feat(xai-oauth): add xAI Grok OAuth (SuperGrok Subscription) provider

Adds a new authentication provider that lets SuperGrok subscribers sign
in to Hermes with their xAI account via the standard OAuth 2.0 PKCE
loopback flow, instead of pasting a raw API key from console.x.ai.

Highlights
----------
* OAuth 2.0 PKCE loopback login against accounts.x.ai with discovery,
  state/nonce, and a strict CORS-origin allowlist on the callback.
* Authorize URL carries `plan=generic` (required for non-allowlisted
  loopback clients) and `referrer=hermes-agent` for best-effort
  attribution in xAI's OAuth server logs.
* Token storage in `auth.json` with file-locked atomic writes; JWT
  `exp`-based expiry detection