCODE_GO_API_KEY", "opencode_go"),
("KILOCODE_API_KEY", "kilocode"),

55
hermes_cli/main.py
View file

@ -2374,8 +2374,6 @@ def select_provider_and_model(args=None):
# Step 2: Provider-specific setup + model selection
if selected_provider == "openrouter":
_model_flow_openrouter(config, current_model)
elif selected_provider == "ai-gateway":
_model_flow_ai_gateway(config, current_model)
elif selected_provider == "nous":
_model_flow_nous(config, current_model, args=args)
elif selected_provider == "openai-codex":
@ -2962,59 +2960,6 @@ def _model_flow_openrouter(config, current_model=""):
print("No change.")
def _model_flow_ai_gateway(config, current_model=""):
"""Vercel AI Gateway provider: ensure API key, then pick model with pricing."""
from hermes_constants import AI_GATEWAY_BASE_URL
from hermes_cli.auth import (
PROVIDER_REGISTRY,
_prompt_model_selection,
_save_model_choice,
deactivate_provider,
)
from hermes_cli.config import get_env_value
# Route through _prompt_api_key so users can replace a stale/broken key
# in-flow (K/R/C) instead of having to edit ~/.hermes/.env by hand.
pconfig = PROVIDER_REGISTRY["ai-gateway"]
existing_key = get_env_value("AI_GATEWAY_API_KEY") or ""
if not existing_key:
print(
"Create API key here: https://vercel.com/d?to=%2F%5Bteam%5D%2F%7E%2Fai-gateway&title=AI+Gateway"
)
print("Add a payment method to get $5 in free credits.")
print()
_resolved, abort = _prompt_api_key(pconfig, existing_key, provider_id="ai-gateway")
if abort:
return
from hermes_cli.models import ai_gateway_model_ids, get_pricing_for_provider
models_list = ai_gateway_model_ids(force_refresh=True)
pricing = get_pricing_for_provider("ai-gateway", force_refresh=True)
selected = _prompt_model_selection(
models_list, current_model=current_model, pricing=pricing
)
if selected:
_save_model_choice(selected)
from hermes_cli.config import load_config, save_config
cfg = load_config()
model = cfg.get("model")
if not isinstance(model, dict):
model = {"default": model} if model else {}
cfg["model"] = model
model["provider"] = "ai-gateway"
model["base_url"] = AI_GATEWAY_BASE_URL
model["api_mode"] = "chat_completions"
save_config(cfg)
deactivate_provider()
print(f"Default model set to: {selected} (via Vercel AI Gateway)")
else:
print("No change.")
def _model_flow_nous(config, current_model="", args=None):
"""Nous Portal provider: ensure logged in, then pick model."""
from hermes_cli.auth import (

1
hermes_cli/model_normalize.py
View file

@ -67,7 +67,6 @@ _VENDOR_PREFIXES: dict[str, str] = {
_AGGREGATOR_PROVIDERS: frozenset[str] = frozenset({
"openrouter",
"nous",
"ai-gateway",
"kilocode",
})

219
hermes_cli/models.py
View file

@ -69,29 +69,6 @@ OPENROUTER_MODELS: list[tuple[str, str]] = [
_openrouter_catalog_cache: list[tuple[str, str]] | None = None
# Fallback Vercel AI Gateway snapshot used when the live catalog is unavailable.
# OSS / open-weight models prioritized first, then closed-source by family.
# Slugs match Vercel's actual /v1/models catalog (e.g. alibaba/ for Qwen,
# zai/ and xai/ without hyphens).
VERCEL_AI_GATEWAY_MODELS: list[tuple[str, str]] = [
("moonshotai/kimi-k2.6", "recommended"),
("alibaba/qwen3.6-plus", ""),
("zai/glm-5.1", ""),
("minimax/minimax-m2.7", ""),
("anthropic/claude-sonnet-4.6", ""),
("anthropic/claude-opus-4.7", ""),
("anthropic/claude-opus-4.6", ""),
("anthropic/claude-haiku-4.5", ""),
("openai/gpt-5.4", ""),
("openai/gpt-5.4-mini", ""),
("openai/gpt-5.3-codex", ""),
("google/gemini-3.1-pro-preview", ""),
("google/gemini-3-flash", ""),
("google/gemini-3.1-flash-lite-preview", ""),
("xai/grok-4.20-reasoning", ""),
]
_ai_gateway_catalog_cache: list[tuple[str, str]] | None = None
def _codex_curated_models() -> list[str]:
@ -479,12 +456,6 @@ _PROVIDER_MODELS: dict[str, list[str]] = {
],
}
# Vercel AI Gateway: derive the bare-model-id catalog from the curated
# ``VERCEL_AI_GATEWAY_MODELS`` snapshot so both the picker (tuples with descriptions)
# and the static fallback catalog (bare ids) stay in sync from a single
# source of truth.
_PROVIDER_MODELS["ai-gateway"] = [mid for mid, _ in VERCEL_AI_GATEWAY_MODELS]
# ---------------------------------------------------------------------------
# Nous Portal free-model helper
# ---------------------------------------------------------------------------
@ -969,7 +940,6 @@ CANONICAL_PROVIDERS: list[ProviderEntry] = [
ProviderEntry("opencode-go", "OpenCode Go", "OpenCode Go (open models, $10/month subscription)"),
ProviderEntry("bedrock", "AWS Bedrock", "AWS Bedrock (Claude, Nova, Llama, DeepSeek — IAM or API key)"),
ProviderEntry("azure-foundry", "Azure Foundry", "Azure Foundry (OpenAI-style or Anthropic-style endpoint — your Azure AI deployment)"),
ProviderEntry("ai-gateway", "Vercel AI Gateway", "Vercel AI Gateway"),
ProviderEntry("qwen-oauth", "Qwen OAuth (Portal)", "Qwen OAuth (reuses local Qwen CLI login)"),
]
@ -1033,9 +1003,6 @@ _PROVIDER_ALIASES = {
"zen": "opencode-zen",
"go": "opencode-go",
"opencode-go-sub": "opencode-go",
"aigateway": "ai-gateway",
"vercel": "ai-gateway",
"vercel-ai-gateway": "ai-gateway",
"kilo": "kilocode",
"kilo-code": "kilocode",
"kilo-gateway": "kilocode",
@ -1220,95 +1187,6 @@ def get_curated_nous_model_ids() -> list[str]:
return list(_PROVIDER_MODELS.get("nous", []))
def _ai_gateway_model_is_free(pricing: Any) -> bool:
"""Return True if an AI Gateway model has $0 input AND output pricing."""
if not isinstance(pricing, dict):
return False
try:
return float(pricing.get("input", "0")) == 0 and float(pricing.get("output", "0")) == 0
except (TypeError, ValueError):
return False
def fetch_ai_gateway_models(
timeout: float = 8.0,
*,
force_refresh: bool = False,
) -> list[tuple[str, str]]:
"""Return the curated AI Gateway picker list, refreshed from the live catalog when possible."""
global _ai_gateway_catalog_cache
if _ai_gateway_catalog_cache is not None and not force_refresh:
return list(_ai_gateway_catalog_cache)
from hermes_constants import AI_GATEWAY_BASE_URL
fallback = list(VERCEL_AI_GATEWAY_MODELS)
preferred_ids = [mid for mid, _ in fallback]
try:
req = urllib.request.Request(
f"{AI_GATEWAY_BASE_URL.rstrip('/')}/models",
headers={"Accept": "application/json"},
)
with urllib.request.urlopen(req, timeout=timeout) as resp:
payload = json.loads(resp.read().decode())
except Exception:
return list(_ai_gateway_catalog_cache or fallback)
live_items = payload.get("data", [])
if not isinstance(live_items, list):
return list(_ai_gateway_catalog_cache or fallback)
live_by_id: dict[str, dict[str, Any]] = {}
for item in live_items:
if not isinstance(item, dict):
continue
mid = str(item.get("id") or "").strip()
if not mid:
continue
live_by_id[mid] = item
curated: list[tuple[str, str]] = []
for preferred_id in preferred_ids:
live_item = live_by_id.get(preferred_id)
if live_item is None:
continue
desc = "free" if _ai_gateway_model_is_free(live_item.get("pricing")) else ""
curated.append((preferred_id, desc))
if not curated:
return list(_ai_gateway_catalog_cache or fallback)
# If the live catalog offers a free Moonshot model, auto-promote it to
# position #1 as "recommended" — dynamic discovery without a PR.
free_moonshot = next(
(
mid
for mid, item in live_by_id.items()
if mid.startswith("moonshotai/")
and _ai_gateway_model_is_free(item.get("pricing"))
),
None,
)
if free_moonshot:
curated = [(mid, desc) for mid, desc in curated if mid != free_moonshot]
curated.insert(0, (free_moonshot, "recommended"))
else:
first_id, _ = curated[0]
curated[0] = (first_id, "recommended")
_ai_gateway_catalog_cache = curated
return list(curated)
def ai_gateway_model_ids(*, force_refresh: bool = False) -> list[str]:
"""Return just the AI Gateway model-id strings."""
return [mid for mid, _ in fetch_ai_gateway_models(force_refresh=force_refresh)]
# ---------------------------------------------------------------------------
# Pricing helpers — fetch live pricing from OpenRouter-compatible /v1/models
# ---------------------------------------------------------------------------
@ -1454,56 +1332,6 @@ def fetch_models_with_pricing(
return result
def fetch_ai_gateway_pricing(
timeout: float = 8.0,
*,
force_refresh: bool = False,
) -> dict[str, dict[str, str]]:
"""Fetch Vercel AI Gateway /v1/models and return hermes-shaped pricing.
Vercel uses ``input`` / ``output`` field names; hermes's picker expects
``prompt`` / ``completion``. This translates. Cache read/write field names
already match.
"""
from hermes_constants import AI_GATEWAY_BASE_URL
cache_key = AI_GATEWAY_BASE_URL.rstrip("/")
if not force_refresh and cache_key in _pricing_cache:
return _pricing_cache[cache_key]
try:
req = urllib.request.Request(
f"{cache_key}/models",
headers={"Accept": "application/json"},
)
with urllib.request.urlopen(req, timeout=timeout) as resp:
payload = json.loads(resp.read().decode())
except Exception:
_pricing_cache[cache_key] = {}
return {}
result: dict[str, dict[str, str]] = {}
for item in payload.get("data", []):
if not isinstance(item, dict):
continue
mid = item.get("id")
pricing = item.get("pricing")
if not (mid and isinstance(pricing, dict)):
continue
entry: dict[str, str] = {
"prompt": str(pricing.get("input", "")),
"completion": str(pricing.get("output", "")),
}
if pricing.get("input_cache_read"):
entry["input_cache_read"] = str(pricing["input_cache_read"])
if pricing.get("input_cache_write"):
entry["input_cache_write"] = str(pricing["input_cache_write"])
result[mid] = entry
_pricing_cache[cache_key] = result
return result
def _resolve_openrouter_api_key() -> str:
"""Best-effort OpenRouter API key for pricing fetch."""
return os.getenv("OPENROUTER_API_KEY", "").strip()
@ -1535,7 +1363,7 @@ def _resolve_nous_pricing_credentials() -> tuple[str, str]:
def get_pricing_for_provider(provider: str, *, force_refresh: bool = False) -> dict[str, dict[str, str]]:
"""Return live pricing for providers that support it (openrouter, nous, ai-gateway, novita)."""
"""Return live pricing for providers that support it (openrouter, nous, novita)."""
normalized = normalize_provider(provider)
if normalized == "openrouter":
return fetch_models_with_pricing(
@ -1543,8 +1371,6 @@ def get_pricing_for_provider(provider: str, *, force_refresh: bool = False) -> d
base_url="https://openrouter.ai/api",
force_refresh=force_refresh,
)
if normalized == "ai-gateway":
return fetch_ai_gateway_pricing(force_refresh=force_refresh)
if normalized == "novita":
return _fetch_novita_pricing(force_refresh=force_refresh)
if normalized == "nous":
@ -1574,9 +1400,8 @@ def _fetch_novita_pricing(
0.0001 USD. Convert them to the per-token strings used by the shared
pricing formatter.
Results are cached in ``_pricing_cache`` keyed on the resolved base URL,
matching the pattern used by ``fetch_ai_gateway_pricing`` — without this,
every menu render or pricing lookup re-hits the network.
Results are cached in ``_pricing_cache`` keyed on the resolved base URL —
without this, every menu render or pricing lookup re-hits the network.
"""
api_key = os.getenv("NOVITA_API_KEY", "").strip()
if not api_key:
@ -1763,7 +1588,7 @@ def _model_in_provider_catalog(name_lower: str, providers: set[str]) -> bool:
_AGGREGATOR_PROVIDERS = frozenset(
{"nous", "openrouter", "ai-gateway", "copilot", "kilocode"}
{"nous", "openrouter", "copilot", "kilocode"}
)
@ -2110,7 +1935,7 @@ def _resolve_copilot_catalog_api_key() -> str:
# - "nous": curated list and Portal /models endpoint are the source of
# truth for the subscription tier.
# Also excluded: providers that already have dedicated live-endpoint
# branches below (copilot, anthropic, ai-gateway, ollama-cloud, custom,
# branches below (copilot, anthropic, ollama-cloud, custom,
# stepfun, openai-codex) — those paths handle freshness themselves.
_MODELS_DEV_PREFERRED: frozenset[str] = frozenset({
"opencode-go",
@ -2235,10 +2060,6 @@ def provider_model_ids(provider: Optional[str], *, force_refresh: bool = False)
live = _fetch_anthropic_models()
if live:
return live
if normalized == "ai-gateway":
live = _fetch_ai_gateway_models()
if live:
return live
if normalized == "ollama-cloud":
live = fetch_ollama_cloud_models(force_refresh=force_refresh)
if live:
@ -3152,36 +2973,6 @@ def probe_api_models(
}
def _fetch_ai_gateway_models(timeout: float = 5.0) -> Optional[list[str]]:
"""Fetch available language models with tool-use from AI Gateway."""
api_key = os.getenv("AI_GATEWAY_API_KEY", "").strip()
if not api_key:
return None
base_url = os.getenv("AI_GATEWAY_BASE_URL", "").strip()
if not base_url:
from hermes_constants import AI_GATEWAY_BASE_URL
base_url = AI_GATEWAY_BASE_URL
url = base_url.rstrip("/") + "/models"
headers: dict[str, str] = {
"Authorization": f"Bearer {api_key}",
"User-Agent": _HERMES_USER_AGENT,
}
req = urllib.request.Request(url, headers=headers)
try:
with urllib.request.urlopen(req, timeout=timeout) as resp:
data = json.loads(resp.read().decode())
return [
m["id"]
for m in data.get("data", [])
if m.get("id")
and m.get("type") == "language"
and "tool-use" in (m.get("tags") or [])
]
except Exception:
return None
def fetch_api_models(
api_key: Optional[str],
base_url: Optional[str],

9
hermes_cli/providers.py
View file

@ -143,10 +143,6 @@ HERMES_OVERLAYS: Dict[str, HermesOverlay] = {
transport="openai_chat",
base_url_env_var="ALIBABA_CODING_PLAN_BASE_URL",
),
"vercel": HermesOverlay(
transport="openai_chat",
is_aggregator=True,
),
"opencode": HermesOverlay(
transport="openai_chat",
is_aggregator=True,
@ -290,11 +286,6 @@ ALIASES: Dict[str, str] = {
"github": "github-copilot",
"github-copilot-acp": "copilot-acp",
# vercel (models.dev ID for AI Gateway)
"ai-gateway": "vercel",
"aigateway": "vercel",
"vercel-ai-gateway": "vercel",
# opencode (models.dev ID for OpenCode Zen)
"opencode-zen": "opencode",
"zen": "opencode",

140
hermes_cli/setup.py
View file

@ -101,7 +101,6 @@ _DEFAULT_PROVIDER_MODELS = {
"arcee": ["trinity-large-thinking", "trinity-large-preview", "trinity-mini"],
"minimax": ["MiniMax-M2.7", "MiniMax-M2.5", "MiniMax-M2.1", "MiniMax-M2"],
"minimax-cn": ["MiniMax-M2.7", "MiniMax-M2.5", "MiniMax-M2.1", "MiniMax-M2"],
"ai-gateway": ["anthropic/claude-opus-4.6", "anthropic/claude-sonnet-4.6", "openai/gpt-5", "google/gemini-3-flash"],
"kilocode": ["anthropic/claude-opus-4.6", "anthropic/claude-sonnet-4.6", "openai/gpt-5.4", "google/gemini-3-pro-preview", "google/gemini-3-flash-preview"],
"opencode-zen": ["gpt-5.4", "gpt-5.3-codex", "claude-sonnet-4-6", "gemini-3-flash", "glm-5", "kimi-k2.5", "minimax-m2.7"],
"opencode-go": ["kimi-k2.6", "kimi-k2.5", "glm-5.1", "glm-5", "mimo-v2.5-pro", "mimo-v2.5", "mimo-v2-pro", "mimo-v2-omni", "minimax-m2.7", "minimax-m2.5", "qwen3.7-max", "qwen3.6-plus", "qwen3.5-plus"],
@ -679,102 +678,6 @@ def _prompt_container_resources(config: dict):
pass
def _prompt_vercel_sandbox_settings(config: dict):
"""Prompt for Vercel Sandbox settings without exposing unsupported disk sizing."""
terminal = config.setdefault("terminal", {})
print()
print_info("Vercel Sandbox settings:")
print_info(" Filesystem persistence uses Vercel snapshots.")
print_info(" Snapshots restore files only; live processes do not continue after sandbox recreation.")
from tools.terminal_tool import _SUPPORTED_VERCEL_RUNTIMES
current_runtime = terminal.get("vercel_runtime") or "node24"
supported_label = ", ".join(_SUPPORTED_VERCEL_RUNTIMES)
runtime = prompt(f" Runtime ({supported_label})", current_runtime).strip() or current_runtime
if runtime not in _SUPPORTED_VERCEL_RUNTIMES:
print_warning(f"Unsupported Vercel runtime '{runtime}', keeping {current_runtime}.")
runtime = current_runtime if current_runtime in _SUPPORTED_VERCEL_RUNTIMES else "node24"
terminal["vercel_runtime"] = runtime
save_env_value("TERMINAL_VERCEL_RUNTIME", runtime)
current_persist = terminal.get("container_persistent", True)
persist_label = "yes" if current_persist else "no"
terminal["container_persistent"] = prompt(
" Persist filesystem with snapshots? (yes/no)", persist_label
).lower() in {"yes", "true", "y", "1"}
current_cpu = terminal.get("container_cpu", 1)
cpu_str = prompt(" CPU cores", str(current_cpu))
try:
terminal["container_cpu"] = float(cpu_str)
except ValueError:
pass
current_mem = terminal.get("container_memory", 5120)
mem_str = prompt(" Memory in MB (5120 = 5GB)", str(current_mem))
try:
terminal["container_memory"] = int(mem_str)
except ValueError:
pass
if terminal.get("container_disk", 51200) not in {0, 51200}:
print_warning("Vercel Sandbox does not support custom disk sizing; resetting container_disk to 51200.")
terminal["container_disk"] = 51200
print()
print_info("Vercel authentication:")
print_info(" Use a long-lived Vercel access token plus project/team IDs.")
linked_project = _read_nearest_vercel_project()
if linked_project:
print_info(" Found defaults in nearest .vercel/project.json.")
remove_env_value("VERCEL_OIDC_TOKEN")
token = prompt(" Vercel access token", get_env_value("VERCEL_TOKEN") or "", password=True)
project = prompt(
" Vercel project ID",
get_env_value("VERCEL_PROJECT_ID") or linked_project.get("projectId", ""),
)
team = prompt(
" Vercel team ID",
get_env_value("VERCEL_TEAM_ID") or linked_project.get("orgId", ""),
)
if token:
save_env_value("VERCEL_TOKEN", token)
if project:
save_env_value("VERCEL_PROJECT_ID", project)
if team:
save_env_value("VERCEL_TEAM_ID", team)
def _read_nearest_vercel_project(start: Path | None = None) -> dict[str, str]:
"""Read project/team defaults from the nearest Vercel link file."""
current = (start or Path.cwd()).resolve()
if current.is_file():
current = current.parent
for directory in (current, *current.parents):
project_file = directory / ".vercel" / "project.json"
if not project_file.exists():
continue
try:
data = json.loads(project_file.read_text(encoding="utf-8"))
except (OSError, json.JSONDecodeError):
return {}
if not isinstance(data, dict):
return {}
return {
key: value
for key, value in {
"projectId": data.get("projectId"),
"orgId": data.get("orgId"),
}.items()
if isinstance(value, str) and value.strip()
}
return {}
# Tool categories and provider config are now in tools_config.py (shared
# between `hermes tools` and `hermes setup tools`).
@ -936,7 +839,6 @@ def setup_model_provider(config: dict, *, quick: bool = False):
"minimax": "MiniMax",
"minimax-cn": "MiniMax CN",
"anthropic": "Anthropic",
"ai-gateway": "Vercel AI Gateway",
"custom": "your custom endpoint",
}
_prov_display = _prov_names.get(selected_provider, selected_provider or "your provider")
@ -1407,12 +1309,11 @@ def setup_terminal_backend(config: dict):
"Modal - serverless cloud sandbox",
"SSH - run on a remote machine",
"Daytona - persistent cloud development environment",
"Vercel Sandbox - cloud microVM with snapshot filesystem persistence",
]
idx_to_backend = {0: "local", 1: "docker", 2: "modal", 3: "ssh", 4: "daytona", 5: "vercel_sandbox"}
backend_to_idx = {"local": 0, "docker": 1, "modal": 2, "ssh": 3, "daytona": 4, "vercel_sandbox": 5}
idx_to_backend = {0: "local", 1: "docker", 2: "modal", 3: "ssh", 4: "daytona"}
backend_to_idx = {"local": 0, "docker": 1, "modal": 2, "ssh": 3, "daytona": 4}
next_idx = 6
next_idx = 5
if is_linux:
terminal_choices.append("Singularity/Apptainer - HPC-friendly container")
idx_to_backend[next_idx] = "singularity"
@ -1658,39 +1559,6 @@ def setup_terminal_backend(config: dict):
_prompt_container_resources(config)
elif selected_backend == "vercel_sandbox":
print_success("Terminal backend: Vercel Sandbox")
print_info("Cloud microVM sandboxes with snapshot-backed filesystem persistence.")
print_info("Requires the optional SDK: pip install 'hermes-agent[vercel]'")
try:
__import__("vercel")
except ImportError:
print_info("Installing vercel SDK...")
import subprocess
uv_bin = shutil.which("uv")
if uv_bin:
result = subprocess.run(
[uv_bin, "pip", "install", "--python", sys.executable, "vercel"],
capture_output=True,
text=True,
)
else:
result = subprocess.run(
[sys.executable, "-m", "pip", "install", "vercel"],
capture_output=True,
text=True,
)
if result.returncode == 0:
print_success("vercel SDK installed")
else:
print_warning("Install failed — run manually: pip install 'hermes-agent[vercel]'")
if result.stderr:
print_info(f" Error: {result.stderr.strip().splitlines()[-1]}")
_prompt_vercel_sandbox_settings(config)
elif selected_backend == "ssh":
print_success("Terminal backend: SSH")
print_info("Run commands on a remote machine via SSH.")
@ -1744,8 +1612,6 @@ def setup_terminal_backend(config: dict):
save_env_value("TERMINAL_ENV", selected_backend)
if selected_backend == "modal":
save_env_value("TERMINAL_MODAL_MODE", config["terminal"].get("modal_mode", "auto"))
if selected_backend == "vercel_sandbox":
save_env_value("TERMINAL_VERCEL_RUNTIME", config["terminal"].get("vercel_runtime", "node24"))
save_config(config)
print()
print_success(f"Terminal backend set to: {selected_backend}")

18
hermes_cli/status.py
View file

@ -18,7 +18,6 @@ from hermes_cli.config import get_env_path, get_env_value, get_hermes_home, load
from hermes_cli.models import provider_label
from hermes_cli.nous_subscription import get_nous_subscription_features
from hermes_cli.runtime_provider import resolve_requested_provider
from hermes_cli.vercel_auth import describe_vercel_auth
from hermes_constants import OPENROUTER_MODELS_URL
from tools.tool_backend_helpers import managed_nous_tools_enabled
@ -380,23 +379,6 @@ def show_status(args):
elif terminal_env == "daytona":
daytona_image = os.getenv("TERMINAL_DAYTONA_IMAGE", "nikolaik/python-nodejs:python3.11-nodejs20")
print(f" Daytona Image: {daytona_image}")
elif terminal_env == "vercel_sandbox":
runtime = os.getenv("TERMINAL_VERCEL_RUNTIME") or terminal_cfg.get("vercel_runtime") or "node24"
persist = os.getenv("TERMINAL_CONTAINER_PERSISTENT")
if persist is None:
persist_enabled = bool(terminal_cfg.get("container_persistent", True))
else:
persist_enabled = persist.lower() in {"1", "true", "yes", "on"}
auth_status = describe_vercel_auth()
sdk_ok = importlib.util.find_spec("vercel") is not None
sdk_label = "installed" if sdk_ok else "missing (install: pip install 'hermes-agent[vercel]')"
print(f" Runtime: {runtime}")
print(f" SDK: {check_mark(sdk_ok)} {sdk_label}")
print(f" Auth: {check_mark(auth_status.ok)} {auth_status.label}")
for line in auth_status.detail_lines:
print(f" Auth detail: {line}")
print(f" Persistence: {'snapshot filesystem' if persist_enabled else 'ephemeral filesystem'}")
print(" Processes: live processes do not survive cleanup, snapshots, or sandbox recreation")
sudo_password = os.getenv("SUDO_PASSWORD", "")
print(f" Sudo: {check_mark(bool(sudo_password))} {'enabled' if sudo_password else 'disabled'}")

70
hermes_cli/vercel_auth.py
View file

@ -1,70 +0,0 @@
"""Helpers for reporting Vercel Sandbox authentication state."""
from __future__ import annotations
import os
from dataclasses import dataclass
_TOKEN_TUPLE_VARS = ("VERCEL_TOKEN", "VERCEL_PROJECT_ID", "VERCEL_TEAM_ID")
@dataclass(frozen=True)
class VercelAuthStatus:
ok: bool
label: str
detail_lines: tuple[str, ...]
def _present(name: str) -> bool:
return bool(os.getenv(name))
def describe_vercel_auth() -> VercelAuthStatus:
"""Return Vercel auth status without exposing secret values."""
has_oidc = _present("VERCEL_OIDC_TOKEN")
token_states = {name: _present(name) for name in _TOKEN_TUPLE_VARS}
present_token_vars = tuple(name for name, present in token_states.items() if present)
missing_token_vars = tuple(name for name, present in token_states.items() if not present)
if has_oidc:
details = [
"mode: OIDC",
"active env: VERCEL_OIDC_TOKEN",
"note: OIDC tokens are development-only; use access-token auth for deployments and long-running processes",
]
if present_token_vars:
details.append(f"also present: {', '.join(present_token_vars)}")
return VercelAuthStatus(True, "OIDC token via VERCEL_OIDC_TOKEN", tuple(details))
if not missing_token_vars:
return VercelAuthStatus(
True,
"access token + project/team via VERCEL_TOKEN, VERCEL_PROJECT_ID, VERCEL_TEAM_ID",
(
"mode: access token",
"active env: VERCEL_TOKEN, VERCEL_PROJECT_ID, VERCEL_TEAM_ID",
),
)
if present_token_vars:
return VercelAuthStatus(
False,
f"partial access-token auth (missing {', '.join(missing_token_vars)})",
(
"mode: incomplete access token",
f"present env: {', '.join(present_token_vars)}",
f"missing env: {', '.join(missing_token_vars)}",
"recommended: set VERCEL_TOKEN, VERCEL_PROJECT_ID, and VERCEL_TEAM_ID together",
),
)
return VercelAuthStatus(
False,
"not configured",
(
"recommended: set VERCEL_TOKEN, VERCEL_PROJECT_ID, and VERCEL_TEAM_ID",
"development-only alternative: set VERCEL_OIDC_TOKEN",
),
)

7
hermes_cli/web_server.py
View file

@ -266,12 +266,7 @@ _SCHEMA_OVERRIDES: Dict[str, Dict[str, Any]] = {
"terminal.backend": {
"type": "select",
"description": "Terminal execution backend",
"options": ["local", "docker", "ssh", "modal", "daytona", "vercel_sandbox", "singularity"],
},
"terminal.vercel_runtime": {
"type": "select",
"description": "Vercel Sandbox runtime",
"options": ["node24", "node22", "python3.13"], # sync with _SUPPORTED_VERCEL_RUNTIMES in terminal_tool.py
"options": ["local", "docker", "ssh", "modal", "daytona", "singularity"],
},
"terminal.modal_mode": {
"type": "select",

2
hermes_constants.py
View file

@ -461,5 +461,3 @@ FINISH_REASON_LENGTH = "length"
OPENROUTER_BASE_URL = "https://openrouter.ai/api/v1"
OPENROUTER_MODELS_URL = f"{OPENROUTER_BASE_URL}/models"
AI_GATEWAY_BASE_URL = "https://ai-gateway.vercel.sh/v1"

43
plugins/model-providers/ai-gateway/__init__.py
View file

@ -1,43 +0,0 @@
"""Vercel AI Gateway provider profile.
AI Gateway routes to multiple backends. Hermes sends attribution
headers and full reasoning config passthrough.
"""
from typing import Any
from providers import register_provider
from providers.base import ProviderProfile
class VercelAIGatewayProfile(ProviderProfile):
"""Vercel AI Gateway — attribution headers + reasoning passthrough."""
def build_api_kwargs_extras(
self,
*,
reasoning_config: dict | None = None,
supports_reasoning: bool = True,
**ctx: Any,
) -> tuple[dict[str, Any], dict[str, Any]]:
extra_body: dict[str, Any] = {}
if supports_reasoning and reasoning_config is not None:
extra_body["reasoning"] = dict(reasoning_config)
elif supports_reasoning:
extra_body["reasoning"] = {"enabled": True, "effort": "medium"}
return extra_body, {}
vercel = VercelAIGatewayProfile(
name="ai-gateway",
aliases=("vercel", "vercel-ai-gateway", "ai_gateway", "aigateway"),
env_vars=("AI_GATEWAY_API_KEY",),
base_url="https://ai-gateway.vercel.sh/v1",
default_headers={
"HTTP-Referer": "https://hermes-agent.nousresearch.com",
"X-Title": "Hermes Agent",
},
default_aux_model="google/gemini-3-flash",
)
register_provider(vercel)

5
plugins/model-providers/ai-gateway/plugin.yaml
View file

@ -1,5 +0,0 @@
name: ai-gateway-provider
kind: model-provider
version: 1.0.0
description: Vercel AI Gateway
author: Nous Research

3
pyproject.toml
View file

@ -82,7 +82,6 @@ fal = ["fal-client==0.13.1"]
edge-tts = ["edge-tts==7.2.7"]
modal = ["modal==1.3.4"]
daytona = ["daytona==0.155.0"]
vercel = ["vercel==0.5.7"]
hindsight = ["hindsight-client==0.6.1"]
dev = ["debugpy==1.8.20", "pytest==9.0.2", "pytest-asyncio==1.3.0", "pytest-timeout==2.4.0", "mcp==1.26.0", "ty==0.0.21", "ruff==0.15.10"]
messaging = ["python-telegram-bot[webhooks]==22.6", "discord.py[voice]==2.7.1", "aiohttp==3.13.3", "brotlicffi==1.2.0.1", "slack-bolt==1.27.0", "slack-sdk==3.40.1", "qrcode==7.4.2"]
@ -189,7 +188,7 @@ all = [
#
# Removed from [all] on 2026-05-12 (covered by lazy-install):
# anthropic, exa, firecrawl, parallel-web, fal, edge-tts,
# modal, daytona, vercel, messaging (telegram/discord/slack),
# modal, daytona, messaging (telegram/discord/slack),
# matrix, slack, honcho, voice (faster-whisper),
# dingtalk, feishu, bedrock, tts-premium (elevenlabs)
#

5
run_agent.py
View file

@ -2975,15 +2975,12 @@ class AIAgent:
def _apply_client_headers_for_base_url(self, base_url: str) -> None:
from agent.auxiliary_client import (
_AI_GATEWAY_HEADERS,
build_nvidia_nim_headers,
build_or_headers,
)
if base_url_host_matches(base_url, "openrouter.ai"):
self._client_kwargs["default_headers"] = build_or_headers()
elif base_url_host_matches(base_url, "ai-gateway.vercel.sh"):
self._client_kwargs["default_headers"] = dict(_AI_GATEWAY_HEADERS)
elif base_url_host_matches(base_url, "integrate.api.nvidia.com"):
self._client_kwargs["default_headers"] = build_nvidia_nim_headers(base_url)
elif base_url_host_matches(base_url, "api.routermint.com"):
@ -3788,8 +3785,6 @@ class AIAgent:
"""
if base_url_host_matches(self._base_url_lower, "nousresearch.com"):
return True
if base_url_host_matches(self._base_url_lower, "ai-gateway.vercel.sh"):
return True
if (
base_url_host_matches(self._base_url_lower, "models.github.ai")
or base_url_host_matches(self._base_url_lower, "api.githubcopilot.com")

2
setup-hermes.sh
View file

@ -214,7 +214,7 @@ else
# if mistral can't resolve.
_BROKEN_EXTRAS=() # populate when an extra becomes unresolvable
_ALL_EXTRAS=(
modal daytona vercel messaging matrix cron cli dev tts-premium slack
modal daytona messaging matrix cron cli dev tts-premium slack
pty honcho mcp homeassistant sms acp voice dingtalk feishu google
bedrock web youtube
)

3
skills/autonomous-ai-agents/hermes-agent/SKILL.md
View file

@ -389,7 +389,6 @@ Full config reference: https://hermes-agent.nousresearch.com/docs/user-guide/con
| Alibaba / DashScope | API key | `DASHSCOPE_API_KEY` |
| Xiaomi MiMo | API key | `XIAOMI_API_KEY` |
| Kilo Code | API key | `KILOCODE_API_KEY` |
| AI Gateway (Vercel) | API key | `AI_GATEWAY_API_KEY` |
| OpenCode Zen | API key | `OPENCODE_ZEN_API_KEY` |
| OpenCode Go | API key | `OPENCODE_GO_API_KEY` |
| Qwen OAuth | OAuth | `hermes auth add qwen-oauth` |
@ -995,7 +994,7 @@ See `tests/agent/test_prompt_builder.py::TestEnvironmentHints` for a worked exam
Factual guidance about the host OS, user home, cwd, terminal backend, and shell (bash vs. PowerShell on Windows) is emitted from `agent/prompt_builder.py::build_environment_hints()`. This is also where the WSL hint and per-backend probe logic live. The convention:
- **Local terminal backend** → emit host info (OS, `$HOME`, cwd) + Windows-specific notes (hostname ≠ username, `terminal` uses bash not PowerShell).
- **Remote terminal backend** (anything in `_REMOTE_TERMINAL_BACKENDS`: `docker, singularity, modal, daytona, ssh, vercel_sandbox, managed_modal`) → **suppress** host info entirely and describe only the backend. A live `uname`/`whoami`/`pwd` probe runs inside the backend via `tools.environments.get_environment(...).execute(...)`, cached per process in `_BACKEND_PROBE_CACHE`, with a static fallback if the probe times out.
- **Remote terminal backend** (anything in `_REMOTE_TERMINAL_BACKENDS`: `docker, singularity, modal, daytona, ssh, managed_modal`) → **suppress** host info entirely and describe only the backend. A live `uname`/`whoami`/`pwd` probe runs inside the backend via `tools.environments.get_environment(...).execute(...)`, cached per process in `_BACKEND_PROBE_CACHE`, with a static fallback if the probe times out.
- **Key fact for prompt authoring:** when `TERMINAL_ENV != "local"`, *every* file tool (`read_file`, `write_file`, `patch`, `search_files`) runs inside the backend container, not on the host. The system prompt must never describe the host in that case — the agent can't touch it.
Full design notes, the exact emitted strings, and testing pitfalls:

1
tests/agent/test_models_dev.py
View file

@ -94,7 +94,6 @@ class TestProviderMapping:
assert PROVIDER_TO_MODELS_DEV["copilot"] == "github-copilot"
assert PROVIDER_TO_MODELS_DEV["stepfun"] == "stepfun"
assert PROVIDER_TO_MODELS_DEV["kilocode"] == "kilo"
assert PROVIDER_TO_MODELS_DEV["ai-gateway"] == "vercel"
def test_xai_oauth_uses_xai_catalog(self):
assert PROVIDER_TO_MODELS_DEV["xai"] == "xai"

2
tests/agent/test_prompt_builder.py
View file

@ -942,7 +942,7 @@ class TestEnvironmentHints:
def test_remote_backend_list_covers_known_sandboxes(self):
"""Regression guard: if someone adds a remote backend, they must list it here."""
import agent.prompt_builder as _pb
for backend in ("docker", "singularity", "modal", "daytona", "ssh", "vercel_sandbox"):
for backend in ("docker", "singularity", "modal", "daytona", "ssh"):
assert backend in _pb._REMOTE_TERMINAL_BACKENDS, (
f"{backend!r} must be in _REMOTE_TERMINAL_BACKENDS so its host "
f"info is suppressed in the system prompt"

2
tests/agent/test_usage_pricing.py
View file

@ -40,7 +40,7 @@ def test_normalize_usage_openai_subtracts_cached_prompt_tokens():
def test_normalize_usage_openai_reads_top_level_anthropic_cache_fields():
"""Some OpenAI-compatible proxies (OpenRouter, Vercel AI Gateway, Cline) expose
"""Some OpenAI-compatible proxies (OpenRouter, Cline) expose
Anthropic-style cache token counts at the top level of the usage object when
routing Claude models, instead of nesting them in prompt_tokens_details.

24
tests/cli/test_cli_init.py
View file

@ -544,30 +544,6 @@ class TestRootLevelProviderOverride:
assert cfg["model"]["base_url"] == "https://example.com/v1"
def test_terminal_vercel_runtime_bridged_to_env(self, tmp_path, monkeypatch):
"""Classic CLI must expose terminal.vercel_runtime to terminal_tool.py."""
import yaml
hermes_home = tmp_path / ".hermes"
hermes_home.mkdir()
monkeypatch.setenv("HERMES_HOME", str(hermes_home))
monkeypatch.delenv("TERMINAL_VERCEL_RUNTIME", raising=False)
config_path = hermes_home / "config.yaml"
config_path.write_text(yaml.safe_dump({
"terminal": {
"backend": "vercel_sandbox",
"vercel_runtime": "python3.13",
},
}))
import cli
monkeypatch.setattr(cli, "_hermes_home", hermes_home)
cfg = cli.load_cli_config()
assert cfg["terminal"]["vercel_runtime"] == "python3.13"
assert os.environ["TERMINAL_VERCEL_RUNTIME"] == "python3.13"
def test_normalize_root_model_keys_moves_to_model(self):
"""_normalize_root_model_keys migrates root keys into model section."""
from hermes_cli.config import _normalize_root_model_keys

3
tests/conftest.py
View file

@ -147,7 +147,6 @@ _CREDENTIAL_NAMES = frozenset({
"TOOL_GATEWAY_USER_TOKEN",
"TELEGRAM_WEBHOOK_SECRET",
"WEBHOOK_SECRET",
"AI_GATEWAY_API_KEY",
"VOICE_TOOLS_OPENAI_KEY",
"BROWSER_USE_API_KEY",
"CUSTOM_API_KEY",
@ -158,7 +157,6 @@ _CREDENTIAL_NAMES = frozenset({
"OLLAMA_BASE_URL",
"GROQ_BASE_URL",
"XAI_BASE_URL",
"AI_GATEWAY_BASE_URL",
"ANTHROPIC_BASE_URL",
})
@ -217,7 +215,6 @@ _HERMES_BEHAVIORAL_VARS = frozenset({
"HERMES_TENANT",
"TERMINAL_CWD",
"TERMINAL_ENV",
"TERMINAL_VERCEL_RUNTIME",
"TERMINAL_CONTAINER_CPU",
"TERMINAL_CONTAINER_DISK",
"TERMINAL_CONTAINER_MEMORY",

22
tests/gateway/test_config_cwd_bridge.py
View file

@ -33,7 +33,6 @@ def _simulate_config_bridge(cfg: dict, initial_env: dict | None = None):
"backend": "TERMINAL_ENV",
"cwd": "TERMINAL_CWD",
"timeout": "TERMINAL_TIMEOUT",
"vercel_runtime": "TERMINAL_VERCEL_RUNTIME",
"container_persistent": "TERMINAL_CONTAINER_PERSISTENT",
"container_cpu": "TERMINAL_CONTAINER_CPU",
"container_memory": "TERMINAL_CONTAINER_MEMORY",
@ -245,24 +244,3 @@ class TestTildeExpansion:
}
result = _simulate_config_bridge(cfg)
assert result["TERMINAL_CWD"] == os.path.expanduser("~/nested")
class TestVercelTerminalBridge:
def test_vercel_terminal_settings_bridge(self):
cfg = {
"terminal": {
"backend": "vercel_sandbox",
"vercel_runtime": "python3.13",
"container_persistent": True,
"container_cpu": 2,
"container_memory": 4096,
"container_disk": 51200,
}
}
result = _simulate_config_bridge(cfg, {"MESSAGING_CWD": "/from/env"})
assert result["TERMINAL_ENV"] == "vercel_sandbox"
assert result["TERMINAL_VERCEL_RUNTIME"] == "python3.13"
assert result["TERMINAL_CONTAINER_PERSISTENT"] == "True"
assert result["TERMINAL_CONTAINER_CPU"] == "2"
assert result["TERMINAL_CONTAINER_MEMORY"] == "4096"
assert result["TERMINAL_CONTAINER_DISK"] == "51200"

161
tests/hermes_cli/test_ai_gateway_models.py
View file

@ -1,161 +0,0 @@
"""AI Gateway model list and pricing translation.
Vercel AI Gateway exposes ``/v1/models`` with a richer shape than OpenAI's
spec (type, tags, pricing). The pricing object uses ``input`` / ``output``
where hermes's shared picker expects ``prompt`` / ``completion``; these tests
pin the translation and the curated-list filtering.
"""
import json
from unittest.mock import patch, MagicMock
from hermes_cli import models as models_module
from hermes_cli.models import (
VERCEL_AI_GATEWAY_MODELS,
_ai_gateway_model_is_free,
fetch_ai_gateway_models,
fetch_ai_gateway_pricing,
)
def _mock_urlopen(payload):
"""Build a urlopen() context manager mock returning the given payload."""
resp = MagicMock()
resp.read.return_value = json.dumps(payload).encode()
ctx = MagicMock()
ctx.__enter__.return_value = resp
ctx.__exit__.return_value = False
return ctx
def _reset_caches():
models_module._ai_gateway_catalog_cache = None
models_module._pricing_cache.clear()
def test_ai_gateway_pricing_translates_input_output_to_prompt_completion():
_reset_caches()
payload = {
"data": [
{
"id": "moonshotai/kimi-k2.5",
"type": "language",
"pricing": {
"input": "0.0000006",
"output": "0.0000025",
"input_cache_read": "0.00000015",
"input_cache_write": "0.0000006",
},
}
]
}
with patch("urllib.request.urlopen", return_value=_mock_urlopen(payload)):
result = fetch_ai_gateway_pricing(force_refresh=True)
entry = result["moonshotai/kimi-k2.5"]
assert entry["prompt"] == "0.0000006"
assert entry["completion"] == "0.0000025"
assert entry["input_cache_read"] == "0.00000015"
assert entry["input_cache_write"] == "0.0000006"
def test_ai_gateway_pricing_returns_empty_on_fetch_failure():
_reset_caches()
with patch("urllib.request.urlopen", side_effect=OSError("network down")):
result = fetch_ai_gateway_pricing(force_refresh=True)
assert result == {}
def test_ai_gateway_pricing_skips_entries_without_pricing_dict():
_reset_caches()
payload = {
"data": [
{"id": "x/y", "pricing": None},
{"id": "a/b", "pricing": {"input": "0", "output": "0"}},
]
}
with patch("urllib.request.urlopen", return_value=_mock_urlopen(payload)):
result = fetch_ai_gateway_pricing(force_refresh=True)
assert "x/y" not in result
assert result["a/b"] == {"prompt": "0", "completion": "0"}
def test_ai_gateway_free_detector():
assert _ai_gateway_model_is_free({"input": "0", "output": "0"}) is True
assert _ai_gateway_model_is_free({"input": "0", "output": "0.01"}) is False
assert _ai_gateway_model_is_free({"input": "0.01", "output": "0"}) is False
assert _ai_gateway_model_is_free(None) is False
assert _ai_gateway_model_is_free({"input": "not a number"}) is False
def test_fetch_ai_gateway_models_filters_against_live_catalog():
_reset_caches()
preferred = [mid for mid, _ in VERCEL_AI_GATEWAY_MODELS]
live_ids = preferred[:3] # only first three exist live
payload = {
"data": [
{"id": mid, "pricing": {"input": "0.001", "output": "0.002"}}
for mid in live_ids
]
}
with patch("urllib.request.urlopen", return_value=_mock_urlopen(payload)):
result = fetch_ai_gateway_models(force_refresh=True)
assert [mid for mid, _ in result] == live_ids
assert result[0][1] == "recommended"
def test_fetch_ai_gateway_models_tags_free_models():
_reset_caches()
first_id = VERCEL_AI_GATEWAY_MODELS[0][0]
second_id = VERCEL_AI_GATEWAY_MODELS[1][0]
payload = {
"data": [
{"id": first_id, "pricing": {"input": "0.001", "output": "0.002"}},
{"id": second_id, "pricing": {"input": "0", "output": "0"}},
]
}
with patch("urllib.request.urlopen", return_value=_mock_urlopen(payload)):
result = fetch_ai_gateway_models(force_refresh=True)
by_id = dict(result)
assert by_id[first_id] == "recommended"
assert by_id[second_id] == "free"
def test_free_moonshot_model_auto_promoted_to_top_even_if_not_curated():
_reset_caches()
first_curated = VERCEL_AI_GATEWAY_MODELS[0][0]
unlisted_free_moonshot = "moonshotai/kimi-coder-free-preview"
payload = {
"data": [
{"id": first_curated, "pricing": {"input": "0.001", "output": "0.002"}},
{"id": unlisted_free_moonshot, "pricing": {"input": "0", "output": "0"}},
]
}
with patch("urllib.request.urlopen", return_value=_mock_urlopen(payload)):
result = fetch_ai_gateway_models(force_refresh=True)
assert result[0] == (unlisted_free_moonshot, "recommended")
assert any(mid == first_curated for mid, _ in result)
def test_paid_moonshot_does_not_get_auto_promoted():
_reset_caches()
first_curated = VERCEL_AI_GATEWAY_MODELS[0][0]
payload = {
"data": [
{"id": first_curated, "pricing": {"input": "0.001", "output": "0.002"}},
{"id": "moonshotai/some-paid-variant", "pricing": {"input": "0.001", "output": "0.002"}},
]
}
with patch("urllib.request.urlopen", return_value=_mock_urlopen(payload)):
result = fetch_ai_gateway_models(force_refresh=True)
assert result[0][0] == first_curated
def test_fetch_ai_gateway_models_falls_back_on_error():
_reset_caches()
with patch("urllib.request.urlopen", side_effect=OSError("network")):
result = fetch_ai_gateway_models(force_refresh=True)
assert result == list(VERCEL_AI_GATEWAY_MODELS)

39
tests/hermes_cli/test_api_key_providers.py
View file

@ -1,4 +1,4 @@
"""Tests for API-key provider support (z.ai/GLM, Kimi, MiniMax, AI Gateway)."""
"""Tests for API-key provider support (z.ai/GLM, Kimi, MiniMax)."""
import os
@ -40,7 +40,6 @@ class TestProviderRegistry:
("stepfun", "StepFun Step Plan", "api_key"),
("minimax", "MiniMax", "api_key"),
("minimax-cn", "MiniMax (China)", "api_key"),
("ai-gateway", "Vercel AI Gateway", "api_key"),
("kilocode", "Kilo Code", "api_key"),
("gmi", "GMI Cloud", "api_key"),
])
@ -97,11 +96,6 @@ class TestProviderRegistry:
assert pconfig.api_key_env_vars == ("MINIMAX_CN_API_KEY",)
assert pconfig.base_url_env_var == "MINIMAX_CN_BASE_URL"
def test_ai_gateway_env_vars(self):
pconfig = PROVIDER_REGISTRY["ai-gateway"]
assert pconfig.api_key_env_vars == ("AI_GATEWAY_API_KEY",)
assert pconfig.base_url_env_var == "AI_GATEWAY_BASE_URL"
def test_kilocode_env_vars(self):
pconfig = PROVIDER_REGISTRY["kilocode"]
assert pconfig.api_key_env_vars == ("KILOCODE_API_KEY",)
@ -125,7 +119,6 @@ class TestProviderRegistry:
assert PROVIDER_REGISTRY["stepfun"].inference_base_url == STEPFUN_STEP_PLAN_INTL_BASE_URL
assert PROVIDER_REGISTRY["minimax"].inference_base_url == "https://api.minimax.io/anthropic"
assert PROVIDER_REGISTRY["minimax-cn"].inference_base_url == "https://api.minimaxi.com/anthropic"
assert PROVIDER_REGISTRY["ai-gateway"].inference_base_url == "https://ai-gateway.vercel.sh/v1"
assert PROVIDER_REGISTRY["kilocode"].inference_base_url == "https://api.kilo.ai/api/gateway"
assert PROVIDER_REGISTRY["gmi"].inference_base_url == "https://api.gmi-serving.com/v1"
assert PROVIDER_REGISTRY["huggingface"].inference_base_url == "https://router.huggingface.co/v1"
@ -149,7 +142,6 @@ PROVIDER_ENV_VARS = (
"GLM_API_KEY", "ZAI_API_KEY", "Z_AI_API_KEY",
"KIMI_API_KEY", "KIMI_BASE_URL", "STEPFUN_API_KEY", "STEPFUN_BASE_URL",
"MINIMAX_API_KEY", "MINIMAX_CN_API_KEY",
"AI_GATEWAY_API_KEY", "AI_GATEWAY_BASE_URL",
"KILOCODE_API_KEY", "KILOCODE_BASE_URL",
"GMI_API_KEY", "GMI_BASE_URL",
"DASHSCOPE_API_KEY", "OPENCODE_ZEN_API_KEY", "OPENCODE_GO_API_KEY",
@ -184,9 +176,6 @@ class TestResolveProvider:
def test_explicit_minimax_cn(self):
assert resolve_provider("minimax-cn") == "minimax-cn"
def test_explicit_ai_gateway(self):
assert resolve_provider("ai-gateway") == "ai-gateway"
def test_explicit_gmi(self):
assert resolve_provider("gmi") == "gmi"
@ -211,12 +200,6 @@ class TestResolveProvider:
def test_alias_minimax_underscore(self):
assert resolve_provider("minimax_cn") == "minimax-cn"
def test_alias_aigateway(self):
assert resolve_provider("aigateway") == "ai-gateway"
def test_alias_vercel(self):
assert resolve_provider("vercel") == "ai-gateway"
def test_alias_gmi_cloud(self):
assert resolve_provider("gmi-cloud") == "gmi"
@ -291,10 +274,6 @@ class TestResolveProvider:
monkeypatch.setenv("MINIMAX_CN_API_KEY", "test-mm-cn-key")
assert resolve_provider("auto") == "minimax-cn"
def test_auto_detects_ai_gateway_key(self, monkeypatch):
monkeypatch.setenv("AI_GATEWAY_API_KEY", "test-gw-key")
assert resolve_provider("auto") == "ai-gateway"
def test_auto_detects_gmi_key(self, monkeypatch):
monkeypatch.setenv("GMI_API_KEY", "test-gmi-key")
assert resolve_provider("auto") == "gmi"
@ -535,13 +514,6 @@ class TestResolveApiKeyProviderCredentials:
assert creds["api_key"] == "mmcn-secret-key"
assert creds["base_url"] == "https://api.minimaxi.com/anthropic"
def test_resolve_ai_gateway_with_key(self, monkeypatch):
monkeypatch.setenv("AI_GATEWAY_API_KEY", "gw-secret-key")
creds = resolve_api_key_provider_credentials("ai-gateway")
assert creds["provider"] == "ai-gateway"
assert creds["api_key"] == "gw-secret-key"
assert creds["base_url"] == "https://ai-gateway.vercel.sh/v1"
def test_resolve_kilocode_with_key(self, monkeypatch):
monkeypatch.setenv("KILOCODE_API_KEY", "kilo-secret-key")
creds = resolve_api_key_provider_credentials("kilocode")
@ -641,15 +613,6 @@ class TestRuntimeProviderResolution:
assert result["provider"] == "minimax"
assert result["api_key"] == "mm-key"
def test_runtime_ai_gateway(self, monkeypatch):
monkeypatch.setenv("AI_GATEWAY_API_KEY", "gw-key")
from hermes_cli.runtime_provider import resolve_runtime_provider
result = resolve_runtime_provider(requested="ai-gateway")
assert result["provider"] == "ai-gateway"
assert result["api_mode"] == "chat_completions"
assert result["api_key"] == "gw-key"
assert "ai-gateway.vercel.sh" in result["base_url"]
def test_runtime_kilocode(self, monkeypatch):
monkeypatch.setenv("KILOCODE_API_KEY", "kilo-key")
from hermes_cli.runtime_provider import resolve_runtime_provider

2
tests/hermes_cli/test_arcee_provider.py
View file

@ -16,7 +16,7 @@ _OTHER_PROVIDER_KEYS = (
"OPENAI_API_KEY", "ANTHROPIC_API_KEY", "DEEPSEEK_API_KEY",
"GOOGLE_API_KEY", "GEMINI_API_KEY", "DASHSCOPE_API_KEY",
"XAI_API_KEY", "KIMI_API_KEY", "KIMI_CN_API_KEY",
"MINIMAX_API_KEY", "MINIMAX_CN_API_KEY", "AI_GATEWAY_API_KEY",
"MINIMAX_API_KEY", "MINIMAX_CN_API_KEY",
"KILOCODE_API_KEY", "HF_TOKEN", "GLM_API_KEY", "ZAI_API_KEY",
"XIAOMI_API_KEY", "TOKENHUB_API_KEY", "COPILOT_GITHUB_TOKEN", "GH_TOKEN", "GITHUB_TOKEN",
)

35
tests/hermes_cli/test_doctor.py
View file

@ -253,38 +253,6 @@ def test_check_gateway_service_linger_skips_when_service_not_installed(monkeypat
assert issues == []
def test_doctor_reports_vercel_backend_diagnostics(monkeypatch, tmp_path):
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("TERMINAL_VERCEL_RUNTIME", "python3.13")
monkeypatch.setenv("TERMINAL_CONTAINER_DISK", "2048")
monkeypatch.setenv("VERCEL_TOKEN", "super-secret-value")
monkeypatch.delenv("VERCEL_PROJECT_ID", raising=False)
monkeypatch.setenv("VERCEL_TEAM_ID", "team")
monkeypatch.setattr(doctor_mod.importlib.util, "find_spec", lambda name: object() if name == "vercel" else None)
fake_model_tools = types.SimpleNamespace(
check_tool_availability=lambda *a, **kw: ([], []),
TOOLSET_REQUIREMENTS={},
)
monkeypatch.setitem(sys.modules, "model_tools", fake_model_tools)
buf = io.StringIO()
with contextlib.redirect_stdout(buf):
doctor_mod.run_doctor(Namespace(fix=False))
out = buf.getvalue()
assert "Vercel runtime" in out
assert "python3.13" in out
assert "Vercel custom disk unsupported" in out
assert "Vercel auth incomplete" in out
assert "VERCEL_PROJECT_ID" in out
assert "Vercel auth mode: incomplete access token" in out
assert "Vercel auth present env: VERCEL_TOKEN, VERCEL_TEAM_ID" in out
assert "Vercel auth missing env: VERCEL_PROJECT_ID" in out
assert "super-secret-value" not in out
assert "snapshot filesystem only" in out
# ── Memory provider section (doctor should only check the *active* provider) ──
@ -522,7 +490,6 @@ def test_run_doctor_flags_missing_credentials_for_active_openrouter_provider(mon
@pytest.mark.parametrize(
("provider", "default_model"),
[
("ai-gateway", "anthropic/claude-sonnet-4.6"),
("opencode-zen", "anthropic/claude-sonnet-4.6"),
("kilocode", "anthropic/claude-sonnet-4.6"),
("kimi-coding", "kimi-k2"),
@ -566,7 +533,7 @@ def test_run_doctor_accepts_hermes_provider_ids_that_catalog_aliases(
out = buf.getvalue()
assert f"model.provider '{provider}' is not a recognised provider" not in out
assert f"model.provider '{provider}' is unknown" not in out
if provider in {"ai-gateway", "opencode-zen", "kilocode"}:
if provider in {"opencode-zen", "kilocode"}:
assert (
f"model.default '{default_model}' uses a vendor/model slug but provider is '{provider}'"
not in out

1
tests/hermes_cli/test_gmi_provider.py
View file

@ -183,7 +183,6 @@ class TestGmiDoctor:
"DASHSCOPE_API_KEY",
"MINIMAX_API_KEY",
"MINIMAX_CN_API_KEY",
"AI_GATEWAY_API_KEY",
"KILOCODE_API_KEY",
"OPENCODE_ZEN_API_KEY",
"OPENCODE_GO_API_KEY",

44
tests/hermes_cli/test_runtime_provider_resolution.py
View file

@ -226,20 +226,6 @@ def test_qwen_oauth_auto_fallthrough_on_auth_failure(monkeypatch):
assert resolved["provider"] != "qwen-oauth"
def test_resolve_runtime_provider_ai_gateway(monkeypatch):
monkeypatch.setattr(rp, "resolve_provider", lambda *a, **k: "ai-gateway")
monkeypatch.setattr(rp, "_get_model_config", lambda: {})
monkeypatch.setenv("AI_GATEWAY_API_KEY", "test-ai-gw-key")
resolved = rp.resolve_runtime_provider(requested="ai-gateway")
assert resolved["provider"] == "ai-gateway"
assert resolved["api_mode"] == "chat_completions"
assert resolved["base_url"] == "https://ai-gateway.vercel.sh/v1"
assert resolved["api_key"] == "test-ai-gw-key"
assert resolved["requested_provider"] == "ai-gateway"
def test_resolve_runtime_provider_lmstudio_uses_token_when_present(monkeypatch):
monkeypatch.setattr(rp, "resolve_provider", lambda *a, **k: "lmstudio")
monkeypatch.setattr(
@ -351,36 +337,6 @@ def test_resolve_runtime_provider_lmstudio_saved_base_url_wins_over_env(monkeypa
assert resolved["api_key"] == "dummy-lm-api-key"
def test_resolve_runtime_provider_ai_gateway_explicit_override_skips_pool(monkeypatch):
def _unexpected_pool(provider):
raise AssertionError(f"load_pool should not be called for {provider}")
def _unexpected_provider_resolution(provider):
raise AssertionError(f"resolve_api_key_provider_credentials should not be called for {provider}")
monkeypatch.setattr(rp, "resolve_provider", lambda *a, **k: "ai-gateway")
monkeypatch.setattr(rp, "_get_model_config", lambda: {})
monkeypatch.setattr(rp, "load_pool", _unexpected_pool)
monkeypatch.setattr(
rp,
"resolve_api_key_provider_credentials",
_unexpected_provider_resolution,
)
resolved = rp.resolve_runtime_provider(
requested="ai-gateway",
explicit_api_key="ai-gateway-explicit-token",
explicit_base_url="https://proxy.example.com/v1/",
)
assert resolved["provider"] == "ai-gateway"
assert resolved["api_mode"] == "chat_completions"
assert resolved["api_key"] == "ai-gateway-explicit-token"
assert resolved["base_url"] == "https://proxy.example.com/v1"
assert resolved["source"] == "explicit"
assert resolved.get("credential_pool") is None
def test_resolve_runtime_provider_openrouter_explicit(monkeypatch):
monkeypatch.setattr(rp, "resolve_provider", lambda *a, **k: "openrouter")
monkeypatch.setattr(rp, "_get_model_config", lambda: {})

7
tests/hermes_cli/test_set_config_value.py
View file

@ -125,13 +125,6 @@ class TestConfigYamlRouting:
or "TERMINAL_DOCKER_MOUNT_CWD_TO_WORKSPACE=True" in env_content
)
def test_terminal_vercel_runtime_goes_to_config_and_env(self, _isolated_hermes_home):
set_config_value("terminal.vercel_runtime", "python3.13")
config = _read_config(_isolated_hermes_home)
env_content = _read_env(_isolated_hermes_home)
assert "vercel_runtime: python3.13" in config
assert "TERMINAL_VERCEL_RUNTIME=python3.13" in env_content
# ---------------------------------------------------------------------------
# Empty / falsy values — regression tests for #4277

90
tests/hermes_cli/test_setup.py
View file

@ -30,17 +30,6 @@ def _clear_provider_env(monkeypatch):
monkeypatch.delenv(key, raising=False)
def _clear_vercel_env(monkeypatch):
for key in (
"TERMINAL_VERCEL_RUNTIME",
"VERCEL_OIDC_TOKEN",
"VERCEL_TOKEN",
"VERCEL_PROJECT_ID",
"VERCEL_TEAM_ID",
):
monkeypatch.delenv(key, raising=False)
def _stub_tts(monkeypatch):
"""Stub out TTS prompts so setup_model_provider doesn't block."""
monkeypatch.setattr("hermes_cli.setup.prompt_choice", lambda q, c, d=0: (
@ -494,85 +483,6 @@ def test_modal_setup_persists_direct_mode_when_user_chooses_their_own_account(tm
assert config["terminal"]["modal_mode"] == "direct"
def test_vercel_setup_configures_access_token_auth(tmp_path, monkeypatch):
monkeypatch.setenv("HERMES_HOME", str(tmp_path))
_clear_vercel_env(monkeypatch)
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "old-oidc")
monkeypatch.setitem(sys.modules, "vercel", types.ModuleType("vercel"))
config = load_config()
def fake_prompt_choice(question, choices, default=0):
if question == "Select terminal backend:":
return 5
raise AssertionError(f"Unexpected prompt_choice call: {question}")
prompt_values = iter(["python3.13", "yes", "2", "4096", "token", "project", "team"])
monkeypatch.setattr("hermes_cli.setup.prompt_choice", fake_prompt_choice)
monkeypatch.setattr("hermes_cli.setup.prompt", lambda *args, **kwargs: next(prompt_values))
from hermes_cli.setup import setup_terminal_backend
setup_terminal_backend(config)
assert config["terminal"]["backend"] == "vercel_sandbox"
assert config["terminal"]["vercel_runtime"] == "python3.13"
assert config["terminal"]["container_disk"] == 51200
assert os.environ["TERMINAL_VERCEL_RUNTIME"] == "python3.13"
assert "VERCEL_OIDC_TOKEN" not in os.environ
assert os.environ["VERCEL_TOKEN"] == "token"
assert os.environ["VERCEL_PROJECT_ID"] == "project"
assert os.environ["VERCEL_TEAM_ID"] == "team"
def test_vercel_setup_prefills_project_and_team_from_link_file(tmp_path, monkeypatch):
monkeypatch.setenv("HERMES_HOME", str(tmp_path))
_clear_vercel_env(monkeypatch)
project_root = tmp_path / "project"
nested = project_root / "app" / "src"
nested.mkdir(parents=True)
vercel_dir = project_root / ".vercel"
vercel_dir.mkdir()
(vercel_dir / "project.json").write_text(
json.dumps({"projectId": "linked-project", "orgId": "linked-team"}),
encoding="utf-8",
)
monkeypatch.chdir(nested)
monkeypatch.setitem(sys.modules, "vercel", types.ModuleType("vercel"))
config = load_config()
config["terminal"]["container_disk"] = 999
def fake_prompt_choice(question, choices, default=0):
if question == "Select terminal backend:":
return 5
raise AssertionError(f"Unexpected prompt_choice call: {question}")
prompt_values = iter(["node24", "no", "1", "5120", "token", "", ""])
defaults = {}
def fake_prompt(message, default="", **kwargs):
defaults[message] = default
value = next(prompt_values)
return value or default
monkeypatch.setattr("hermes_cli.setup.prompt_choice", fake_prompt_choice)
monkeypatch.setattr("hermes_cli.setup.prompt", fake_prompt)
from hermes_cli.setup import setup_terminal_backend
setup_terminal_backend(config)
assert config["terminal"]["backend"] == "vercel_sandbox"
assert config["terminal"]["container_persistent"] is False
assert config["terminal"]["container_disk"] == 51200
assert "VERCEL_OIDC_TOKEN" not in os.environ
assert os.environ["VERCEL_TOKEN"] == "token"
assert os.environ["VERCEL_PROJECT_ID"] == "linked-project"
assert os.environ["VERCEL_TEAM_ID"] == "linked-team"
assert defaults[" Vercel project ID"] == "linked-project"
assert defaults[" Vercel team ID"] == "linked-team"
def test_setup_slack_saves_home_channel(monkeypatch):
"""_setup_slack() saves SLACK_HOME_CHANNEL when the user provides one."""
saved = {}

31
tests/hermes_cli/test_status.py
View file

@ -83,37 +83,6 @@ def test_show_status_reports_nous_auth_error(monkeypatch, capsys, tmp_path):
assert "Key exp:" in output
def test_show_status_reports_vercel_backend_contract(monkeypatch, capsys, tmp_path):
from hermes_cli import status as status_mod
import hermes_cli.auth as auth_mod
import hermes_cli.gateway as gateway_mod
monkeypatch.setenv("HERMES_HOME", str(tmp_path))
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("TERMINAL_VERCEL_RUNTIME", "python3.13")
monkeypatch.setenv("TERMINAL_CONTAINER_PERSISTENT", "true")
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "oidc-token")
monkeypatch.setattr(status_mod.importlib.util, "find_spec", lambda name: object() if name == "vercel" else None)
monkeypatch.setattr(status_mod, "load_config", lambda: {"terminal": {"backend": "vercel_sandbox"}}, raising=False)
monkeypatch.setattr(auth_mod, "get_nous_auth_status", lambda: {}, raising=False)
monkeypatch.setattr(auth_mod, "get_codex_auth_status", lambda: {}, raising=False)
monkeypatch.setattr(auth_mod, "get_qwen_auth_status", lambda: {}, raising=False)
monkeypatch.setattr(auth_mod, "get_xai_oauth_auth_status", lambda: {}, raising=False)
monkeypatch.setattr(gateway_mod, "find_gateway_pids", lambda exclude_pids=None: [], raising=False)
status_mod.show_status(SimpleNamespace(all=False, deep=False))
output = capsys.readouterr().out
assert "Backend: vercel_sandbox" in output
assert "Runtime: python3.13" in output
assert "Auth:" in output and "OIDC token via VERCEL_OIDC_TOKEN" in output
assert "Auth detail: mode: OIDC" in output
assert "Auth detail: active env: VERCEL_OIDC_TOKEN" in output
assert "oidc-token" not in output
assert "snapshot filesystem" in output
assert "live processes do not survive" in output
# ---------------------------------------------------------------------------
# Helpers shared by xAI OAuth status tests
# ---------------------------------------------------------------------------

2
tests/hermes_cli/test_tencent_tokenhub_provider.py
View file

@ -19,7 +19,7 @@ _OTHER_PROVIDER_KEYS = (
"OPENAI_API_KEY", "ANTHROPIC_API_KEY", "DEEPSEEK_API_KEY",
"GOOGLE_API_KEY", "GEMINI_API_KEY", "DASHSCOPE_API_KEY",
"XAI_API_KEY", "KIMI_API_KEY", "KIMI_CN_API_KEY",
"MINIMAX_API_KEY", "MINIMAX_CN_API_KEY", "AI_GATEWAY_API_KEY",
"MINIMAX_API_KEY", "MINIMAX_CN_API_KEY",
"KILOCODE_API_KEY", "HF_TOKEN", "GLM_API_KEY", "ZAI_API_KEY",
"XIAOMI_API_KEY", "OPENROUTER_API_KEY", "COPILOT_GITHUB_TOKEN",
"GH_TOKEN", "GITHUB_TOKEN", "ARCEEAI_API_KEY",

6
tests/hermes_cli/test_web_server.py
View file

@ -377,12 +377,6 @@ class TestBuildSchemaFromConfig:
assert entry["type"] == "select"
assert "options" in entry
assert "local" in entry["options"]
assert "vercel_sandbox" in entry["options"]
runtime_entry = CONFIG_SCHEMA["terminal.vercel_runtime"]
assert runtime_entry["type"] == "select"
assert "node24" in runtime_entry["options"]
assert "python3.13" in runtime_entry["options"]
assert len(runtime_entry["options"]) >= 3
def test_empty_prefix_produces_correct_keys(self):
from hermes_cli.web_server import _build_schema_from_config

2
tests/hermes_cli/test_xiaomi_provider.py
View file

@ -82,7 +82,7 @@ class TestXiaomiAutoDetection:
for var in ("OPENROUTER_API_KEY", "OPENAI_API_KEY", "ANTHROPIC_API_KEY",
"DEEPSEEK_API_KEY", "GOOGLE_API_KEY", "GEMINI_API_KEY",
"DASHSCOPE_API_KEY", "XAI_API_KEY", "KIMI_API_KEY",
"MINIMAX_API_KEY", "AI_GATEWAY_API_KEY", "KILOCODE_API_KEY",
"MINIMAX_API_KEY", "KILOCODE_API_KEY",
"HF_TOKEN", "GLM_API_KEY", "COPILOT_GITHUB_TOKEN",
"GH_TOKEN", "GITHUB_TOKEN", "MINIMAX_CN_API_KEY",
"TOKENHUB_API_KEY", "ARCEEAI_API_KEY"):

18
tests/providers/test_plugin_discovery.py
View file

@ -46,14 +46,26 @@ def test_bundled_plugins_discovered():
assert (child / "plugin.yaml").exists(), f"{child.name} missing plugin.yaml"
def test_all_34_profiles_register():
"""After discovery, the registry must contain exactly 34 distinct profiles."""
def test_all_profiles_register():
"""After discovery, the registry must contain every bundled provider directory.
This is an invariant — the number of profiles matches the number of plugin
directories, not a hardcoded count. Counts shift when providers are
added/removed; that's expected and shouldn't break CI.
"""
_clear_provider_caches()
from providers import list_providers
plugins_dir = REPO_ROOT / "plugins" / "model-providers"
plugin_dir_count = sum(1 for c in plugins_dir.iterdir() if c.is_dir())
profiles = list_providers()
names = sorted(p.name for p in profiles)
assert len(names) == 34, f"Expected 34 profiles, got {len(names)}: {names}"
# Some plugin __init__.py files register multiple profiles, so the registry
# count is >= the directory count (never less).
assert len(names) >= plugin_dir_count, (
f"Expected at least {plugin_dir_count} profiles (one per plugin dir), got {len(names)}: {names}"
)
# Spot-check representative providers from different categories
for required in (

26
tests/run_agent/test_provider_attribution_headers.py
View file

@ -1,8 +1,4 @@
"""Attribution default_headers applied per provider via base-URL detection.
Mirrors the OpenRouter pattern for the Vercel AI Gateway so that
referrerUrl / appName / User-Agent flow into gateway analytics.
"""
"""Attribution default_headers applied per provider via base-URL detection."""
from types import SimpleNamespace
from unittest.mock import MagicMock, patch
@ -28,26 +24,6 @@ def test_openrouter_base_url_applies_or_headers(mock_openai):
assert headers["X-Title"] == "Hermes Agent"
@patch("run_agent.OpenAI")
def test_ai_gateway_base_url_applies_attribution_headers(mock_openai):
mock_openai.return_value = MagicMock()
agent = AIAgent(
api_key="test-key",
base_url="https://openrouter.ai/api/v1",
model="test/model",
quiet_mode=True,
skip_context_files=True,
skip_memory=True,
)
agent._apply_client_headers_for_base_url("https://ai-gateway.vercel.sh/v1")
headers = agent._client_kwargs["default_headers"]
assert headers["HTTP-Referer"] == "https://hermes-agent.nousresearch.com"
assert headers["X-Title"] == "Hermes Agent"
assert headers["User-Agent"].startswith("HermesAgent/")
@patch("run_agent.OpenAI")
def test_routermint_base_url_applies_user_agent_header(mock_openai):
mock_openai.return_value = MagicMock()

34
tests/run_agent/test_provider_parity.py
View file

@ -313,40 +313,6 @@ class TestBuildApiKwargsKimiNoTemperatureOverride:
assert "temperature" not in kwargs
class TestBuildApiKwargsAIGateway:
def test_uses_chat_completions_format(self, monkeypatch):
agent = _make_agent(monkeypatch, "ai-gateway", base_url="https://ai-gateway.vercel.sh/v1", model="gpt-4o")
messages = [{"role": "user", "content": "hi"}]
kwargs = agent._build_api_kwargs(messages)
assert "messages" in kwargs
assert "model" in kwargs
assert kwargs["messages"][-1]["content"] == "hi"
def test_no_responses_api_fields(self, monkeypatch):
agent = _make_agent(monkeypatch, "ai-gateway", base_url="https://ai-gateway.vercel.sh/v1", model="gpt-4o")
messages = [{"role": "user", "content": "hi"}]
kwargs = agent._build_api_kwargs(messages)
assert "input" not in kwargs
assert "instructions" not in kwargs
assert "store" not in kwargs
def test_includes_reasoning_in_extra_body(self, monkeypatch):
agent = _make_agent(monkeypatch, "ai-gateway", base_url="https://ai-gateway.vercel.sh/v1", model="gpt-4o")
messages = [{"role": "user", "content": "hi"}]
kwargs = agent._build_api_kwargs(messages)
extra = kwargs.get("extra_body", {})
assert "reasoning" in extra
assert extra["reasoning"]["enabled"] is True
def test_includes_tools(self, monkeypatch):
agent = _make_agent(monkeypatch, "ai-gateway", base_url="https://ai-gateway.vercel.sh/v1", model="gpt-4o")
messages = [{"role": "user", "content": "hi"}]
kwargs = agent._build_api_kwargs(messages)
assert "tools" in kwargs
tool_names = [t["function"]["name"] for t in kwargs["tools"]]
assert "web_search" in tool_names
class TestBuildApiKwargsNousPortal:
def test_includes_nous_product_tags(self, monkeypatch):
from agent.portal_tags import nous_portal_tags

2
tests/test_project_metadata.py
View file

@ -70,7 +70,7 @@ def test_lazy_installable_extras_excluded_from_all():
"fal",
"edge-tts", "tts-premium",
"voice", # faster-whisper / sounddevice / numpy
"modal", "daytona", "vercel",
"modal", "daytona",
"messaging", "slack", "matrix", "dingtalk", "feishu",
"honcho", "hindsight",
}

4
tests/tools/test_command_guards.py
View file

@ -73,10 +73,6 @@ class TestContainerSkip:
result = check_all_command_guards("rm -rf /", "daytona")
assert result["approved"] is True
def test_vercel_sandbox_skips_both(self):
result = check_all_command_guards("rm -rf /", "vercel_sandbox")
assert result["approved"] is True
# ---------------------------------------------------------------------------
# tirith allow + safe command

4
tests/tools/test_hardline_blocklist.py
View file

@ -241,7 +241,7 @@ def test_container_backends_still_bypass(clean_session):
Hardline only protects environments with real host impact (local, ssh).
"""
for env in ("docker", "singularity", "modal", "daytona", "vercel_sandbox"):
for env in ("docker", "singularity", "modal", "daytona"):
r1 = check_dangerous_command("rm -rf /", env)
assert r1["approved"] is True, f"container {env} should still bypass"
r2 = check_all_command_guards("rm -rf /", env)
@ -372,7 +372,7 @@ def test_sudo_stdin_guard_not_blocked_by_yolo(clean_session, monkeypatch):
def test_sudo_stdin_guard_container_bypass(clean_session):
"""Containerized backends still bypass — they can't touch the host."""
for env in ("docker", "singularity", "modal", "daytona", "vercel_sandbox"):
for env in ("docker", "singularity", "modal", "daytona"):
for cmd in _SUDO_STDIN_BLOCK:
result = check_all_command_guards(cmd, env)
assert result["approved"] is True, f"container {env} should bypass sudo guard on {cmd!r}"

8
tests/tools/test_local_env_blocklist.py
View file

@ -132,10 +132,6 @@ class TestProviderEnvBlocklist:
"MODAL_TOKEN_ID": "modal-id",
"MODAL_TOKEN_SECRET": "modal-secret",
"DAYTONA_API_KEY": "daytona-key",
"VERCEL_OIDC_TOKEN": "vercel-oidc-token",
"VERCEL_TOKEN": "vercel-token",
"VERCEL_PROJECT_ID": "vercel-project",
"VERCEL_TEAM_ID": "vercel-team",
}
result_env = _run_with_env(extra_os_env=leaked_vars)
@ -291,10 +287,6 @@ class TestBlocklistCoverage:
"MODAL_TOKEN_ID",
"MODAL_TOKEN_SECRET",
"DAYTONA_API_KEY",
"VERCEL_OIDC_TOKEN",
"VERCEL_TOKEN",
"VERCEL_PROJECT_ID",
"VERCEL_TEAM_ID",
}
assert extras.issubset(_HERMES_PROVIDER_ENV_BLOCKLIST)

21
tests/tools/test_modal_sandbox_fixes.py
View file

@ -7,7 +7,6 @@ Covers the bugs discovered while setting up TBLite evaluation:
4. ensurepip fix in Modal image builder
5. No swe-rex dependency — uses native Modal SDK
6. /home/ added to host prefix check
7. Vercel sandbox cwd normalization
"""
import os
@ -102,26 +101,6 @@ class TestCwdHandling:
config = _tt_mod._get_env_config()
assert config["cwd"] == "/root"
def test_host_path_replaced_for_vercel_sandbox(self, monkeypatch):
"""Host paths should be discarded for Vercel Sandbox."""
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("TERMINAL_CWD", "/Users/someone/projects")
config = _tt_mod._get_env_config()
assert config["cwd"] == "/vercel/sandbox"
def test_relative_path_replaced_for_vercel_sandbox(self, monkeypatch):
"""Relative cwd should not map into a remote Vercel sandbox."""
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("TERMINAL_CWD", "src")
config = _tt_mod._get_env_config()
assert config["cwd"] == "/vercel/sandbox"
def test_default_cwd_is_workspace_root_for_vercel_sandbox(self, monkeypatch):
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.delenv("TERMINAL_CWD", raising=False)
config = _tt_mod._get_env_config()
assert config["cwd"] == "/vercel/sandbox"
@pytest.mark.parametrize("backend", ["modal", "docker", "singularity", "daytona"])
def test_default_cwd_is_root_for_container_backends(self, backend, monkeypatch):
"""Container backends should default to /root, not ~."""

2
tests/tools/test_skills_tool.py
View file

@ -958,7 +958,7 @@ class TestSkillViewPrerequisites:
@pytest.mark.parametrize(
"backend",
["ssh", "daytona", "docker", "singularity", "modal", "vercel_sandbox"],
["ssh", "daytona", "docker", "singularity", "modal"],
)
def test_remote_backend_becomes_available_after_local_secret_capture(
self, tmp_path, monkeypatch, backend

128
tests/tools/test_terminal_requirements.py
View file

@ -21,13 +21,8 @@ def _clear_terminal_env(monkeypatch):
"TERMINAL_SSH_PORT",
"TERMINAL_SSH_USER",
"TERMINAL_TIMEOUT",
"TERMINAL_VERCEL_RUNTIME",
"MODAL_TOKEN_ID",
"MODAL_TOKEN_SECRET",
"VERCEL_OIDC_TOKEN",
"VERCEL_TOKEN",
"VERCEL_PROJECT_ID",
"VERCEL_TEAM_ID",
"HOME",
"USERPROFILE",
]
@ -191,126 +186,3 @@ def test_modal_backend_managed_mode_without_feature_flag_logs_clear_error(monkey
"paid Nous subscription is required" in record.getMessage()
for record in caplog.records
)
def test_vercel_backend_without_sdk_logs_specific_error(monkeypatch, caplog):
_clear_terminal_env(monkeypatch)
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setattr(terminal_tool_module.importlib.util, "find_spec", lambda _name: None)
with caplog.at_level(logging.ERROR):
ok = terminal_tool_module.check_terminal_requirements()
assert ok is False
assert any(
"vercel is required for the Vercel Sandbox terminal backend" in record.getMessage()
for record in caplog.records
)
def test_vercel_backend_without_auth_logs_specific_error(monkeypatch, caplog):
_clear_terminal_env(monkeypatch)
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setattr(terminal_tool_module.importlib.util, "find_spec", lambda _name: object())
with caplog.at_level(logging.ERROR):
ok = terminal_tool_module.check_terminal_requirements()
assert ok is False
assert any(
"no supported auth configuration was found" in record.getMessage()
for record in caplog.records
)
def test_vercel_backend_accepts_oidc_auth(monkeypatch):
_clear_terminal_env(monkeypatch)
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "oidc-token")
monkeypatch.setattr(terminal_tool_module.importlib.util, "find_spec", lambda _name: object())
assert terminal_tool_module.check_terminal_requirements() is True
def test_vercel_backend_accepts_token_tuple_auth(monkeypatch):
_clear_terminal_env(monkeypatch)
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("VERCEL_TOKEN", "token")
monkeypatch.setenv("VERCEL_PROJECT_ID", "project")
monkeypatch.setenv("VERCEL_TEAM_ID", "team")
monkeypatch.setattr(terminal_tool_module.importlib.util, "find_spec", lambda _name: object())
assert terminal_tool_module.check_terminal_requirements() is True
@pytest.mark.parametrize("runtime", ["node24", "node22", "python3.13"])
def test_vercel_backend_accepts_supported_runtimes(monkeypatch, runtime):
_clear_terminal_env(monkeypatch)
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("TERMINAL_VERCEL_RUNTIME", runtime)
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "oidc-token")
monkeypatch.setattr(terminal_tool_module.importlib.util, "find_spec", lambda _name: object())
assert terminal_tool_module.check_terminal_requirements() is True
def test_vercel_backend_accepts_blank_runtime(monkeypatch):
_clear_terminal_env(monkeypatch)
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("TERMINAL_VERCEL_RUNTIME", " ")
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "oidc-token")
monkeypatch.setattr(terminal_tool_module.importlib.util, "find_spec", lambda _name: object())
assert terminal_tool_module.check_terminal_requirements() is True
def test_vercel_backend_rejects_unsupported_runtime(monkeypatch, caplog):
_clear_terminal_env(monkeypatch)
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("TERMINAL_VERCEL_RUNTIME", "node20")
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "oidc-token")
monkeypatch.setattr(terminal_tool_module.importlib.util, "find_spec", lambda _name: object())
with caplog.at_level(logging.ERROR):
ok = terminal_tool_module.check_terminal_requirements()
assert ok is False
assert any(
"Vercel Sandbox runtime 'node20' is not supported" in record.getMessage()
and "node24, node22, python3.13" in record.getMessage()
for record in caplog.records
)
def test_vercel_backend_rejects_nondefault_disk(monkeypatch, caplog):
_clear_terminal_env(monkeypatch)
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("TERMINAL_CONTAINER_DISK", "8192")
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "oidc-token")
monkeypatch.setattr(terminal_tool_module.importlib.util, "find_spec", lambda _name: object())
with caplog.at_level(logging.ERROR):
ok = terminal_tool_module.check_terminal_requirements()
assert ok is False
assert any(
"does not support custom TERMINAL_CONTAINER_DISK=8192" in record.getMessage()
for record in caplog.records
)
def test_vercel_backend_rejects_malformed_disk_without_raising(monkeypatch, caplog):
_clear_terminal_env(monkeypatch)
monkeypatch.setenv("TERMINAL_ENV", "vercel_sandbox")
monkeypatch.setenv("TERMINAL_CONTAINER_DISK", "large")
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "oidc-token")
monkeypatch.setattr(terminal_tool_module.importlib.util, "find_spec", lambda _name: object())
with caplog.at_level(logging.ERROR):
ok = terminal_tool_module.check_terminal_requirements()
assert ok is False
assert any(
"Invalid value for TERMINAL_CONTAINER_DISK" in record.getMessage()
for record in caplog.records
)

65
tests/tools/test_terminal_tool_requirements.py
View file

@ -64,68 +64,3 @@ class TestTerminalRequirements:
assert "terminal" in names
assert "execute_code" in names
def test_terminal_and_execute_code_tools_resolve_for_vercel_sandbox(self, monkeypatch):
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "oidc-token")
monkeypatch.setattr(
terminal_tool_module,
"_get_env_config",
lambda: {"env_type": "vercel_sandbox", "container_disk": 51200},
)
monkeypatch.setattr(
terminal_tool_module.importlib.util,
"find_spec",
lambda _name: object(),
)
tools = get_tool_definitions(enabled_toolsets=["terminal", "code_execution"], quiet_mode=True)
names = {tool["function"]["name"] for tool in tools}
assert "terminal" in names
assert "execute_code" in names
def test_terminal_and_execute_code_tools_hide_for_unsupported_vercel_runtime(self, monkeypatch):
monkeypatch.setenv("VERCEL_OIDC_TOKEN", "oidc-token")
monkeypatch.setattr(
terminal_tool_module,
"_get_env_config",
lambda: {
"env_type": "vercel_sandbox",
"container_disk": 51200,
"vercel_runtime": "node20",
},
)
monkeypatch.setattr(
terminal_tool_module.importlib.util,
"find_spec",
lambda _name: object(),
)
tools = get_tool_definitions(enabled_toolsets=["terminal", "code_execution"], quiet_mode=True)
names = {tool["function"]["name"] for tool in tools}
assert "terminal" not in names
assert "execute_code" not in names
def test_terminal_and_execute_code_tools_hide_for_vercel_without_auth(self, monkeypatch):
monkeypatch.delenv("VERCEL_OIDC_TOKEN", raising=False)
monkeypatch.delenv("VERCEL_TOKEN", raising=False)
monkeypatch.delenv("VERCEL_PROJECT_ID", raising=False)
monkeypatch.delenv("VERCEL_TEAM_ID", raising=False)
monkeypatch.setattr(
terminal_tool_module,
"_get_env_config",
lambda: {
"env_type": "vercel_sandbox",
"container_disk": 51200,
"vercel_runtime": "node22",
},
)
monkeypatch.setattr(
terminal_tool_module.importlib.util,
"find_spec",
lambda _name: object(),
)
tools = get_tool_definitions(enabled_toolsets=["terminal", "code_execution"], quiet_mode=True)
names = {tool["function"]["name"] for tool in tools}
assert "terminal" not in names
assert "execute_code" not in names

606
tests/tools/test_vercel_sandbox_environment.py
View file

@ -1,606 +0,0 @@
"""Unit tests for the Vercel Sandbox terminal backend."""
from __future__ import annotations
import importlib
import io
import re
import sys
import tarfile
import threading
import types
from dataclasses import dataclass
from enum import StrEnum
from pathlib import Path
from types import SimpleNamespace
import pytest
class _FakeRunResult:
def __init__(self, output: str | bytes = "", exit_code: int = 0):
self._output = output
self.exit_code = exit_code
def output(self) -> str | bytes:
return self._output
class _FakeSandboxStatus(StrEnum):
PENDING = "pending"
RUNNING = "running"
STOPPING = "stopping"
STOPPED = "stopped"
FAILED = "failed"
ABORTED = "aborted"
SNAPSHOTTING = "snapshotting"
@dataclass(frozen=True)
class _FakeSnapshot:
snapshot_id: str
class _FakeSandbox:
def __init__(
self,
*,
cwd: str = "/vercel/sandbox",
home: str = "/home/vercel",
status: _FakeSandboxStatus = _FakeSandboxStatus.RUNNING,
):
self.sandbox = SimpleNamespace(cwd=cwd, id="sb-123")
self.status = status
self.home = home
self.closed = 0
self.client = SimpleNamespace(close=self._close)
self.run_command_calls: list[tuple[str, list[str], dict]] = []
self.run_command_side_effects: list[object] = []
self.write_files_calls: list[list[dict[str, object]]] = []
self.write_files_side_effects: list[object] = []
self.download_file_calls: list[tuple[str, Path]] = []
self.download_file_side_effects: list[object] = []
self.download_file_content = b""
self.stop_calls: list[tuple[tuple, dict]] = []
self.snapshot_calls: list[tuple[tuple, dict]] = []
self.snapshot_side_effects: list[object] = []
self.snapshot_id = "snap_default"
self.refresh_calls = 0
self.wait_for_status_calls: list[tuple[object, object, object]] = []
self.wait_for_status_side_effects: list[object] = []
def _close(self) -> None:
self.closed += 1
def refresh(self) -> None:
self.refresh_calls += 1
def wait_for_status(self, status: _FakeSandboxStatus | str, *, timeout, poll_interval) -> None:
self.wait_for_status_calls.append((status, timeout, poll_interval))
if self.wait_for_status_side_effects:
effect = self.wait_for_status_side_effects.pop(0)
if isinstance(effect, Exception):
raise effect
if callable(effect):
effect(status, timeout, poll_interval)
return
self.status = _FakeSandboxStatus(status)
def run_command(self, cmd: str, args: list[str] | None = None, **kwargs):
args = list(args or [])
self.run_command_calls.append((cmd, args, kwargs))
if self.run_command_side_effects:
effect = self.run_command_side_effects.pop(0)
if isinstance(effect, Exception):
raise effect
if callable(effect):
return effect(cmd, args, kwargs)
return effect
script = args[1] if len(args) > 1 else ""
if 'printf %s "$HOME"' in script:
return _FakeRunResult(self.home)
return _FakeRunResult("")
def write_files(self, files: list[dict[str, object]]) -> None:
self.write_files_calls.append(files)
if self.write_files_side_effects:
effect = self.write_files_side_effects.pop(0)
if isinstance(effect, Exception):
raise effect
if callable(effect):
effect(files)
def download_file(self, remote_path: str, local_path) -> str:
destination = Path(local_path)
self.download_file_calls.append((remote_path, destination))
if self.download_file_side_effects:
effect = self.download_file_side_effects.pop(0)
if isinstance(effect, Exception):
raise effect
if callable(effect):
return effect(remote_path, destination)
destination.write_bytes(self.download_file_content)
return str(destination.resolve())
def stop(self, *args, **kwargs) -> None:
self.stop_calls.append((args, kwargs))
def snapshot(self, *args, **kwargs):
self.snapshot_calls.append((args, kwargs))
if self.snapshot_side_effects:
effect = self.snapshot_side_effects.pop(0)
if isinstance(effect, Exception):
raise effect
if callable(effect):
return effect(*args, **kwargs)
if isinstance(effect, str):
return _FakeSnapshot(effect)
return effect
return _FakeSnapshot(self.snapshot_id)
@dataclass(frozen=True)
class _FakeResources:
vcpus: float | None = None
memory: int | None = None
@dataclass(frozen=True)
class _FakeWriteFile:
path: str
content: bytes
class _FakeSDK:
def __init__(self):
self.create_kwargs: list[dict[str, object]] = []
self.create_side_effects: list[object] = []
self.sandboxes: list[_FakeSandbox] = []
@property
def current(self) -> _FakeSandbox:
return self.sandboxes[-1]
def create(self, **kwargs):
self.create_kwargs.append(kwargs)
if self.create_side_effects:
effect = self.create_side_effects.pop(0)
if isinstance(effect, Exception):
raise effect
if isinstance(effect, _FakeSandbox):
self.sandboxes.append(effect)
return effect
sandbox = _FakeSandbox()
self.sandboxes.append(sandbox)
return sandbox
def _cwd_result(body: str = "", *, cwd: str = "/vercel/sandbox", exit_code: int = 0):
def _result(_cmd: str, args: list[str], _kwargs: dict):
script = args[1] if len(args) > 1 else ""
match = re.search(r"__HERMES_CWD_[A-Za-z0-9]+__", script)
marker = match.group(0) if match else "__HERMES_CWD_MISSING__"
prefix = f"{body}\n\n" if body else "\n"
return _FakeRunResult(f"{prefix}{marker}{cwd}{marker}\n", exit_code)
return _result
def _tar_bytes(entries: dict[str, bytes]) -> bytes:
buffer = io.BytesIO()
with tarfile.open(fileobj=buffer, mode="w") as tar:
for name, content in entries.items():
info = tarfile.TarInfo(name)
info.size = len(content)
tar.addfile(info, io.BytesIO(content))
return buffer.getvalue()
@pytest.fixture()
def vercel_sdk(monkeypatch):
fake_sdk = _FakeSDK()
sandbox_mod = types.ModuleType("vercel.sandbox")
sandbox_mod.Sandbox = types.SimpleNamespace(create=fake_sdk.create)
sandbox_mod.Resources = _FakeResources
sandbox_mod.WriteFile = _FakeWriteFile
sandbox_mod.SandboxStatus = _FakeSandboxStatus
vercel_mod = types.ModuleType("vercel")
vercel_mod.sandbox = sandbox_mod
monkeypatch.setitem(sys.modules, "vercel", vercel_mod)
monkeypatch.setitem(sys.modules, "vercel.sandbox", sandbox_mod)
return fake_sdk
@pytest.fixture()
def vercel_module(vercel_sdk, monkeypatch):
monkeypatch.setattr("tools.environments.base.is_interrupted", lambda: False)
monkeypatch.setattr("tools.credential_files.get_credential_file_mounts", lambda: [])
monkeypatch.setattr("tools.credential_files.iter_skills_files", lambda **kwargs: [])
monkeypatch.setattr("tools.credential_files.iter_cache_files", lambda **kwargs: [])
module = importlib.import_module("tools.environments.vercel_sandbox")
return importlib.reload(module)
@pytest.fixture()
def make_env(vercel_module, request):
envs = []
def _cleanup_envs():
for env in envs:
env._sync_manager = None
env.cleanup()
request.addfinalizer(_cleanup_envs)
def _factory(**kwargs):
kwargs.setdefault("runtime", "node22")
kwargs.setdefault("cwd", vercel_module.DEFAULT_VERCEL_CWD)
kwargs.setdefault("timeout", 30)
kwargs.setdefault("task_id", "task-123")
env = vercel_module.VercelSandboxEnvironment(**kwargs)
envs.append(env)
return env
return _factory
class TestStartup:
def test_default_cwd_tracks_remote_workspace_root(self, make_env, vercel_sdk):
sandbox = _FakeSandbox(cwd="/workspace")
vercel_sdk.create_side_effects.append(sandbox)
env = make_env()
assert env.cwd == "/workspace"
def test_tilde_cwd_resolves_against_remote_home(self, make_env, vercel_sdk):
sandbox = _FakeSandbox(home="/home/custom")
vercel_sdk.create_side_effects.append(sandbox)
env = make_env(cwd="~")
assert env.cwd == "/home/custom"
def test_pending_sandbox_timeout_raises_descriptive_error(
self, make_env, vercel_sdk
):
sandbox = _FakeSandbox(status=_FakeSandboxStatus.PENDING)
sandbox.wait_for_status_side_effects.append(TimeoutError("still pending"))
vercel_sdk.create_side_effects.append(sandbox)
with pytest.raises(RuntimeError, match="Sandbox did not reach running state"):
make_env()
class TestFileSync:
def test_initial_sync_uploads_managed_files_under_remote_home(
self, make_env, vercel_sdk, monkeypatch, tmp_path
):
src = tmp_path / "token.txt"
src.write_text("secret-token")
monkeypatch.setattr(
"tools.credential_files.get_credential_file_mounts",
lambda: [
{
"host_path": str(src),
"container_path": "/root/.hermes/credentials/token.txt",
}
],
)
monkeypatch.setattr("tools.credential_files.iter_skills_files", lambda **kwargs: [])
monkeypatch.setattr("tools.credential_files.iter_cache_files", lambda **kwargs: [])
make_env()
uploaded = vercel_sdk.current.write_files_calls[0]
assert uploaded == [
{
"path": "/home/vercel/.hermes/credentials/token.txt",
"content": b"secret-token",
}
]
def test_execute_resyncs_changed_managed_files(
self, make_env, vercel_sdk, monkeypatch, tmp_path
):
src = tmp_path / "token.txt"
src.write_text("secret-token")
monkeypatch.setattr(
"tools.credential_files.get_credential_file_mounts",
lambda: [
{
"host_path": str(src),
"container_path": "/root/.hermes/credentials/token.txt",
}
],
)
monkeypatch.setattr("tools.credential_files.iter_skills_files", lambda **kwargs: [])
monkeypatch.setattr("tools.credential_files.iter_cache_files", lambda **kwargs: [])
env = make_env()
src.write_text("updated-secret-token")
monkeypatch.setenv("HERMES_FORCE_FILE_SYNC", "1")
vercel_sdk.current.run_command_side_effects.append(_cwd_result("hello"))
result = env.execute("echo hello")
assert result == {"output": "hello\n", "returncode": 0}
assert vercel_sdk.current.write_files_calls[-1] == [
{
"path": "/home/vercel/.hermes/credentials/token.txt",
"content": b"updated-secret-token",
}
]
def test_cleanup_syncs_back_snapshots_closes_and_is_idempotent(
self, make_env, vercel_module, vercel_sdk, monkeypatch, tmp_path
):
hermes_home = tmp_path / ".hermes"
monkeypatch.setenv("HERMES_HOME", str(hermes_home))
src = tmp_path / "token.txt"
src.write_text("host-token")
monkeypatch.setattr(
"tools.credential_files.get_credential_file_mounts",
lambda: [
{
"host_path": str(src),
"container_path": "/root/.hermes/credentials/token.txt",
}
],
)
monkeypatch.setattr(
"tools.credential_files.iter_skills_files",
lambda **kwargs: [],
)
monkeypatch.setattr(
"tools.credential_files.iter_cache_files",
lambda **kwargs: [],
)
env = make_env()
sandbox = vercel_sdk.current
sandbox.snapshot_id = "snap_cleanup"
vercel_sdk.current.download_file_content = _tar_bytes(
{
"home/vercel/.hermes/credentials/token.txt": b"remote-token",
"home/vercel/.hermes/credentials/new.txt": b"new-remote",
"home/vercel/.hermes/unmapped/skip.txt": b"skip",
}
)
env.cleanup()
env.cleanup()
assert src.read_text() == "remote-token"
assert (tmp_path / "new.txt").read_text() == "new-remote"
assert not (tmp_path / "skip.txt").exists()
assert len(sandbox.snapshot_calls) == 1
assert len(sandbox.stop_calls) == 1 # always stop after snapshot to avoid resource leaks
assert sandbox.closed == 1
assert vercel_module._load_snapshots() == {"task-123": "snap_cleanup"}
def test_cleanup_sync_back_failure_from_download_does_not_block_snapshot(
self, make_env, vercel_sdk, monkeypatch, tmp_path
):
src = tmp_path / "token.txt"
src.write_text("host-token")
monkeypatch.setattr(
"tools.credential_files.get_credential_file_mounts",
lambda: [
{
"host_path": str(src),
"container_path": "/root/.hermes/credentials/token.txt",
}
],
)
monkeypatch.setattr(
"tools.credential_files.iter_skills_files",
lambda **kwargs: [],
)
monkeypatch.setattr(
"tools.credential_files.iter_cache_files",
lambda **kwargs: [],
)
env = make_env()
sandbox = vercel_sdk.current
sandbox.run_command_side_effects.extend(
[
_FakeRunResult("tar failed", exit_code=2),
_FakeRunResult(""),
_FakeRunResult("tar failed", exit_code=2),
_FakeRunResult(""),
_FakeRunResult("tar failed", exit_code=2),
_FakeRunResult(""),
]
)
monkeypatch.setattr("tools.environments.file_sync.time.sleep", lambda _delay: None)
env.cleanup()
assert src.read_text() == "host-token"
assert len(sandbox.snapshot_calls) == 1
assert sandbox.closed == 1
assert len(sandbox.download_file_calls) == 0
class TestExecute:
@pytest.mark.parametrize(
("make_unhealthy", "label"),
[
(
lambda sandbox: setattr(
sandbox, "status", _FakeSandboxStatus.STOPPED
),
"terminal state",
),
(
lambda sandbox: setattr(
sandbox,
"refresh",
lambda: (_ for _ in ()).throw(RuntimeError("refresh failed")),
),
"refresh failure",
),
],
ids=["terminal-state", "refresh-failure"],
)
def test_execute_recreates_unhealthy_sandbox_before_running_command(
self, make_env, vercel_sdk, make_unhealthy, label
):
env = make_env()
original = vercel_sdk.current
make_unhealthy(original)
replacement = _FakeSandbox()
replacement.run_command_side_effects.extend(
[
_FakeRunResult(replacement.home),
_cwd_result("hello"),
]
)
vercel_sdk.create_side_effects.append(replacement)
result = env.execute("echo hello")
assert result == {"output": "hello\n", "returncode": 0}, label
assert original.closed == 1
assert vercel_sdk.current is replacement
def test_run_bash_handle_uses_captured_sandbox_for_exec_and_cancel(
self, make_env
):
env = make_env()
original = env._sandbox
assert original is not None
replacement = _FakeSandbox()
started = threading.Event()
release = threading.Event()
def blocking_command(_cmd: str, _args: list[str], _kwargs: dict):
started.set()
release.wait(timeout=5)
return _FakeRunResult("done")
original.run_command_side_effects.append(blocking_command)
handle = env._run_bash("echo done")
assert started.wait(timeout=1)
env._sandbox = replacement
handle.kill()
release.set()
assert handle.wait(timeout=2) == 0
assert len(original.stop_calls) == 1
assert replacement.stop_calls == []
cmd, args, kwargs = original.run_command_calls[-1]
assert cmd == "bash"
assert args == ["-c", "echo done"]
assert kwargs["cwd"] == "/vercel/sandbox"
class TestSnapshotPersistence:
def test_create_restores_from_saved_snapshot(
self, make_env, vercel_module, vercel_sdk, monkeypatch, tmp_path
):
hermes_home = tmp_path / ".hermes"
monkeypatch.setenv("HERMES_HOME", str(hermes_home))
vercel_module._store_snapshot("task-123", "snap_saved")
restored = _FakeSandbox(cwd="/restored")
vercel_sdk.create_side_effects.append(restored)
env = make_env()
assert env.cwd == "/restored"
assert vercel_sdk.create_kwargs[0]["source"] == {
"type": "snapshot",
"snapshot_id": "snap_saved",
}
assert vercel_module._load_snapshots() == {"task-123": "snap_saved"}
def test_restore_failure_prunes_snapshot_and_falls_back_to_fresh_sandbox(
self, make_env, vercel_module, vercel_sdk, monkeypatch, tmp_path
):
hermes_home = tmp_path / ".hermes"
monkeypatch.setenv("HERMES_HOME", str(hermes_home))
vercel_module._store_snapshot("task-123", "snap_stale")
fresh = _FakeSandbox(cwd="/fresh")
vercel_sdk.create_side_effects.extend(
[RuntimeError("snapshot missing"), fresh]
)
env = make_env()
assert env.cwd == "/fresh"
assert vercel_sdk.create_kwargs[0]["source"] == {
"type": "snapshot",
"snapshot_id": "snap_stale",
}
assert "source" not in vercel_sdk.create_kwargs[1]
assert vercel_module._load_snapshots() == {}
def test_cleanup_stops_when_snapshot_fails_without_storing_metadata(
self, make_env, vercel_module, vercel_sdk, monkeypatch, tmp_path
):
hermes_home = tmp_path / ".hermes"
monkeypatch.setenv("HERMES_HOME", str(hermes_home))
env = make_env()
sandbox = vercel_sdk.current
sandbox.snapshot_side_effects.append(RuntimeError("snapshot failed"))
env.cleanup()
assert len(sandbox.snapshot_calls) == 1
assert len(sandbox.stop_calls) == 1
assert sandbox.closed == 1
assert vercel_module._load_snapshots() == {}
def test_non_persistent_cleanup_stops_without_snapshot(
self, make_env, vercel_module, vercel_sdk, monkeypatch, tmp_path
):
hermes_home = tmp_path / ".hermes"
monkeypatch.setenv("HERMES_HOME", str(hermes_home))
env = make_env(persistent_filesystem=False)
sandbox = vercel_sdk.current
env.cleanup()
assert sandbox.snapshot_calls == []
assert len(sandbox.stop_calls) == 1
assert sandbox.closed == 1
assert vercel_module._load_snapshots() == {}
def test_persistent_cleanup_without_task_id_stops_without_snapshot(
self, make_env, vercel_module, vercel_sdk, monkeypatch, tmp_path
):
hermes_home = tmp_path / ".hermes"
monkeypatch.setenv("HERMES_HOME", str(hermes_home))
env = make_env(task_id="")
sandbox = vercel_sdk.current
env.cleanup()
assert sandbox.snapshot_calls == []
assert len(sandbox.stop_calls) == 1
assert sandbox.closed == 1
assert vercel_module._load_snapshots() == {}
class TestCleanup:
def test_cleanup_continues_when_sync_back_raises(self, make_env, vercel_sdk):
env = make_env()
sandbox = vercel_sdk.current
class FailingSyncManager:
def sync_back(self):
raise RuntimeError("download failed")
env._sync_manager = FailingSyncManager()
env.cleanup()
assert len(sandbox.snapshot_calls) == 1
assert sandbox.closed == 1

4
tools/approval.py
View file

@ -930,7 +930,7 @@ def check_dangerous_command(command: str, env_type: str,
Returns:
{"approved": True/False, "message": str or None, ...}
"""
if env_type in {"docker", "singularity", "modal", "daytona", "vercel_sandbox"}:
if env_type in {"docker", "singularity", "modal", "daytona"}:
return {"approved": True, "message": None}
# Hardline floor: commands with no recovery path (rm -rf /, mkfs, dd
@ -1060,7 +1060,7 @@ def check_all_command_guards(command: str, env_type: str,
other was shown to the user.
"""
# Skip containers for both checks
if env_type in {"docker", "singularity", "modal", "daytona", "vercel_sandbox"}:
if env_type in {"docker", "singularity", "modal", "daytona"}:
return {"approved": True, "message": None}
# Hardline floor: unconditional block for catastrophic commands

18
tools/code_execution_tool.py
View file

@ -157,21 +157,6 @@ def check_sandbox_requirements() -> bool:
"""Code execution sandbox requires a POSIX OS for Unix domain sockets."""
if not SANDBOX_AVAILABLE:
return False
try:
from tools.terminal_tool import (
_check_vercel_sandbox_requirements,
_get_env_config,
)
config = _get_env_config()
except Exception:
logger.debug("Could not resolve terminal config for execute_code availability", exc_info=True)
return False
if config.get("env_type") == "vercel_sandbox":
return _check_vercel_sandbox_requirements(config)
return True
@ -612,13 +597,12 @@ def _get_or_create_env(task_id: str):
cwd = overrides.get("cwd") or config["cwd"]
container_config = None
if env_type in {"docker", "singularity", "modal", "daytona", "vercel_sandbox"}:
if env_type in {"docker", "singularity", "modal", "daytona"}:
container_config = {
"container_cpu": config.get("container_cpu", 1),
"container_memory": config.get("container_memory", 5120),
"container_disk": config.get("container_disk", 51200),
"container_persistent": config.get("container_persistent", True),
"vercel_runtime": config.get("vercel_runtime", ""),
"docker_volumes": config.get("docker_volumes", []),
"docker_run_as_host_user": config.get("docker_run_as_host_user", False),
}

2
tools/credential_files.py
View file

@ -385,7 +385,7 @@ def to_agent_visible_cache_path(
translation (only Docker for now).
"""
# Only Docker backend requires translation at this time. Other backends
# (Modal, Daytona, Vercel) use different mount semantics and will be
# (Modal, Daytona) use different mount semantics and will be
# addressed separately if needed. Backend is identified by TERMINAL_ENV
# (same env var tools/terminal_tool.py reads in _get_environment_config).
if os.environ.get("TERMINAL_ENV", "local") != "docker":

4
tools/environments/__init__.py
View file

@ -2,8 +2,8 @@
Each backend provides the same interface (BaseEnvironment ABC) for running
shell commands in a specific execution context: local, Docker, SSH,
Singularity, Modal, Daytona, or Vercel Sandbox. (Modal additionally has
direct and Nous-managed modes, selected via terminal.modal_mode.)
Singularity, Modal, or Daytona. (Modal additionally has direct and
Nous-managed modes, selected via terminal.modal_mode.)
The terminal_tool.py factory (_create_environment) selects the backend
based on the TERMINAL_ENV configuration.

4
tools/environments/local.py
View file

@ -160,10 +160,6 @@ def _build_provider_env_blocklist() -> frozenset:
"MODAL_TOKEN_ID",
"MODAL_TOKEN_SECRET",
"DAYTONA_API_KEY",
"VERCEL_OIDC_TOKEN",
"VERCEL_TOKEN",
"VERCEL_PROJECT_ID",
"VERCEL_TEAM_ID",
})
return frozenset(blocked)

654
tools/environments/vercel_sandbox.py
View file

@ -1,654 +0,0 @@
"""Vercel Sandbox execution environment.
Uses the Vercel Python SDK to run commands in cloud sandboxes through Hermes'
shared ``BaseEnvironment`` shell contract. When persistence is enabled, the
backend stores task-scoped snapshot metadata under ``HERMES_HOME`` and restores
new sandboxes from those snapshots on later task reuse.
"""
from __future__ import annotations
from functools import cache
from dataclasses import dataclass
from datetime import timedelta
import logging
import math
import os
import shlex
import threading
import time
from pathlib import Path
from typing import TYPE_CHECKING, Any
import httpx
from hermes_constants import get_hermes_home
from tools.environments.base import (
BaseEnvironment,
_ThreadedProcessHandle,
_load_json_store,
_save_json_store,
)
from tools.environments.file_sync import (
FileSyncManager,
iter_sync_files,
quoted_rm_command,
)
logger = logging.getLogger(__name__)
if TYPE_CHECKING:
from vercel.sandbox import Resources, Sandbox, SandboxStatus, WriteFile
DEFAULT_VERCEL_CWD = "/vercel/sandbox"
_DEFAULT_CONTAINER_DISK_MB = 51200
def _ensure_vercel_sdk() -> None:
"""Lazy-install vercel SDK on demand. Idempotent."""
try:
from tools.lazy_deps import ensure as _lazy_ensure
_lazy_ensure("terminal.vercel", prompt=False)
except ImportError:
pass
except Exception as e:
raise ImportError(str(e))
_CREATE_RETRY_ATTEMPTS = 3
_WRITE_RETRY_ATTEMPTS = 3
_TRANSIENT_STATUS_CODES = frozenset({408, 425, 429, 500, 502, 503, 504})
_RETRY_BACKOFF_STEP = timedelta(milliseconds=100)
_MIN_SANDBOX_TIMEOUT = timedelta(minutes=5)
_MIN_RUNNING_WAIT = timedelta(seconds=1)
_RUNNING_WAIT_TIMEOUT = timedelta(seconds=30)
_RUNNING_WAIT_POLL_INTERVAL = timedelta(milliseconds=250)
_STOP_TIMEOUT = timedelta(seconds=15)
_STOP_POLL_INTERVAL = timedelta(milliseconds=500)
_SNAPSHOT_STORE_NAME = "vercel_sandbox_snapshots.json"
def _exception_chain(exc: BaseException) -> list[BaseException]:
chain: list[BaseException] = []
current: BaseException | None = exc
seen: set[int] = set()
while current is not None and id(current) not in seen:
chain.append(current)
seen.add(id(current))
current = current.__cause__ or current.__context__
return chain
def _extract_status_code(exc: BaseException) -> int | None:
response = getattr(exc, "response", None)
for value in (getattr(exc, "status_code", None), getattr(response, "status_code", None)):
if isinstance(value, int):
return value
return None
def _is_transient_vercel_error(exc: BaseException) -> bool:
for error in _exception_chain(exc):
status_code = _extract_status_code(error)
if status_code in _TRANSIENT_STATUS_CODES:
return True
if isinstance(
error,
(httpx.NetworkError, httpx.ProtocolError, httpx.ReadError),
):
return True
error_name = type(error).__name__.lower()
if "ratelimit" in error_name or "servererror" in error_name:
return True
return False
def _retry_vercel_call(
label: str,
callback,
*,
attempts: int,
):
backoff_seconds = _RETRY_BACKOFF_STEP.total_seconds()
for attempt in range(1, attempts + 1):
try:
return callback()
except Exception as exc:
if attempt >= attempts or not _is_transient_vercel_error(exc):
raise
logger.warning(
"Vercel: %s failed (%s); retrying %d/%d",
label,
exc,
attempt,
attempts,
)
time.sleep(backoff_seconds * attempt)
def _coerce_text(value: Any) -> str:
if value is None:
return ""
if isinstance(value, bytes):
return value.decode("utf-8", errors="replace")
return str(value)
def _extract_result_output(result: Any) -> str:
try:
return _coerce_text(result.output())
except (AttributeError, TypeError):
return _coerce_text(result)
def _extract_result_returncode(result: Any) -> int:
try:
exit_code = result.exit_code
except AttributeError:
try:
exit_code = result.returncode
except AttributeError:
return 1
return exit_code if isinstance(exit_code, int) else 1
def _snapshot_store_path() -> Path:
return get_hermes_home() / _SNAPSHOT_STORE_NAME
def _load_snapshots() -> dict:
return _load_json_store(_snapshot_store_path())
def _save_snapshots(data: dict) -> None:
_save_json_store(_snapshot_store_path(), data)
def _get_snapshot_id(task_id: str) -> str | None:
if not task_id:
return None
snapshot_id = _load_snapshots().get(task_id)
return snapshot_id if isinstance(snapshot_id, str) and snapshot_id else None
def _store_snapshot(task_id: str, snapshot_id: str) -> None:
if not task_id or not snapshot_id:
return
snapshots = _load_snapshots()
snapshots[task_id] = snapshot_id
_save_snapshots(snapshots)
def _delete_snapshot(task_id: str, snapshot_id: str | None = None) -> None:
if not task_id:
return
snapshots = _load_snapshots()
existing = snapshots.get(task_id)
if existing is None:
return
if snapshot_id is not None and existing != snapshot_id:
return
snapshots.pop(task_id, None)
_save_snapshots(snapshots)
def _extract_snapshot_id(snapshot: Any) -> str | None:
for attr in ("snapshot_id", "snapshotId", "id"):
value = getattr(snapshot, attr, None)
if isinstance(value, str) and value:
return value
if isinstance(snapshot, dict):
for key in ("snapshot_id", "snapshotId", "id"):
value = snapshot.get(key)
if isinstance(value, str) and value:
return value
return None
@cache
def _sandbox_status_type() -> type[SandboxStatus]:
_ensure_vercel_sdk()
from vercel.sandbox import SandboxStatus
return SandboxStatus
@cache
def _terminal_sandbox_states() -> frozenset[SandboxStatus]:
SandboxStatus = _sandbox_status_type()
return frozenset(
{
SandboxStatus.ABORTED,
SandboxStatus.FAILED,
SandboxStatus.STOPPED,
}
)
@dataclass(frozen=True, slots=True)
class _SandboxCreateParams:
timeout: timedelta
runtime: str | None = None
resources: Resources | None = None
class VercelSandboxEnvironment(BaseEnvironment):
"""Vercel cloud sandbox backend."""
_stdin_mode = "heredoc"
def __init__(
self,
runtime: str | None = None,
cwd: str = DEFAULT_VERCEL_CWD,
timeout: int = 60,
cpu: float = 1,
memory: int = 5120,
disk: int = _DEFAULT_CONTAINER_DISK_MB,
persistent_filesystem: bool = True,
task_id: str = "default",
):
requested_cwd = cwd
super().__init__(cwd=cwd, timeout=timeout)
self._runtime = runtime or None
self._persistent = persistent_filesystem
self._task_id = task_id
self._requested_cwd = requested_cwd
self._lock = threading.Lock()
self._sandbox: Sandbox | None = None
self._workspace_root = DEFAULT_VERCEL_CWD
self._remote_home = DEFAULT_VERCEL_CWD
self._sync_manager: FileSyncManager | None = None
self._create_params = self._build_create_params(cpu=cpu, memory=memory, disk=disk)
self._sandbox = self._create_sandbox()
self._configure_attached_sandbox(requested_cwd=requested_cwd)
self._sync_manager.sync(force=True)
self.init_session()
def _build_create_params(self, *, cpu: float, memory: int, disk: int) -> _SandboxCreateParams:
if disk not in {0, _DEFAULT_CONTAINER_DISK_MB}:
raise ValueError(
"Vercel Sandbox does not support configurable container_disk. "
"Use the default shared setting."
)
_ensure_vercel_sdk()
from vercel.sandbox import Resources
sandbox_timeout = max(
timedelta(seconds=max(self.timeout, 0)),
_MIN_SANDBOX_TIMEOUT,
)
vcpus = math.floor(cpu) if cpu > 0 else None
memory_mb = memory if memory > 0 else None
resources = (
Resources(vcpus=vcpus, memory=memory_mb)
if vcpus is not None or memory_mb is not None
else None
)
return _SandboxCreateParams(
timeout=sandbox_timeout,
runtime=self._runtime,
resources=resources,
)
def _create_sandbox(self) -> Sandbox:
_ensure_vercel_sdk()
from vercel.sandbox import Sandbox
snapshot_id = _get_snapshot_id(self._task_id) if self._persistent else None
if snapshot_id:
try:
return _retry_vercel_call(
"sandbox restore",
lambda: Sandbox.create(
timeout=self._create_params.timeout,
runtime=self._create_params.runtime,
resources=self._create_params.resources,
source={"type": "snapshot", "snapshot_id": snapshot_id},
),
attempts=_CREATE_RETRY_ATTEMPTS,
)
except Exception as exc:
logger.warning(
"Vercel: failed to restore snapshot %s for task %s; "
"falling back to a fresh sandbox: %s",
snapshot_id,
self._task_id,
exc,
)
_delete_snapshot(self._task_id, snapshot_id)
params = self._create_params
return _retry_vercel_call(
"sandbox create",
lambda: Sandbox.create(
timeout=params.timeout,
runtime=params.runtime,
resources=params.resources,
),
attempts=_CREATE_RETRY_ATTEMPTS,
)
def _configure_attached_sandbox(self, *, requested_cwd: str) -> None:
self._wait_for_running()
self._workspace_root = self._detect_workspace_root()
self._remote_home = self._detect_remote_home()
if self._remote_home == "/":
container_base = "/.hermes"
else:
container_base = f"{self._remote_home.rstrip('/')}/.hermes"
self._sync_manager = FileSyncManager(
get_files_fn=lambda: iter_sync_files(container_base),
upload_fn=self._vercel_upload,
delete_fn=self._vercel_delete,
bulk_upload_fn=self._vercel_bulk_upload,
bulk_download_fn=self._vercel_bulk_download,
)
if requested_cwd == "~":
self.cwd = self._remote_home
elif requested_cwd in {"", DEFAULT_VERCEL_CWD}:
self.cwd = self._workspace_root
else:
self.cwd = requested_cwd
def _detect_workspace_root(self) -> str:
sandbox = self._sandbox
if sandbox is None:
raise RuntimeError("Vercel sandbox is not attached")
cwd = sandbox.sandbox.cwd
return cwd if cwd.startswith("/") else DEFAULT_VERCEL_CWD
def _detect_remote_home(self) -> str:
sandbox = self._sandbox
if sandbox is None:
raise RuntimeError("Vercel sandbox is not attached")
try:
result = sandbox.run_command(
"sh",
["-lc", 'printf %s "$HOME"'],
cwd=self._workspace_root,
)
except Exception as exc:
logger.debug(
"Vercel: home detection failed for task %s: %s",
self._task_id,
exc,
)
return self._workspace_root
home = _extract_result_output(result).strip()
if home.startswith("/"):
return home
return self._workspace_root
def _wait_for_running(self, timeout: timedelta = _RUNNING_WAIT_TIMEOUT) -> None:
sandbox = self._sandbox
if sandbox is None:
raise RuntimeError("Vercel sandbox is not attached")
SandboxStatus = _sandbox_status_type()
status = sandbox.status
if status is None or status == SandboxStatus.RUNNING:
return
if status in _terminal_sandbox_states():
raise RuntimeError(f"Sandbox entered terminal state: {status}")
try:
sandbox.wait_for_status(
SandboxStatus.RUNNING,
timeout=max(timeout, _MIN_RUNNING_WAIT),
poll_interval=_RUNNING_WAIT_POLL_INTERVAL,
)
except TimeoutError as exc:
status = sandbox.status
if status in _terminal_sandbox_states():
raise RuntimeError(f"Sandbox entered terminal state: {status}") from exc
raise RuntimeError(
f"Sandbox did not reach running state (last status: {status})"
) from exc
def _close_sandbox_client(self, sandbox: Sandbox | None) -> None:
if sandbox is None:
return
try:
sandbox.client.close()
except Exception:
pass
def _stop_sandbox(self, sandbox: Sandbox | None) -> None:
if sandbox is None:
return
try:
sandbox.stop(
blocking=True,
timeout=_STOP_TIMEOUT,
poll_interval=_STOP_POLL_INTERVAL,
)
except TypeError:
try:
sandbox.stop()
except Exception:
pass
except Exception:
pass
def _snapshot_sandbox(self, sandbox: Sandbox) -> str | None:
if not self._persistent or not self._task_id:
return None
try:
snapshot = sandbox.snapshot()
except Exception as exc:
logger.warning(
"Vercel: filesystem snapshot failed for task %s: %s",
self._task_id,
exc,
)
return None
snapshot_id = _extract_snapshot_id(snapshot)
if not snapshot_id:
logger.warning(
"Vercel: filesystem snapshot for task %s did not return a snapshot id",
self._task_id,
)
return None
_store_snapshot(self._task_id, snapshot_id)
logger.info(
"Vercel: saved filesystem snapshot %s for task %s",
snapshot_id,
self._task_id,
)
return snapshot_id
def _ensure_sandbox_ready(self) -> None:
sandbox = self._sandbox
requested_cwd = self.cwd or self._requested_cwd or DEFAULT_VERCEL_CWD
if sandbox is None:
self._sandbox = self._create_sandbox()
self._configure_attached_sandbox(requested_cwd=requested_cwd)
return
try:
sandbox.refresh()
except Exception as exc:
logger.warning(
"Vercel: sandbox refresh failed for task %s: %s; recreating",
self._task_id,
exc,
)
self._close_sandbox_client(sandbox)
self._sandbox = self._create_sandbox()
self._configure_attached_sandbox(requested_cwd=requested_cwd)
return
status = sandbox.status
if status in _terminal_sandbox_states():
logger.warning(
"Vercel: sandbox entered state %s for task %s; recreating",
status,
self._task_id,
)
self._close_sandbox_client(sandbox)
self._sandbox = self._create_sandbox()
self._configure_attached_sandbox(requested_cwd=requested_cwd)
return
self._wait_for_running()
def _vercel_upload(self, host_path: str, remote_path: str) -> None:
self._vercel_bulk_upload([(host_path, remote_path)])
def _vercel_bulk_upload(self, files: list[tuple[str, str]]) -> None:
if not files:
return
payload: list[WriteFile] = [
{
"path": remote_path,
"content": Path(host_path).read_bytes(),
}
for host_path, remote_path in files
]
sandbox = self._sandbox
if sandbox is None:
raise RuntimeError("Vercel sandbox is not attached")
_retry_vercel_call(
"write_files",
lambda: sandbox.write_files(payload),
attempts=_WRITE_RETRY_ATTEMPTS,
)
def _vercel_delete(self, remote_paths: list[str]) -> None:
if not remote_paths:
return
sandbox = self._sandbox
if sandbox is None:
raise RuntimeError("Vercel sandbox is not attached")
result = sandbox.run_command(
"bash",
["-lc", quoted_rm_command(remote_paths)],
cwd=self._workspace_root,
)
if _extract_result_returncode(result) != 0:
raise RuntimeError(
f"Vercel delete failed: {_extract_result_output(result).strip()}"
)
def _vercel_bulk_download(self, dest_tar_path: Path) -> None:
remote_hermes = (
"/.hermes"
if self._remote_home == "/"
else f"{self._remote_home.rstrip('/')}/.hermes"
)
archive_member = remote_hermes.lstrip("/")
remote_tar = f"/tmp/.hermes_sync.{os.getpid()}.tar"
sandbox = self._sandbox
if sandbox is None:
raise RuntimeError("Vercel sandbox is not attached")
try:
result = sandbox.run_command(
"bash",
[
"-lc",
f"tar cf {shlex.quote(remote_tar)} -C / {shlex.quote(archive_member)}",
],
cwd=self._workspace_root,
)
if _extract_result_returncode(result) != 0:
raise RuntimeError(
f"Vercel bulk download failed: {_extract_result_output(result).strip()}"
)
sandbox.download_file(remote_tar, dest_tar_path)
finally:
try:
sandbox.run_command(
"bash",
["-lc", f"rm -f {shlex.quote(remote_tar)}"],
cwd=self._workspace_root,
)
except Exception:
pass
def _before_execute(self) -> None:
with self._lock:
self._ensure_sandbox_ready()
if self._sync_manager is not None:
self._sync_manager.sync()
def _run_bash(
self,
cmd_string: str,
*,
login: bool = False,
timeout: int = 120,
stdin_data: str | None = None,
):
"""Run a bash command in the Vercel sandbox.
``timeout`` is not forwarded to the Vercel SDK (which does not expose
a per-exec timeout parameter); the base class ``_wait_for_process``
enforces timeout by killing the sandbox via ``cancel_fn``.
``stdin_data`` is intentionally discarded here because
``_stdin_mode = "heredoc"`` causes the base class ``execute()`` to
embed any stdin payload into the command string before calling this
method.
"""
del timeout
del stdin_data
sandbox = self._sandbox
if sandbox is None:
raise RuntimeError("Vercel sandbox is not attached")
workspace_root = self._workspace_root
lock = self._lock
def cancel() -> None:
with lock:
self._stop_sandbox(sandbox)
def exec_fn() -> tuple[str, int]:
result = sandbox.run_command(
"bash",
["-lc" if login else "-c", cmd_string],
cwd=workspace_root,
)
return _extract_result_output(result), _extract_result_returncode(result)
return _ThreadedProcessHandle(exec_fn, cancel_fn=cancel)
def cleanup(self):
with self._lock:
sandbox = self._sandbox
sync_manager = self._sync_manager
if sandbox is not None and sync_manager is not None:
try:
sync_manager.sync_back()
except Exception as exc:
logger.warning(
"Vercel: sync_back failed for task %s: %s",
self._task_id,
exc,
)
self._sandbox = None
self._sync_manager = None
if sandbox is None:
return
snapshot_id = self._snapshot_sandbox(sandbox)
# Always stop the sandbox during cleanup to avoid resource leaks,
# matching the Modal and Daytona patterns.
self._stop_sandbox(sandbox)
self._close_sandbox_client(sandbox)

2
tools/file_operations.py
View file

@ -3,7 +3,7 @@
File Operations Module
Provides file manipulation capabilities (read, write, patch, search) that work
across all terminal backends (local, docker, ssh, singularity, modal, daytona, vercel_sandbox).
across all terminal backends (local, docker, ssh, singularity, modal, daytona).
The key insight is that all file operations can be expressed as shell commands,
so we wrap the terminal backend's execute() interface to provide a unified file API.

3
tools/file_tools.py
View file

@ -467,13 +467,12 @@ def _get_file_ops(task_id: str = "default") -> ShellFileOperations:
logger.info("Creating new %s environment for task %s...", env_type, task_id[:8])
container_config = None
if env_type in {"docker", "singularity", "modal", "daytona", "vercel_sandbox"}:
if env_type in {"docker", "singularity", "modal", "daytona"}:
container_config = {
"container_cpu": config.get("container_cpu", 1),
"container_memory": config.get("container_memory", 5120),
"container_disk": config.get("container_disk", 51200),
"container_persistent": config.get("container_persistent", True),
"vercel_runtime": config.get("vercel_runtime", ""),
"docker_volumes": config.get("docker_volumes", []),
"docker_mount_cwd_to_workspace": config.get("docker_mount_cwd_to_workspace", False),
"docker_forward_env": config.get("docker_forward_env", []),

1
tools/lazy_deps.py
View file

@ -156,7 +156,6 @@ LAZY_DEPS: dict[str, tuple[str, ...]] = {
# ─── Terminal backends ─────────────────────────────────────────────────
"terminal.modal": ("modal==1.3.4",),
"terminal.daytona": ("daytona==0.155.0",),
"terminal.vercel": ("vercel==0.5.7",),
# ─── Skills ────────────────────────────────────────────────────────────
"skill.google_workspace": (

2
tools/skills_tool.py
View file

@ -103,7 +103,7 @@ _PLATFORM_MAP = {
}
_ENV_VAR_NAME_RE = re.compile(r"^[A-Za-z_][A-Za-z0-9_]*$")
_REMOTE_ENV_BACKENDS = frozenset(
{"docker", "singularity", "modal", "ssh", "daytona", "vercel_sandbox"}
{"docker", "singularity", "modal", "ssh", "daytona"}
)
_secret_capture_callback = None

121
tools/terminal_tool.py
View file

@ -3,18 +3,16 @@
Terminal Tool Module
A terminal tool that executes commands in local, Docker, Modal, SSH,
Singularity, Daytona, and Vercel Sandbox environments. Supports local
execution, containerized backends, and cloud sandboxes, including managed
Modal mode.
Singularity, and Daytona environments. Supports local execution,
containerized backends, and cloud sandboxes, including managed Modal mode.
Environment Selection (via TERMINAL_ENV environment variable):
Supported environments:
- "local": Execute directly on the host machine (default, fastest)
- "docker": Execute in Docker containers (isolated, requires Docker)
- "modal": Execute in Modal cloud sandboxes (direct Modal or managed gateway)
- "vercel_sandbox": Execute in Vercel Sandbox cloud sandboxes
Features:
- Multiple execution backends (local, docker, modal, vercel_sandbox)
- Multiple execution backends (local, docker, modal)
- Background task support
- VM/container lifecycle management
- Automatic cleanup after inactivity
@ -119,68 +117,6 @@ DISK_USAGE_WARNING_THRESHOLD_GB = _safe_parse_import_env(
float,
"number",
)
_VERCEL_SANDBOX_DEFAULT_CWD = "/vercel/sandbox"
_SUPPORTED_VERCEL_RUNTIMES = ("node24", "node22", "python3.13")
def _is_supported_vercel_runtime(runtime: str) -> bool:
return not runtime or runtime in _SUPPORTED_VERCEL_RUNTIMES
def _check_vercel_sandbox_requirements(config: dict[str, Any]) -> bool:
"""Validate Vercel Sandbox terminal backend requirements."""
runtime = (config.get("vercel_runtime") or "").strip()
if not _is_supported_vercel_runtime(runtime):
supported = ", ".join(_SUPPORTED_VERCEL_RUNTIMES)
logger.error(
"Vercel Sandbox runtime %r is not supported. "
"Set TERMINAL_VERCEL_RUNTIME to one of: %s.",
runtime,
supported,
)
return False
disk = config.get("container_disk", 51200)
if disk not in {0, 51200}:
logger.error(
"Vercel Sandbox does not support custom TERMINAL_CONTAINER_DISK=%s. "
"Use the default shared setting (51200 MB).",
disk,
)
return False
if importlib.util.find_spec("vercel") is None:
logger.error(
"vercel is required for the Vercel Sandbox terminal backend: pip install vercel"
)
return False
has_oidc = bool(os.getenv("VERCEL_OIDC_TOKEN"))
has_token = bool(os.getenv("VERCEL_TOKEN"))
has_project = bool(os.getenv("VERCEL_PROJECT_ID"))
has_team = bool(os.getenv("VERCEL_TEAM_ID"))
if has_oidc:
return True
if has_token or has_project or has_team:
if has_token and has_project and has_team:
return True
logger.error(
"Vercel Sandbox backend selected with token auth, but "
"VERCEL_TOKEN, VERCEL_PROJECT_ID, and VERCEL_TEAM_ID must all "
"be set together. VERCEL_OIDC_TOKEN is supported for one-off "
"local development only."
)
return False
logger.error(
"Vercel Sandbox backend selected but no supported auth configuration "
"was found. Set VERCEL_TOKEN, VERCEL_PROJECT_ID, and VERCEL_TEAM_ID "
"for normal use. VERCEL_OIDC_TOKEN is supported for one-off local "
"development only."
)
return False
def _check_disk_usage_warning():
@ -837,10 +773,9 @@ def _transform_sudo_command(command: str | None) -> tuple[str | None, str | None
should prepend sudo_stdin to their stdin_data and pass the merged bytes to
Popen's stdin pipe.
Callers that cannot pipe subprocess stdin (modal, daytona,
vercel_sandbox) must embed the password in the command string
themselves; see their execute() methods for how they handle the
non-None sudo_stdin case.
Callers that cannot pipe subprocess stdin (modal, daytona) must embed
the password in the command string themselves; see their execute()
methods for how they handle the non-None sudo_stdin case.
If SUDO_PASSWORD is not set and in interactive mode (HERMES_INTERACTIVE=1):
Prompts user for password with 45s timeout, caches for session.
@ -1015,14 +950,12 @@ def _get_env_config() -> Dict[str, Any]:
mount_docker_cwd = os.getenv("TERMINAL_DOCKER_MOUNT_CWD_TO_WORKSPACE", "false").lower() in {"true", "1", "yes"}
# Default cwd: local uses the host's current directory, ssh uses the
# remote home, Vercel uses its documented workspace root, and everything
# else starts in the backend's default root-like cwd.
# remote home, and everything else starts in the backend's default
# root-like cwd.
if env_type == "local":
default_cwd = os.getcwd()
elif env_type == "ssh":
default_cwd = "~"
elif env_type == "vercel_sandbox":
default_cwd = _VERCEL_SANDBOX_DEFAULT_CWD
else:
default_cwd = "/root"
@ -1044,7 +977,7 @@ def _get_env_config() -> Dict[str, Any]:
):
host_cwd = candidate
cwd = "/workspace"
elif env_type in {"modal", "docker", "singularity", "daytona", "vercel_sandbox"} and cwd:
elif env_type in {"modal", "docker", "singularity", "daytona"} and cwd:
# Host paths and relative paths that won't work inside containers
is_host_path = any(cwd.startswith(p) for p in host_prefixes)
is_relative = not os.path.isabs(cwd) # e.g. "." or "src/"
@ -1062,7 +995,6 @@ def _get_env_config() -> Dict[str, Any]:
"singularity_image": os.getenv("TERMINAL_SINGULARITY_IMAGE", f"docker://{default_image}"),
"modal_image": os.getenv("TERMINAL_MODAL_IMAGE", default_image),
"daytona_image": os.getenv("TERMINAL_DAYTONA_IMAGE", default_image),
"vercel_runtime": os.getenv("TERMINAL_VERCEL_RUNTIME", "").strip(),
"cwd": cwd,
"host_cwd": host_cwd,
"docker_mount_cwd_to_workspace": mount_docker_cwd,
@ -1082,7 +1014,7 @@ def _get_env_config() -> Dict[str, Any]:
).lower() in {"true", "1", "yes"},
"local_persistent": os.getenv("TERMINAL_LOCAL_PERSISTENT", "false").lower() in {"true", "1", "yes"},
# Container resource config (applies to docker, singularity, modal,
# daytona, and vercel_sandbox -- ignored for local/ssh)
# daytona -- ignored for local/ssh)
"container_cpu": _parse_env_var("TERMINAL_CONTAINER_CPU", "1", float, "number"),
"container_memory": _parse_env_var("TERMINAL_CONTAINER_MEMORY", "5120"), # MB (default 5GB)
"container_disk": _parse_env_var("TERMINAL_CONTAINER_DISK", "51200"), # MB (default 50GB)
@ -1113,8 +1045,8 @@ def _create_environment(env_type: str, image: str, cwd: str, timeout: int,
Args:
env_type: One of "local", "docker", "singularity", "modal",
"daytona", "vercel_sandbox", "ssh"
image: Docker/Singularity/Modal image name (ignored for local/ssh/vercel)
"daytona", "ssh"
image: Docker/Singularity/Modal image name (ignored for local/ssh)
cwd: Working directory
timeout: Default command timeout
ssh_config: SSH connection config (for env_type="ssh")
@ -1220,21 +1152,6 @@ def _create_environment(env_type: str, image: str, cwd: str, timeout: int,
persistent_filesystem=persistent, task_id=task_id,
)
elif env_type == "vercel_sandbox":
from tools.environments.vercel_sandbox import (
VercelSandboxEnvironment as _VercelSandboxEnvironment,
)
return _VercelSandboxEnvironment(
runtime=cc.get("vercel_runtime") or None,
cwd=cwd,
timeout=timeout,
cpu=cpu,
memory=memory,
disk=disk,
persistent_filesystem=persistent,
task_id=task_id,
)
elif env_type == "ssh":
if not ssh_config or not ssh_config.get("host") or not ssh_config.get("user"):
raise ValueError("SSH environment requires ssh_host and ssh_user to be configured")
@ -1250,7 +1167,7 @@ def _create_environment(env_type: str, image: str, cwd: str, timeout: int,
else:
raise ValueError(
f"Unknown environment type: {env_type}. Use 'local', 'docker', "
f"'singularity', 'modal', 'daytona', 'vercel_sandbox', or 'ssh'"
f"'singularity', 'modal', 'daytona', or 'ssh'"
)
@ -1809,14 +1726,13 @@ def terminal_tool(
}
container_config = None
if env_type in {"docker", "singularity", "modal", "daytona", "vercel_sandbox"}:
if env_type in {"docker", "singularity", "modal", "daytona"}:
container_config = {
"container_cpu": config.get("container_cpu", 1),
"container_memory": config.get("container_memory", 5120),
"container_disk": config.get("container_disk", 51200),
"container_persistent": config.get("container_persistent", True),
"modal_mode": config.get("modal_mode", "auto"),
"vercel_runtime": config.get("vercel_runtime", ""),
"docker_volumes": config.get("docker_volumes", []),
"docker_mount_cwd_to_workspace": config.get("docker_mount_cwd_to_workspace", False),
"docker_forward_env": config.get("docker_forward_env", []),
@ -2265,9 +2181,6 @@ def check_terminal_requirements() -> bool:
return True
elif env_type == "vercel_sandbox":
return _check_vercel_sandbox_requirements(config)
elif env_type == "daytona":
from daytona import Daytona # noqa: F401 — SDK presence check
return os.getenv("DAYTONA_API_KEY") is not None
@ -2275,7 +2188,7 @@ def check_terminal_requirements() -> bool:
else:
logger.error(
"Unknown TERMINAL_ENV '%s'. Use one of: local, docker, singularity, "
"modal, daytona, vercel_sandbox, ssh.",
"modal, daytona, ssh.",
env_type,
)
return False
@ -2318,7 +2231,7 @@ if __name__ == "__main__":
print(
" TERMINAL_ENV: "
f"{os.getenv('TERMINAL_ENV', 'local')} "
"(local/docker/singularity/modal/daytona/vercel_sandbox/ssh)"
"(local/docker/singularity/modal/daytona/ssh)"
)
print(f" TERMINAL_DOCKER_IMAGE: {os.getenv('TERMINAL_DOCKER_IMAGE', default_img)}")
print(f" TERMINAL_SINGULARITY_IMAGE: {os.getenv('TERMINAL_SINGULARITY_IMAGE', f'docker://{default_img}')}")

39
uv.lock generated
View file

@ -1760,9 +1760,6 @@ termux-all = [
tts-premium = [
{ name = "elevenlabs" },
]
vercel = [
{ name = "vercel" },
]
voice = [
{ name = "faster-whisper" },
{ name = "numpy" },
@ -1877,10 +1874,9 @@ requires-dist = [
{ name = "ty", marker = "extra == 'dev'", specifier = "==0.0.21" },
{ name = "tzdata", marker = "sys_platform == 'win32'", specifier = "==2025.3" },
{ name = "uvicorn", extras = ["standard"], marker = "extra == 'web'", specifier = "==0.41.0" },
{ name = "vercel", marker = "extra == 'vercel'", specifier = "==0.5.7" },
{ name = "youtube-transcript-api", marker = "extra == 'youtube'", specifier = "==1.2.4" },
]
provides-extras = ["anthropic", "exa", "firecrawl", "parallel-web", "fal", "edge-tts", "modal", "daytona", "vercel", "hindsight", "dev", "messaging", "cron", "slack", "matrix", "wecom", "cli", "tts-premium", "voice", "pty", "honcho", "mcp", "homeassistant", "sms", "computer-use", "acp", "bedrock", "azure-identity", "termux", "termux-all", "dingtalk", "feishu", "google", "youtube", "web", "all"]
provides-extras = ["anthropic", "exa", "firecrawl", "parallel-web", "fal", "edge-tts", "modal", "daytona", "hindsight", "dev", "messaging", "cron", "slack", "matrix", "wecom", "cli", "tts-premium", "voice", "pty", "honcho", "mcp", "homeassistant", "sms", "computer-use", "acp", "bedrock", "azure-identity", "termux", "termux-all", "dingtalk", "feishu", "google", "youtube", "web", "all"]
[[package]]
name = "hf-xet"
@ -4370,39 +4366,6 @@ wheels = [
{ url = "https://files.pythonhosted.org/packages/e4/16/c1fd27e9549f3c4baf1dc9c20c456cd2f822dbf8de9f463824b0c0357e06/uvloop-0.22.1-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:6cde23eeda1a25c75b2e07d39970f3374105d5eafbaab2a4482be82f272d5a5e", size = 4296730, upload-time = "2025-10-16T22:17:00.744Z" },
]
[[package]]
name = "vercel"
version = "0.5.7"
source = { registry = "https://pypi.org/simple" }
dependencies = [
{ name = "anyio" },
{ name = "cbor2" },
{ name = "httpx" },
{ name = "pydantic" },
{ name = "python-dotenv" },
{ name = "vercel-workers", marker = "python_full_version >= '3.12'" },
{ name = "websockets" },
]
sdist = { url = "https://files.pythonhosted.org/packages/d7/68/a671ebc656afbb5e25fb88c681b61511cc13670ea771c87b2f711782022b/vercel-0.5.7.tar.gz", hash = "sha256:8070ea1b33962adfed98498f9273f24ea2066a20c74d38643d479d8280801c6e", size = 118597, upload-time = "2026-04-15T17:58:20.424Z" }
wheels = [
{ url = "https://files.pythonhosted.org/packages/c7/2e/bacf1ccc0ec95464a68398e64bf5e36f859cd51f3e379623f103802f85f1/vercel-0.5.7-py3-none-any.whl", hash = "sha256:90eb2689c34e403db2170fec3eb47e1a91092c200d91baf4b4501fb3e2a44d28", size = 139698, upload-time = "2026-04-15T17:58:18.945Z" },
]
[[package]]
name = "vercel-workers"
version = "0.0.16"
source = { registry = "https://pypi.org/simple" }
dependencies = [
{ name = "anyio", marker = "python_full_version >= '3.12'" },
{ name = "httpx", marker = "python_full_version >= '3.12'" },
{ name = "python-dotenv", marker = "python_full_version >= '3.12'" },
{ name = "vercel", marker = "python_full_version >= '3.12'" },
]
sdist = { url = "https://files.pythonhosted.org/packages/73/d8/17ba256fceff42be231ca8ff0567dcf2da54ee8de633e949fa08b9403b1f/vercel_workers-0.0.16.tar.gz", hash = "sha256:38df45dbf42fbae39ffa0e419f0908bf1beb047e38fc5ddd0a479feac340fb8c", size = 51615, upload-time = "2026-04-13T21:23:27.649Z" }
wheels = [
{ url = "https://files.pythonhosted.org/packages/65/3a/0137d5b157845e1d41a70130d8dce8ba15d8712f34619693cda04ecb8f02/vercel_workers-0.0.16-py3-none-any.whl", hash = "sha256:542be839e46e236a68cc308695ccc3c970d76de72c978d7f416cc6ce09688896", size = 50141, upload-time = "2026-04-13T21:23:28.652Z" },
]
[[package]]
name = "watchfiles"
version = "1.1.1"

2
website/docs/developer-guide/architecture.md
View file

@ -211,7 +211,7 @@ A shared runtime resolver used by CLI, gateway, cron, ACP, and auxiliary calls.
### Tool System
Central tool registry (`tools/registry.py`) with 70+ registered tools across ~28 toolsets. Each tool file self-registers at import time. The registry handles schema collection, dispatch, availability checking, and error wrapping. Terminal tools support 7 backends (local, Docker, SSH, Daytona, Modal, Singularity, Vercel Sandbox).
Central tool registry (`tools/registry.py`) with 70+ registered tools across ~28 toolsets. Each tool file self-registers at import time. The registry handles schema collection, dispatch, availability checking, and error wrapping. Terminal tools support 6 backends (local, Docker, SSH, Daytona, Modal, Singularity).
→ [Tools Runtime](./tools-runtime.md)

12
website/docs/developer-guide/provider-runtime.md
View file

@ -42,7 +42,6 @@ That ordering matters because Hermes treats the saved model/provider choice as t
Current provider families include (see `plugins/model-providers/` for the complete bundled set):
- AI Gateway (Vercel)
- OpenRouter
- Nous Portal
- OpenAI Codex
@ -93,18 +92,13 @@ This resolver is the main reason Hermes can share auth/runtime logic between:
- ACP editor sessions
- auxiliary model tasks
## AI Gateway
## OpenRouter and custom OpenAI-compatible base URLs
Set `AI_GATEWAY_API_KEY` in `~/.hermes/.env` and run with `--provider ai-gateway`. Hermes fetches available models from the gateway's `/models` endpoint, filtering to language models with tool-use support.
## OpenRouter, AI Gateway, and custom OpenAI-compatible base URLs
Hermes contains logic to avoid leaking the wrong API key to a custom endpoint when multiple provider keys exist (e.g. `OPENROUTER_API_KEY`, `AI_GATEWAY_API_KEY`, and `OPENAI_API_KEY`).
Hermes contains logic to avoid leaking the wrong API key to a custom endpoint when multiple provider keys exist (e.g. `OPENROUTER_API_KEY` and `OPENAI_API_KEY`).
Each provider's API key is scoped to its own base URL:
- `OPENROUTER_API_KEY` is only sent to `openrouter.ai` endpoints
- `AI_GATEWAY_API_KEY` is only sent to `ai-gateway.vercel.sh` endpoints
- `OPENAI_API_KEY` is used for custom endpoints and as a fallback
Hermes also distinguishes between:
@ -115,7 +109,7 @@ Hermes also distinguishes between:
That distinction is especially important for:
- local model servers
- non-OpenRouter/non-AI Gateway OpenAI-compatible APIs
- non-OpenRouter OpenAI-compatible APIs
- switching providers without re-running setup
- config-saved custom endpoints that should keep working even when `OPENAI_BASE_URL` is not exported in the current shell

1
website/docs/developer-guide/tools-runtime.md
View file

@ -213,7 +213,6 @@ The terminal system supports multiple backends:
- singularity
- modal
- daytona
- vercel_sandbox
It also supports:

1
website/docs/getting-started/quickstart.md
View file

@ -124,7 +124,6 @@ Good defaults:
| **NVIDIA NIM** | Nemotron models via build.nvidia.com or local NIM | Set `NVIDIA_API_KEY` (optional: `NVIDIA_BASE_URL`) |
| **GitHub Copilot** | GitHub Copilot subscription (GPT-5.x, Claude, Gemini, etc.) | OAuth via `hermes model`, or `COPILOT_GITHUB_TOKEN` / `GH_TOKEN` |
| **GitHub Copilot ACP** | Copilot ACP agent backend (spawns local `copilot` CLI) | `hermes model` (requires `copilot` CLI + `copilot login`) |
| **Vercel AI Gateway** | Vercel AI Gateway routing | Set `AI_GATEWAY_API_KEY` |
| **Custom Endpoint** | VLLM, SGLang, Ollama, or any OpenAI-compatible API | Set base URL + API key |
For most first-time users: choose a provider, accept the defaults unless you know why you're changing them. The full provider catalog with env vars and setup steps lives on the [Providers](../integrations/providers.md) page.

3
website/docs/integrations/providers.md
View file

@ -21,7 +21,6 @@ You need at least one way to connect to an LLM. Use `hermes model` to switch pro
| **Anthropic** | `hermes model` (Claude Max + extra usage credits via OAuth; also supports Anthropic API key or manual setup-token — see note below) |
| **OpenRouter** | `OPENROUTER_API_KEY` in `~/.hermes/.env` |
| **NovitaAI** | `NOVITA_API_KEY` in `~/.hermes/.env` (provider: `novita`, 200+ models, Model API, Agent Sandbox, GPU Cloud) |
| **AI Gateway** | `AI_GATEWAY_API_KEY` in `~/.hermes/.env` (provider: `ai-gateway`) |
| **z.ai / GLM** | `GLM_API_KEY` in `~/.hermes/.env` (provider: `zai`) |
| **Kimi / Moonshot** | `KIMI_API_KEY` in `~/.hermes/.env` (provider: `kimi-coding`) |
| **Kimi / Moonshot (China)** | `KIMI_CN_API_KEY` in `~/.hermes/.env` (provider: `kimi-coding-cn`; aliases: `kimi-cn`, `moonshot-cn`) |
@ -1490,7 +1489,7 @@ fallback_model:
When activated, the fallback swaps the model and provider mid-session without losing your conversation. The chain is tried entry-by-entry; activation is one-shot per session.
Supported providers: `openrouter`, `nous`, `openai-codex`, `copilot`, `copilot-acp`, `anthropic`, `gemini`, `google-gemini-cli`, `qwen-oauth`, `huggingface`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth`, `deepseek`, `nvidia`, `xai`, `xai-oauth`, `ollama-cloud`, `bedrock`, `ai-gateway`, `azure-foundry`, `opencode-zen`, `opencode-go`, `kilocode`, `xiaomi`, `arcee`, `gmi`, `stepfun`, `lmstudio`, `alibaba`, `alibaba-coding-plan`, `tencent-tokenhub`, `custom`.
Supported providers: `openrouter`, `nous`, `openai-codex`, `copilot`, `copilot-acp`, `anthropic`, `gemini`, `google-gemini-cli`, `qwen-oauth`, `huggingface`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth`, `deepseek`, `nvidia`, `xai`, `xai-oauth`, `ollama-cloud`, `bedrock`, `azure-foundry`, `opencode-zen`, `opencode-go`, `kilocode`, `xiaomi`, `arcee`, `gmi`, `stepfun`, `lmstudio`, `alibaba`, `alibaba-coding-plan`, `tencent-tokenhub`, `custom`.
:::tip
Fallback is configured exclusively through `config.yaml` — or interactively via `hermes fallback`. For full details on when it triggers, how the chain advances, and how it interacts with auxiliary tasks and delegation, see [Fallback Providers](/user-guide/features/fallback-providers).

2
website/docs/reference/cli-commands.md
View file

@ -95,7 +95,7 @@ Common options:
| `-q`, `--query "..."` | One-shot, non-interactive prompt. |
| `-m`, `--model <model>` | Override the model for this run. |
| `-t`, `--toolsets <csv>` | Enable a comma-separated set of toolsets. |
| `--provider <provider>` | Force a provider: `auto`, `openrouter`, `nous`, `openai-codex`, `copilot-acp`, `copilot`, `anthropic`, `gemini`, `google-gemini-cli`, `huggingface`, `novita`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth`, `kilocode`, `xiaomi`, `arcee`, `gmi`, `alibaba`, `alibaba-coding-plan` (alias `alibaba_coding`), `deepseek`, `nvidia`, `ollama-cloud`, `xai` (alias `grok`), `xai-oauth` (alias `grok-oauth`), `qwen-oauth`, `bedrock`, `opencode-zen`, `opencode-go`, `ai-gateway`, `azure-foundry`, `lmstudio`, `stepfun`, `tencent-tokenhub` (alias `tencent`, `tokenhub`). |
| `--provider <provider>` | Force a provider: `auto`, `openrouter`, `nous`, `openai-codex`, `copilot-acp`, `copilot`, `anthropic`, `gemini`, `google-gemini-cli`, `huggingface`, `novita`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth`, `kilocode`, `xiaomi`, `arcee`, `gmi`, `alibaba`, `alibaba-coding-plan` (alias `alibaba_coding`), `deepseek`, `nvidia`, `ollama-cloud`, `xai` (alias `grok`), `xai-oauth` (alias `grok-oauth`), `qwen-oauth`, `bedrock`, `opencode-zen`, `opencode-go`, `azure-foundry`, `lmstudio`, `stepfun`, `tencent-tokenhub` (alias `tencent`, `tokenhub`). |
| `-s`, `--skills <name>` | Preload one or more skills for the session (can be repeated or comma-separated). |
| `-v`, `--verbose` | Verbose output. |
| `-Q`, `--quiet` | Programmatic mode: suppress banner/spinner/tool previews. |

9
website/docs/reference/environment-variables.md
View file

@ -18,8 +18,6 @@ All variables go in `~/.hermes/.env`. You can also set them with `hermes config
| `HERMES_OPENROUTER_CACHE_TTL` | Cache TTL in seconds (1-86400). Overrides `openrouter.response_cache_ttl` in config.yaml. |
| `NOUS_BASE_URL` | Override Nous Portal base URL (rarely needed; development/testing only) |
| `NOUS_INFERENCE_BASE_URL` | Override Nous inference endpoint directly |
| `AI_GATEWAY_API_KEY` | Vercel AI Gateway API key ([ai-gateway.vercel.sh](https://ai-gateway.vercel.sh)) |
| `AI_GATEWAY_BASE_URL` | Override AI Gateway base URL (default: `https://ai-gateway.vercel.sh/v1`) |
| `OPENAI_API_KEY` | API key for custom OpenAI-compatible endpoints (used with `OPENAI_BASE_URL`) |
| `OPENAI_BASE_URL` | Base URL for custom endpoint (VLLM, SGLang, etc.) |
| `COPILOT_GITHUB_TOKEN` | GitHub token for Copilot API — first priority (OAuth `gho_*` or fine-grained PAT `github_pat_*`; classic PATs `ghp_*` are **not supported**) |
@ -156,10 +154,6 @@ For native Anthropic auth, Hermes prefers Claude Code's own credential files whe
| `HINDSIGHT_TIMEOUT` | Timeout in seconds for Hindsight memory-provider API calls (default: `60`). Bump this if your Hindsight instance is slow to respond during `/sync` or `on_session_switch` and you're seeing timeouts in `errors.log`. |
| `SUPERMEMORY_API_KEY` | Semantic long-term memory with profile recall and session ingest ([supermemory.ai](https://supermemory.ai)) |
| `DAYTONA_API_KEY` | Daytona cloud sandboxes ([daytona.io](https://daytona.io/)) |
| `VERCEL_TOKEN` | Vercel Sandbox access token ([vercel.com](https://vercel.com/)) |
| `VERCEL_PROJECT_ID` | Vercel project ID (required with `VERCEL_TOKEN`) |
| `VERCEL_TEAM_ID` | Vercel team ID (required with `VERCEL_TOKEN`) |
| `VERCEL_OIDC_TOKEN` | Vercel short-lived OIDC token (development-only alternative) |
### Langfuse Observability
@ -192,7 +186,7 @@ These variables configure the [Tool Gateway](/user-guide/features/tool-gateway)
| Variable | Description |
|----------|-------------|
| `TERMINAL_ENV` | Backend: `local`, `docker`, `ssh`, `singularity`, `modal`, `daytona`, `vercel_sandbox` |
| `TERMINAL_ENV` | Backend: `local`, `docker`, `ssh`, `singularity`, `modal`, `daytona` |
| `HERMES_DOCKER_BINARY` | Override the container binary Hermes shells out to (e.g. `podman`, `/usr/local/bin/docker`). When unset, Hermes auto-discovers `docker` or `podman` on `PATH`. Needed when both are installed and you want the non-default, or when the binary lives outside `PATH`. |
| `TERMINAL_DOCKER_IMAGE` | Docker image (default: `nikolaik/python-nodejs:python3.11-nodejs20`) |
| `TERMINAL_DOCKER_FORWARD_ENV` | JSON array of env var names to explicitly forward into Docker terminal sessions. Note: skill-declared `required_environment_variables` are forwarded automatically — you only need this for vars not declared by any skill. |
@ -201,7 +195,6 @@ These variables configure the [Tool Gateway](/user-guide/features/tool-gateway)
| `TERMINAL_SINGULARITY_IMAGE` | Singularity image or `.sif` path |
| `TERMINAL_MODAL_IMAGE` | Modal container image |
| `TERMINAL_DAYTONA_IMAGE` | Daytona sandbox image |
| `TERMINAL_VERCEL_RUNTIME` | Vercel Sandbox runtime (`node24`, `node22`, `python3.13`) |
| `TERMINAL_TIMEOUT` | Command timeout in seconds |
| `TERMINAL_LIFETIME_SECONDS` | Max lifetime for terminal sessions in seconds |
| `TERMINAL_CWD` | Working directory for terminal sessions (gateway/cron only; CLI uses launch dir) |

52
website/docs/user-guide/configuration.md
View file

@ -83,11 +83,11 @@ Leaving these unset keeps the legacy defaults (`HERMES_API_TIMEOUT=1800`s, `HERM
## Terminal Backend Configuration
Hermes supports seven terminal backends. Each determines where the agent's shell commands actually execute — your local machine, a Docker container, a remote server via SSH, a Modal cloud sandbox (direct or via the Nous-managed gateway), a Daytona workspace, a Vercel Sandbox, or a Singularity/Apptainer container.
Hermes supports six terminal backends. Each determines where the agent's shell commands actually execute — your local machine, a Docker container, a remote server via SSH, a Modal cloud sandbox (direct or via the Nous-managed gateway), a Daytona workspace, or a Singularity/Apptainer container.
```yaml
terminal:
backend: local # local | docker | ssh | modal | daytona | vercel_sandbox | singularity
backend: local # local | docker | ssh | modal | daytona | singularity
cwd: "." # Gateway/cron working directory (CLI always uses launch dir)
timeout: 180 # Per-command timeout in seconds
env_passthrough: [] # Env var names to forward to sandboxed execution (terminal + execute_code)
@ -96,7 +96,7 @@ terminal:
daytona_image: "nikolaik/python-nodejs:python3.11-nodejs20" # Container image for Daytona backend
```
For cloud sandboxes such as Modal, Daytona, and Vercel Sandbox, `container_persistent: true` means Hermes will try to preserve filesystem state across sandbox recreation. It does not promise that the same live sandbox, PID space, or background processes will still be running later.
For cloud sandboxes such as Modal and Daytona, `container_persistent: true` means Hermes will try to preserve filesystem state across sandbox recreation. It does not promise that the same live sandbox, PID space, or background processes will still be running later.
### Backend Overview
@ -107,7 +107,6 @@ For cloud sandboxes such as Modal, Daytona, and Vercel Sandbox, `container_persi
| **ssh** | Remote server via SSH | Network boundary | Remote dev, powerful hardware |
| **modal** | Modal cloud sandbox | Full (cloud VM) | Ephemeral cloud compute, evals |
| **daytona** | Daytona workspace | Full (cloud container) | Managed cloud dev environments |
| **vercel_sandbox** | Vercel Sandbox | Full (cloud microVM) | Cloud execution with snapshot-backed filesystem persistence |
| **singularity** | Singularity/Apptainer container | Namespaces (--containall) | HPC clusters, shared machines |
### Local Backend
@ -232,49 +231,6 @@ terminal:
**Disk limit:** Daytona enforces a 10 GiB maximum. Requests above this are capped with a warning.
### Vercel Sandbox Backend
Runs commands in a [Vercel Sandbox](https://vercel.com/docs/vercel-sandbox) cloud microVM. Hermes uses the normal terminal and file tool surfaces; there are no Vercel-specific model-facing tools.
```yaml
terminal:
backend: vercel_sandbox
vercel_runtime: node24 # node24 | node22 | python3.13
cwd: /vercel/sandbox # default workspace root
container_persistent: true # Snapshot/restore filesystem
container_disk: 51200 # Shared default only; custom disk is unsupported
```
**Required install:** Install the optional SDK extra:
```bash
pip install 'hermes-agent[vercel]'
```
**Required authentication:** Configure access-token auth with all three of `VERCEL_TOKEN`, `VERCEL_PROJECT_ID`, and `VERCEL_TEAM_ID`. This is the supported setup for deployments and normal long-running Hermes processes on Render, Railway, Docker, and similar hosts.
For one-off local development, Hermes also accepts short-lived Vercel OIDC tokens:
```bash
VERCEL_OIDC_TOKEN="$(vc project token <project-name>)" hermes chat
```
From a linked Vercel project directory, you can omit the project name:
```bash
VERCEL_OIDC_TOKEN="$(vc project token)" hermes chat
```
OIDC tokens are short-lived and should not be used as the documented deployment path.
**Runtime:** `terminal.vercel_runtime` supports `node24`, `node22`, and `python3.13`. If unset, Hermes defaults to `node24`.
**Persistence:** When `container_persistent: true`, Hermes snapshots the sandbox filesystem during cleanup and restores a later sandbox for the same task from that snapshot. Snapshot contents can include Hermes-synced credentials, skills, and cache files that were copied into the sandbox. This preserves filesystem state only; it does not preserve live sandbox identity, PID space, shell state, or running background processes.
**Background commands:** `terminal(background=true)` uses Hermes' generic non-local background process flow. You can spawn, poll, wait, view logs, and kill processes through the normal process tool while the sandbox is alive. Hermes does not provide native Vercel detached-process recovery after cleanup or restart.
**Disk sizing:** Vercel Sandbox does not currently support Hermes' `container_disk` resource knob. Leave `container_disk` unset or at the shared default `51200`; non-default values fail diagnostics and backend creation instead of being silently ignored.
### Singularity/Apptainer Backend
Runs commands in a [Singularity/Apptainer](https://apptainer.org) container. Designed for HPC clusters and shared machines where Docker isn't available.
@ -829,7 +785,7 @@ Every model slot in Hermes — auxiliary tasks, compression, fallback — uses t
When `base_url` is set, Hermes ignores the provider and calls that endpoint directly (using `api_key` or `OPENAI_API_KEY` for auth). When only `provider` is set, Hermes uses that provider's built-in auth and base URL.
Available providers for auxiliary tasks: `auto`, `main`, plus any provider in the [provider registry](/reference/environment-variables) — `openrouter`, `nous`, `openai-codex`, `copilot`, `copilot-acp`, `anthropic`, `gemini`, `google-gemini-cli`, `qwen-oauth`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth`, `deepseek`, `nvidia`, `xai`, `xai-oauth`, `ollama-cloud`, `alibaba`, `bedrock`, `huggingface`, `arcee`, `xiaomi`, `kilocode`, `opencode-zen`, `opencode-go`, `ai-gateway`, `azure-foundry` — or any named custom provider from your `custom_providers` list (e.g. `provider: "beans"`).
Available providers for auxiliary tasks: `auto`, `main`, plus any provider in the [provider registry](/reference/environment-variables) — `openrouter`, `nous`, `openai-codex`, `copilot`, `copilot-acp`, `anthropic`, `gemini`, `google-gemini-cli`, `qwen-oauth`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth`, `deepseek`, `nvidia`, `xai`, `xai-oauth`, `ollama-cloud`, `alibaba`, `bedrock`, `huggingface`, `arcee`, `xiaomi`, `kilocode`, `opencode-zen`, `opencode-go`, `azure-foundry` — or any named custom provider from your `custom_providers` list (e.g. `provider: "beans"`).
:::tip MiniMax OAuth
`minimax-oauth` logs in via browser OAuth (no API key needed). Run `hermes model` and select **MiniMax (OAuth)** to authenticate. Auxiliary tasks use `MiniMax-M2.7-highspeed` automatically. See the [MiniMax OAuth guide](../guides/minimax-oauth.md).

1
website/docs/user-guide/features/fallback-providers.md
View file

@ -47,7 +47,6 @@ Both `provider` and `model` are **required**. If either is missing, the fallback
| Provider | Value | Requirements |
|----------|-------|-------------|
| AI Gateway | `ai-gateway` | `AI_GATEWAY_API_KEY` |
| OpenRouter | `openrouter` | `OPENROUTER_API_KEY` |
| Nous Portal | `nous` | `hermes setup --portal` (fresh) or `hermes auth add nous` (OAuth) |
| OpenAI Codex | `openai-codex` | `hermes model` (ChatGPT OAuth) |

33
website/docs/user-guide/features/tools.md
View file

@ -65,14 +65,13 @@ The terminal tool can execute commands in different environments:
| `singularity` | HPC containers | Cluster computing, rootless |
| `modal` | Cloud execution | Serverless, scale |
| `daytona` | Cloud sandbox workspace | Persistent remote dev environments |
| `vercel_sandbox` | Vercel Sandbox cloud microVM | Cloud execution with snapshot-backed filesystem persistence |
### Configuration
```yaml
# In ~/.hermes/config.yaml
terminal:
backend: local # or: docker, ssh, singularity, modal, daytona, vercel_sandbox
backend: local # or: docker, ssh, singularity, modal, daytona
cwd: "." # Working directory
timeout: 180 # Command timeout in seconds
```
@ -123,41 +122,13 @@ modal setup
hermes config set terminal.backend modal
```
### Vercel Sandbox
```bash
pip install 'hermes-agent[vercel]'
hermes config set terminal.backend vercel_sandbox
hermes config set terminal.vercel_runtime node24
```
Authenticate with all three of `VERCEL_TOKEN`, `VERCEL_PROJECT_ID`, and `VERCEL_TEAM_ID`. This access-token setup is the supported path for deployments and normal long-running Hermes processes on Render, Railway, Docker, and similar hosts. Supported runtimes are `node24`, `node22`, and `python3.13`; Hermes defaults to `/vercel/sandbox` as the remote workspace root.
For one-off local development, Hermes also accepts short-lived Vercel OIDC tokens:
```bash
VERCEL_OIDC_TOKEN="$(vc project token <project-name>)" hermes chat
```
From a linked Vercel project directory:
```bash
VERCEL_OIDC_TOKEN="$(vc project token)" hermes chat
```
With `container_persistent: true`, Hermes uses Vercel snapshots to preserve filesystem state across sandbox recreation for the same task. This can include Hermes-synced credentials, skills, and cache files inside the sandbox. Snapshots do not preserve live processes, PID space, or the same live sandbox identity.
Background terminal commands use Hermes' generic non-local process flow: spawn, poll, wait, log, and kill work through the normal process tool while the sandbox is alive, but Hermes does not provide native Vercel detached-process recovery after cleanup or restart.
Leave `container_disk` unset or at the shared default `51200`; custom disk sizing is unsupported for Vercel Sandbox and will fail diagnostics/backend creation.
### Container Resources
Configure CPU, memory, disk, and persistence for all container backends:
```yaml
terminal:
backend: docker # or singularity, modal, daytona, vercel_sandbox
backend: docker # or singularity, modal, daytona
container_cpu: 1 # CPU cores (default: 1)
container_memory: 5120 # Memory in MB (default: 5GB)
container_disk: 51200 # Disk in MB (default: 50GB)

5
website/docs/user-guide/security.md
View file

@ -144,7 +144,7 @@ The following patterns trigger approval prompts (defined in `tools/approval.py`)
| `gateway run` with `&`/`disown`/`nohup`/`setsid` | Prevents starting gateway outside service manager |
:::info
**Container bypass**: When running in `docker`, `singularity`, `modal`, `daytona`, or `vercel_sandbox` backends, dangerous command checks are **skipped** because the container itself is the security boundary. Destructive commands inside a container can't harm the host.
**Container bypass**: When running in `docker`, `singularity`, `modal`, or `daytona` backends, dangerous command checks are **skipped** because the container itself is the security boundary. Destructive commands inside a container can't harm the host.
:::
### Approval Flow (CLI)
@ -340,7 +340,7 @@ terminal:
- **Ephemeral mode** (`container_persistent: false`): Uses tmpfs for workspace — everything is lost on cleanup
:::tip
For production gateway deployments, use `docker`, `modal`, `daytona`, or `vercel_sandbox` backend to isolate agent commands from your host system. This eliminates the need for dangerous command approval entirely.
For production gateway deployments, use `docker`, `modal`, or `daytona` backend to isolate agent commands from your host system. This eliminates the need for dangerous command approval entirely.
:::
:::warning
@ -357,7 +357,6 @@ If you add names to `terminal.docker_forward_env`, those variables are intention
| **singularity** | Container | ❌ Skipped | HPC environments |
| **modal** | Cloud sandbox | ❌ Skipped | Scalable cloud isolation |
| **daytona** | Cloud sandbox | ❌ Skipped | Persistent cloud workspaces |
| **vercel_sandbox** | Cloud microVM | ❌ Skipped | Cloud execution with snapshot persistence |
## Environment Variable Passthrough {#environment-variable-passthrough}

3
website/docs/user-guide/skills/bundled/autonomous-ai-agents/autonomous-ai-agents-hermes-agent.md
View file

@ -406,7 +406,6 @@ Full config reference: https://hermes-agent.nousresearch.com/docs/user-guide/con
| Alibaba / DashScope | API key | `DASHSCOPE_API_KEY` |
| Xiaomi MiMo | API key | `XIAOMI_API_KEY` |
| Kilo Code | API key | `KILOCODE_API_KEY` |
| AI Gateway (Vercel) | API key | `AI_GATEWAY_API_KEY` |
| OpenCode Zen | API key | `OPENCODE_ZEN_API_KEY` |
| OpenCode Go | API key | `OPENCODE_GO_API_KEY` |
| Qwen OAuth | OAuth | `hermes auth add qwen-oauth` |
@ -1014,7 +1013,7 @@ See `tests/agent/test_prompt_builder.py::TestEnvironmentHints` for a worked exam
Factual guidance about the host OS, user home, cwd, terminal backend, and shell (bash vs. PowerShell on Windows) is emitted from `agent/prompt_builder.py::build_environment_hints()`. This is also where the WSL hint and per-backend probe logic live. The convention:
- **Local terminal backend** → emit host info (OS, `$HOME`, cwd) + Windows-specific notes (hostname ≠ username, `terminal` uses bash not PowerShell).
- **Remote terminal backend** (anything in `_REMOTE_TERMINAL_BACKENDS`: `docker, singularity, modal, daytona, ssh, vercel_sandbox, managed_modal`) → **suppress** host info entirely and describe only the backend. A live `uname`/`whoami`/`pwd` probe runs inside the backend via `tools.environments.get_environment(...).execute(...)`, cached per process in `_BACKEND_PROBE_CACHE`, with a static fallback if the probe times out.
- **Remote terminal backend** (anything in `_REMOTE_TERMINAL_BACKENDS`: `docker, singularity, modal, daytona, ssh, managed_modal`) → **suppress** host info entirely and describe only the backend. A live `uname`/`whoami`/`pwd` probe runs inside the backend via `tools.environments.get_environment(...).execute(...)`, cached per process in `_BACKEND_PROBE_CACHE`, with a static fallback if the probe times out.
- **Key fact for prompt authoring:** when `TERMINAL_ENV != "local"`, *every* file tool (`read_file`, `write_file`, `patch`, `search_files`) runs inside the backend container, not on the host. The system prompt must never describe the host in that case — the agent can't touch it.
Full design notes, the exact emitted strings, and testing pitfalls:

2
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/developer-guide/architecture.md
View file

@ -211,7 +211,7 @@ CLI、gateway、cron、ACP 及辅助调用共用的运行时解析器。将 `(pr
### 工具系统
中央工具注册表（`tools/registry.py`），包含约 28 个 toolset 中的 70+ 个已注册工具。每个工具文件在导入时自行注册。注册表负责 schema 收集、分发、可用性检查和错误包装。终端工具支持 7 种后端（local、Docker、SSH、Daytona、Modal、Singularity、Vercel Sandbox）。
中央工具注册表（`tools/registry.py`），包含约 28 个 toolset 中的 70+ 个已注册工具。每个工具文件在导入时自行注册。注册表负责 schema 收集、分发、可用性检查和错误包装。终端工具支持 6 种后端（local、Docker、SSH、Daytona、Modal、Singularity）。
→ [工具运行时](./tools-runtime.md)

12
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/developer-guide/provider-runtime.md
View file

@ -42,7 +42,6 @@ Hermes 拥有一个共享的 provider 运行时解析器，用于以下场景：
当前 provider 系列包括（完整内置集合见 `plugins/model-providers/`）：
- AI Gateway（Vercel）
- OpenRouter
- Nous Portal
- OpenAI Codex
@ -93,18 +92,13 @@ Hermes 拥有一个共享的 provider 运行时解析器，用于以下场景：
- ACP 编辑器会话
- 辅助模型任务
## AI Gateway
## OpenRouter 与自定义 OpenAI 兼容 base URL
在 `~/.hermes/.env` 中设置 `AI_GATEWAY_API_KEY`，并使用 `--provider ai-gateway` 运行。Hermes 从 gateway 的 `/models` 端点获取可用模型，筛选出支持工具调用的语言模型。
## OpenRouter、AI Gateway 与自定义 OpenAI 兼容 base URL
Hermes 包含相关逻辑，以避免在存在多个 provider 密钥时（例如同时存在 `OPENROUTER_API_KEY`、`AI_GATEWAY_API_KEY` 和 `OPENAI_API_KEY`）将错误的 API key 泄露给自定义端点。
Hermes 包含相关逻辑，以避免在存在多个 provider 密钥时（例如同时存在 `OPENROUTER_API_KEY` 和 `OPENAI_API_KEY`）将错误的 API key 泄露给自定义端点。
每个 provider 的 API key 仅作用于其自身的 base URL：
- `OPENROUTER_API_KEY` 仅发送至 `openrouter.ai` 端点
- `AI_GATEWAY_API_KEY` 仅发送至 `ai-gateway.vercel.sh` 端点
- `OPENAI_API_KEY` 用于自定义端点及作为回退
Hermes 还区分以下两种情况：
@ -115,7 +109,7 @@ Hermes 还区分以下两种情况：
这种区分对以下场景尤为重要：
- 本地模型服务器
- 非 OpenRouter/非 AI Gateway 的 OpenAI 兼容 API
- 非 OpenRouter 的 OpenAI 兼容 API
- 无需重新运行 setup 即可切换 provider
- 通过 config 保存的自定义端点，即使当前 shell 中未导出 `OPENAI_BASE_URL` 也应正常工作

1
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/developer-guide/tools-runtime.md
View file

@ -213,7 +213,6 @@ registry.dispatch(name, args, **kwargs)
- singularity
- modal
- daytona
- vercel_sandbox
还支持：

1
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/getting-started/quickstart.md
View file

@ -124,7 +124,6 @@ hermes setup --portal
| **NVIDIA NIM** | 通过 build.nvidia.com 或本地 NIM 使用 Nemotron 模型 | 设置 `NVIDIA_API_KEY`（可选：`NVIDIA_BASE_URL`） |
| **GitHub Copilot** | GitHub Copilot 订阅（GPT-5.x、Claude、Gemini 等） | 通过 `hermes model` 进行 OAuth，或设置 `COPILOT_GITHUB_TOKEN` / `GH_TOKEN` |
| **GitHub Copilot ACP** | Copilot ACP agent 后端（在本地启动 `copilot` CLI） | `hermes model`（需要 `copilot` CLI + `copilot login`） |
| **Vercel AI Gateway** | Vercel AI Gateway 路由 | 设置 `AI_GATEWAY_API_KEY` |
| **Custom Endpoint** | VLLM、SGLang、Ollama 或任何兼容 OpenAI 的 API | 设置 base URL + API key |
对于大多数初次使用的用户：选择一个 provider，接受默认值（除非你明确知道为何要修改）。完整的 provider 目录及环境变量和配置步骤请参阅 [Providers](../integrations/providers.md) 页面。

3
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/integrations/providers.md
View file

@ -21,7 +21,6 @@ sidebar_position: 1
| **Anthropic** | `hermes model`（Claude Max + 额外用量积分，通过 OAuth；也支持 Anthropic API key 或手动 setup-token——见下方说明） |
| **OpenRouter** | `~/.hermes/.env` 中的 `OPENROUTER_API_KEY` |
| **NovitaAI** | `~/.hermes/.env` 中的 `NOVITA_API_KEY`（provider: `novita`，200+ 模型，Model API、Agent Sandbox、GPU Cloud） |
| **AI Gateway** | `~/.hermes/.env` 中的 `AI_GATEWAY_API_KEY`（provider: `ai-gateway`） |
| **z.ai / GLM** | `~/.hermes/.env` 中的 `GLM_API_KEY`（provider: `zai`） |
| **Kimi / Moonshot** | `~/.hermes/.env` 中的 `KIMI_API_KEY`（provider: `kimi-coding`） |
| **Kimi / Moonshot（中国）** | `~/.hermes/.env` 中的 `KIMI_CN_API_KEY`（provider: `kimi-coding-cn`；别名：`kimi-cn`、`moonshot-cn`） |
@ -1478,7 +1477,7 @@ fallback_model:
激活时，故障转移在不丢失对话的情况下中途切换模型和提供商。链按条目逐一尝试；每个会话激活一次。
支持的提供商：`openrouter`、`nous`、`openai-codex`、`copilot`、`copilot-acp`、`anthropic`、`gemini`、`google-gemini-cli`、`qwen-oauth`、`huggingface`、`zai`、`kimi-coding`、`kimi-coding-cn`、`minimax`、`minimax-cn`、`minimax-oauth`、`deepseek`、`nvidia`、`xai`、`xai-oauth`、`ollama-cloud`、`bedrock`、`ai-gateway`、`azure-foundry`、`opencode-zen`、`opencode-go`、`kilocode`、`xiaomi`、`arcee`、`gmi`、`stepfun`、`lmstudio`、`alibaba`、`alibaba-coding-plan`、`tencent-tokenhub`、`custom`。
支持的提供商：`openrouter`、`nous`、`openai-codex`、`copilot`、`copilot-acp`、`anthropic`、`gemini`、`google-gemini-cli`、`qwen-oauth`、`huggingface`、`zai`、`kimi-coding`、`kimi-coding-cn`、`minimax`、`minimax-cn`、`minimax-oauth`、`deepseek`、`nvidia`、`xai`、`xai-oauth`、`ollama-cloud`、`bedrock`、`azure-foundry`、`opencode-zen`、`opencode-go`、`kilocode`、`xiaomi`、`arcee`、`gmi`、`stepfun`、`lmstudio`、`alibaba`、`alibaba-coding-plan`、`tencent-tokenhub`、`custom`。
:::tip
故障转移仅通过 `config.yaml` 配置——或通过 `hermes fallback` 交互式配置。有关触发时机、链推进方式以及与辅助任务和委托的交互，参见[故障转移提供商](/user-guide/features/fallback-providers)。

2
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/reference/cli-commands.md
View file

@ -95,7 +95,7 @@ hermes chat [options]
| `-q`, `--query "..."` | 单次非交互式 prompt。 |
| `-m`, `--model <model>` | 覆盖本次运行的模型。 |
| `-t`, `--toolsets <csv>` | 启用逗号分隔的 toolset 集合。 |
| `--provider <provider>` | 强制指定 provider：`auto`、`openrouter`、`nous`、`openai-codex`、`copilot-acp`、`copilot`、`anthropic`、`gemini`、`google-gemini-cli`、`huggingface`、`novita`、`zai`、`kimi-coding`、`kimi-coding-cn`、`minimax`、`minimax-cn`、`minimax-oauth`、`kilocode`、`xiaomi`、`arcee`、`gmi`、`alibaba`、`alibaba-coding-plan`（别名 `alibaba_coding`）、`deepseek`、`nvidia`、`ollama-cloud`、`xai`（别名 `grok`）、`xai-oauth`（别名 `grok-oauth`）、`qwen-oauth`、`bedrock`、`opencode-zen`、`opencode-go`、`ai-gateway`、`azure-foundry`、`lmstudio`、`stepfun`、`tencent-tokenhub`（别名 `tencent`、`tokenhub`）。 |
| `--provider <provider>` | 强制指定 provider：`auto`、`openrouter`、`nous`、`openai-codex`、`copilot-acp`、`copilot`、`anthropic`、`gemini`、`google-gemini-cli`、`huggingface`、`novita`、`zai`、`kimi-coding`、`kimi-coding-cn`、`minimax`、`minimax-cn`、`minimax-oauth`、`kilocode`、`xiaomi`、`arcee`、`gmi`、`alibaba`、`alibaba-coding-plan`（别名 `alibaba_coding`）、`deepseek`、`nvidia`、`ollama-cloud`、`xai`（别名 `grok`）、`xai-oauth`（别名 `grok-oauth`）、`qwen-oauth`、`bedrock`、`opencode-zen`、`opencode-go`、`azure-foundry`、`lmstudio`、`stepfun`、`tencent-tokenhub`（别名 `tencent`、`tokenhub`）。 |
| `-s`, `--skills <name>` | 为会话预加载一个或多个 skill（可重复或逗号分隔）。 |
| `-v`, `--verbose` | 详细输出。 |
| `-Q`, `--quiet` | 程序化模式：抑制横幅/spinner/工具预览。 |

9
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/reference/environment-variables.md
View file

@ -18,8 +18,6 @@ description: "Hermes Agent 使用的所有环境变量完整参考"
| `HERMES_OPENROUTER_CACHE_TTL` | 缓存 TTL（秒，1-86400）。覆盖 config.yaml 中的 `openrouter.response_cache_ttl`。 |
| `NOUS_BASE_URL` | 覆盖 Nous Portal base URL（极少使用；仅用于开发/测试） |
| `NOUS_INFERENCE_BASE_URL` | 直接覆盖 Nous 推理端点 |
| `AI_GATEWAY_API_KEY` | Vercel AI Gateway API 密钥（[ai-gateway.vercel.sh](https://ai-gateway.vercel.sh)） |
| `AI_GATEWAY_BASE_URL` | 覆盖 AI Gateway base URL（默认：`https://ai-gateway.vercel.sh/v1`） |
| `OPENAI_API_KEY` | 自定义 OpenAI 兼容端点的 API 密钥（与 `OPENAI_BASE_URL` 配合使用） |
| `OPENAI_BASE_URL` | 自定义端点的 base URL（VLLM、SGLang 等） |
| `COPILOT_GITHUB_TOKEN` | 用于 Copilot API 的 GitHub token——最高优先级（OAuth `gho_*` 或细粒度 PAT `github_pat_*`；经典 PAT `ghp_*` **不支持**） |
@ -156,10 +154,6 @@ description: "Hermes Agent 使用的所有环境变量完整参考"
| `HINDSIGHT_TIMEOUT` | Hindsight 内存提供商 API 调用超时（秒，默认：`60`）。如果 Hindsight 实例在 `/sync` 或 `on_session_switch` 期间响应缓慢并出现超时，请增大此值，并检查 `errors.log`。 |
| `SUPERMEMORY_API_KEY` | 支持 profile 召回和会话摄取的语义长期记忆（[supermemory.ai](https://supermemory.ai)） |
| `DAYTONA_API_KEY` | Daytona 云沙箱（[daytona.io](https://daytona.io/)） |
| `VERCEL_TOKEN` | Vercel Sandbox 访问 token（[vercel.com](https://vercel.com/)） |
| `VERCEL_PROJECT_ID` | Vercel 项目 ID（与 `VERCEL_TOKEN` 配合使用） |
| `VERCEL_TEAM_ID` | Vercel 团队 ID（与 `VERCEL_TOKEN` 配合使用） |
| `VERCEL_OIDC_TOKEN` | Vercel 短期 OIDC token（仅用于开发的替代方案） |
### Langfuse 可观测性
@ -192,7 +186,7 @@ description: "Hermes Agent 使用的所有环境变量完整参考"
| 变量 | 描述 |
|----------|-------------|
| `TERMINAL_ENV` | 后端：`local`、`docker`、`ssh`、`singularity`、`modal`、`daytona`、`vercel_sandbox` |
| `TERMINAL_ENV` | 后端：`local`、`docker`、`ssh`、`singularity`、`modal`、`daytona` |
| `HERMES_DOCKER_BINARY` | 覆盖 Hermes 调用的容器二进制（例如 `podman`、`/usr/local/bin/docker`）。未设置时，Hermes 自动在 `PATH` 上发现 `docker` 或 `podman`。当两者都已安装且需要非默认选项，或二进制不在 `PATH` 中时使用。 |
| `TERMINAL_DOCKER_IMAGE` | Docker 镜像（默认：`nikolaik/python-nodejs:python3.11-nodejs20`） |
| `TERMINAL_DOCKER_FORWARD_ENV` | 显式转发到 Docker 终端会话的环境变量名 JSON 数组。注意：技能声明的 `required_environment_variables` 会自动转发——仅对未被任何技能声明的变量使用此项。 |
@ -201,7 +195,6 @@ description: "Hermes Agent 使用的所有环境变量完整参考"
| `TERMINAL_SINGULARITY_IMAGE` | Singularity 镜像或 `.sif` 路径 |
| `TERMINAL_MODAL_IMAGE` | Modal 容器镜像 |
| `TERMINAL_DAYTONA_IMAGE` | Daytona 沙箱镜像 |
| `TERMINAL_VERCEL_RUNTIME` | Vercel Sandbox 运行时（`node24`、`node22`、`python3.13`） |
| `TERMINAL_TIMEOUT` | 命令超时（秒） |
| `TERMINAL_LIFETIME_SECONDS` | 终端会话最大生命周期（秒） |
| `TERMINAL_CWD` | 终端会话的工作目录（仅 gateway/cron；CLI 使用启动目录） |

52
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/user-guide/configuration.md
View file

@ -83,11 +83,11 @@ delegation:
## 终端后端配置
Hermes 支持七种终端后端。每种后端决定 agent 的 shell 命令实际在哪里执行 —— 本地机器、Docker 容器、通过 SSH 的远程服务器、Modal 云沙箱（直接或通过 Nous 托管的 gateway）、Daytona 工作区、Vercel Sandbox，或 Singularity/Apptainer 容器。
Hermes 支持六种终端后端。每种后端决定 agent 的 shell 命令实际在哪里执行 —— 本地机器、Docker 容器、通过 SSH 的远程服务器、Modal 云沙箱（直接或通过 Nous 托管的 gateway）、Daytona 工作区，或 Singularity/Apptainer 容器。
```yaml
terminal:
backend: local # local | docker | ssh | modal | daytona | vercel_sandbox | singularity
backend: local # local | docker | ssh | modal | daytona | singularity
cwd: "." # Gateway/cron 工作目录（CLI 始终使用启动目录）
timeout: 180 # 每条命令的超时时间（秒）
env_passthrough: [] # 转发到沙箱执行的环境变量名（terminal + execute_code）
@ -96,7 +96,7 @@ terminal:
daytona_image: "nikolaik/python-nodejs:python3.11-nodejs20" # Daytona 后端的容器镜像
```
对于 Modal、Daytona 和 Vercel Sandbox 等云沙箱，`container_persistent: true` 表示 Hermes 将尝试在沙箱重建后保留文件系统状态。这并不保证相同的活跃沙箱、PID 空间或后台进程之后仍在运行。
对于 Modal 和 Daytona 等云沙箱，`container_persistent: true` 表示 Hermes 将尝试在沙箱重建后保留文件系统状态。这并不保证相同的活跃沙箱、PID 空间或后台进程之后仍在运行。
### 后端概览
@ -107,7 +107,6 @@ terminal:
| **ssh** | 通过 SSH 的远程服务器 | 网络边界 | 远程开发、强大硬件 |
| **modal** | Modal 云沙箱 | 完全（云 VM） | 临时云计算、评估 |
| **daytona** | Daytona 工作区 | 完全（云容器） | 托管云开发环境 |
| **vercel_sandbox** | Vercel Sandbox | 完全（云 microVM） | 带快照文件系统持久化的云执行 |
| **singularity** | Singularity/Apptainer 容器 | 命名空间（--containall） | HPC 集群、共享机器 |
### Local 后端
@ -232,49 +231,6 @@ terminal:
**磁盘限制：** Daytona 强制执行 10 GiB 最大值。超过此值的请求将被截断并发出警告。
### Vercel Sandbox 后端
在 [Vercel Sandbox](https://vercel.com/docs/vercel-sandbox) 云 microVM 中运行命令。Hermes 使用普通的终端和文件工具接口；没有 Vercel 特定的面向模型的工具。
```yaml
terminal:
backend: vercel_sandbox
vercel_runtime: node24 # node24 | node22 | python3.13
cwd: /vercel/sandbox # 默认工作区根目录
container_persistent: true # 快照/恢复文件系统
container_disk: 51200 # 仅共享默认值；不支持自定义磁盘
```
**必需安装：** 安装可选 SDK 扩展：
```bash
pip install 'hermes-agent[vercel]'
```
**必需认证：** 使用 `VERCEL_TOKEN`、`VERCEL_PROJECT_ID` 和 `VERCEL_TEAM_ID` 三者全部配置访问令牌认证。这是在 Render、Railway、Docker 及类似宿主上部署和正常长期运行 Hermes 进程的受支持设置。
对于一次性本地开发，Hermes 也接受短期 Vercel OIDC token：
```bash
VERCEL_OIDC_TOKEN="$(vc project token <project-name>)" hermes chat
```
在已链接的 Vercel 项目目录中，可以省略项目名称：
```bash
VERCEL_OIDC_TOKEN="$(vc project token)" hermes chat
```
OIDC token 是短期的，不应作为文档化的部署路径使用。
**运行时：** `terminal.vercel_runtime` 支持 `node24`、`node22` 和 `python3.13`。未设置时，Hermes 默认使用 `node24`。
**持久化：** 当 `container_persistent: true` 时，Hermes 在清理期间对沙箱文件系统进行快照，并从该快照为同一任务恢复后续沙箱。快照内容可以包括复制到沙箱中的 Hermes 同步凭据、技能和缓存文件。这仅保留文件系统状态；不保留活跃沙箱身份、PID 空间、shell 状态或正在运行的后台进程。
**后台命令：** `terminal(background=true)` 使用 Hermes 的通用非本地后台进程流程。您可以在沙箱存活期间通过普通进程工具生成、轮询、等待、查看日志和终止进程。Hermes 不提供清理或重启后的原生 Vercel 分离进程恢复。
**磁盘大小：** Vercel Sandbox 目前不支持 Hermes 的 `container_disk` 资源旋钮。将 `container_disk` 保持未设置或使用共享默认值 `51200`；非默认值会导致诊断和后端创建失败，而不是被静默忽略。
### Singularity/Apptainer 后端
在 [Singularity/Apptainer](https://apptainer.org) 容器中运行命令。专为 Docker 不可用的 HPC 集群和共享机器设计。
@ -818,7 +774,7 @@ Hermes 中的每个模型槽位 —— 辅助任务、压缩、回退 —— 使
当设置 `base_url` 时，Hermes 忽略 provider 并直接调用该端点（使用 `api_key` 或 `OPENAI_API_KEY` 进行认证）。当仅设置 `provider` 时，Hermes 使用该 provider 的内置认证和基础 URL。
辅助任务的可用 providers：`auto`、`main`，以及[provider 注册表](/reference/environment-variables)中的任何 provider —— `openrouter`、`nous`、`openai-codex`、`copilot`、`copilot-acp`、`anthropic`、`gemini`、`google-gemini-cli`、`qwen-oauth`、`zai`、`kimi-coding`、`kimi-coding-cn`、`minimax`、`minimax-cn`、`minimax-oauth`、`deepseek`、`nvidia`、`xai`、`xai-oauth`、`ollama-cloud`、`alibaba`、`bedrock`、`huggingface`、`arcee`、`xiaomi`、`kilocode`、`opencode-zen`、`opencode-go`、`ai-gateway`、`azure-foundry` —— 或您 `custom_providers` 列表中任何命名的自定义 provider（例如 `provider: "beans"`）。
辅助任务的可用 providers：`auto`、`main`，以及[provider 注册表](/reference/environment-variables)中的任何 provider —— `openrouter`、`nous`、`openai-codex`、`copilot`、`copilot-acp`、`anthropic`、`gemini`、`google-gemini-cli`、`qwen-oauth`、`zai`、`kimi-coding`、`kimi-coding-cn`、`minimax`、`minimax-cn`、`minimax-oauth`、`deepseek`、`nvidia`、`xai`、`xai-oauth`、`ollama-cloud`、`alibaba`、`bedrock`、`huggingface`、`arcee`、`xiaomi`、`kilocode`、`opencode-zen`、`opencode-go`、`azure-foundry` —— 或您 `custom_providers` 列表中任何命名的自定义 provider（例如 `provider: "beans"`）。
:::tip MiniMax OAuth
`minimax-oauth` 通过浏览器 OAuth 登录（无需 API 密钥）。运行 `hermes model` 并选择 **MiniMax (OAuth)** 进行认证。辅助任务自动使用 `MiniMax-M2.7-highspeed`。参阅 [MiniMax OAuth 指南](../guides/minimax-oauth.md)。

1
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/user-guide/features/fallback-providers.md
View file

@ -47,7 +47,6 @@ fallback_model:
| 提供商 | 值 | 要求 |
|----------|-------|-------------|
| AI Gateway | `ai-gateway` | `AI_GATEWAY_API_KEY` |
| OpenRouter | `openrouter` | `OPENROUTER_API_KEY` |
| Nous Portal | `nous` | `hermes setup --portal`（全新安装）或 `hermes auth add nous`（OAuth） |
| OpenAI Codex | `openai-codex` | `hermes model`（ChatGPT OAuth） |

33
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/user-guide/features/tools.md
View file

@ -65,14 +65,13 @@ hermes tools
| `singularity` | HPC 容器 | 集群计算、无 root 权限 |
| `modal` | 云端执行 | 无服务器、弹性扩展 |
| `daytona` | 云端沙箱工作区 | 持久化远程开发环境 |
| `vercel_sandbox` | Vercel Sandbox 云微虚拟机 | 带快照文件系统持久化的云端执行 |
### 配置
```yaml
# 在 ~/.hermes/config.yaml 中
terminal:
backend: local # 或：docker, ssh, singularity, modal, daytona, vercel_sandbox
backend: local # 或：docker, ssh, singularity, modal, daytona
cwd: "." # 工作目录
timeout: 180 # 命令超时时间（秒）
```
@ -123,41 +122,13 @@ modal setup
hermes config set terminal.backend modal
```
### Vercel Sandbox
```bash
pip install 'hermes-agent[vercel]'
hermes config set terminal.backend vercel_sandbox
hermes config set terminal.vercel_runtime node24
```
需同时配置 `VERCEL_TOKEN`、`VERCEL_PROJECT_ID` 和 `VERCEL_TEAM_ID` 三个凭据。此访问令牌配置方式是在 Render、Railway、Docker 及类似平台上进行部署和正常长期运行 Hermes 进程的推荐路径。支持的运行时为 `node24`、`node22` 和 `python3.13`；Hermes 默认使用 `/vercel/sandbox` 作为远程工作区根目录。
对于本地一次性开发，Hermes 也接受短期 Vercel OIDC token：
```bash
VERCEL_OIDC_TOKEN="$(vc project token <project-name>)" hermes chat
```
在已关联的 Vercel 项目目录中：
```bash
VERCEL_OIDC_TOKEN="$(vc project token)" hermes chat
```
启用 `container_persistent: true` 后，Hermes 使用 Vercel 快照在同一任务的沙箱重建时保留文件系统状态，其中可包含沙箱内 Hermes 同步的凭据、技能和缓存文件。快照不保留活跃进程、PID 空间或相同的活跃沙箱标识。
后台终端命令使用 Hermes 通用的非本地进程流程：在沙箱存活期间，spawn、poll、wait、log 和 kill 均通过标准 process 工具运行，但 Hermes 不提供清理或重启后的原生 Vercel 后台进程恢复能力。
`container_disk` 保持未设置或使用共享默认值 `51200`；Vercel Sandbox 不支持自定义磁盘大小，设置后将导致诊断/后端创建失败。
### 容器资源
为所有容器后端配置 CPU、内存、磁盘和持久化：
```yaml
terminal:
backend: docker # 或 singularity, modal, daytona, vercel_sandbox
backend: docker # 或 singularity, modal, daytona
container_cpu: 1 # CPU 核心数（默认：1）
container_memory: 5120 # 内存（MB，默认：5GB）
container_disk: 51200 # 磁盘（MB，默认：50GB）

5
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/user-guide/security.md
View file

@ -144,7 +144,7 @@ approvals:
| `gateway run` 配合 `&`/`disown`/`nohup`/`setsid` | 防止在服务管理器外启动 gateway |
:::info
**容器绕过**：在 `docker`、`singularity`、`modal`、`daytona` 或 `vercel_sandbox` 后端运行时，危险命令检查会被**跳过**，因为容器本身就是安全边界。容器内的破坏性命令不会危害宿主机。
**容器绕过**：在 `docker`、`singularity`、`modal` 或 `daytona` 后端运行时，危险命令检查会被**跳过**，因为容器本身就是安全边界。容器内的破坏性命令不会危害宿主机。
:::
### 审批流程（CLI）
@ -340,7 +340,7 @@ terminal:
- **临时模式**（`container_persistent: false`）：工作区使用 tmpfs——清理后所有内容丢失
:::tip
对于生产 gateway 部署，使用 `docker`、`modal`、`daytona` 或 `vercel_sandbox` 后端，将 Agent 命令与宿主机系统隔离。这样可以完全消除危险命令审批的需要。
对于生产 gateway 部署，使用 `docker`、`modal` 或 `daytona` 后端，将 Agent 命令与宿主机系统隔离。这样可以完全消除危险命令审批的需要。
:::
:::warning
@ -357,7 +357,6 @@ terminal:
| **singularity** | 容器 | ❌ 跳过 | HPC 环境 |
| **modal** | 云沙箱 | ❌ 跳过 | 可扩展的云隔离 |
| **daytona** | 云沙箱 | ❌ 跳过 | 持久化云工作区 |
| **vercel_sandbox** | 云微虚拟机 | ❌ 跳过 | 带快照持久化的云执行 |
## 环境变量透传 {#environment-variable-passthrough}

3
website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/user-guide/skills/bundled/autonomous-ai-agents/autonomous-ai-agents-hermes-agent.md
View file

@ -401,7 +401,6 @@ Profiles 使用 `~/.hermes/profiles/<name>/`，布局相同。
| Alibaba / DashScope | API key | `DASHSCOPE_API_KEY` |
| Xiaomi MiMo | API key | `XIAOMI_API_KEY` |
| Kilo Code | API key | `KILOCODE_API_KEY` |
| AI Gateway (Vercel) | API key | `AI_GATEWAY_API_KEY` |
| OpenCode Zen | API key | `OPENCODE_ZEN_API_KEY` |
| OpenCode Go | API key | `OPENCODE_GO_API_KEY` |
| Qwen OAuth | OAuth | `hermes auth add qwen-oauth` |
@ -922,7 +921,7 @@ monkeypatch.setattr(platform, "release", lambda: "6.8.0-generic")
关于宿主 OS、用户 home、cwd、终端后端和 shell（Windows 上的 bash vs PowerShell）的事实性指导从 `agent/prompt_builder.py::build_environment_hints()` 输出。WSL 提示和每个后端的探测逻辑也在此处。约定：
- **本地终端后端** → 输出宿主信息（OS、`$HOME`、cwd）+ Windows 特有说明（hostname ≠ username，`terminal` 使用 bash 而非 PowerShell）。
- **远程终端后端**（`_REMOTE_TERMINAL_BACKENDS` 中的任何内容：`docker, singularity, modal, daytona, ssh, vercel_sandbox, managed_modal`）→ **完全抑制**宿主信息，仅描述后端。通过 `tools.environments.get_environment(...).execute(...)` 在后端内运行实时 `uname`/`whoami`/`pwd` 探测，每进程缓存在 `_BACKEND_PROBE_CACHE` 中，探测超时时使用静态回退。
- **远程终端后端**（`_REMOTE_TERMINAL_BACKENDS` 中的任何内容：`docker, singularity, modal, daytona, ssh, managed_modal`）→ **完全抑制**宿主信息，仅描述后端。通过 `tools.environments.get_environment(...).execute(...)` 在后端内运行实时 `uname`/`whoami`/`pwd` 探测，每进程缓存在 `_BACKEND_PROBE_CACHE` 中，探测超时时使用静态回退。
- **prompt 编写的关键事实：** 当 `TERMINAL_ENV != "local"` 时，*每个*文件工具（`read_file`、`write_file`、`patch`、`search_files`）都在后端容器内运行，而非宿主上。在这种情况下，系统 prompt 绝不能描述宿主——agent 无法访问它。
完整设计说明、确切输出字符串和测试陷阱：`references/prompt-builder-environment-hints.md`。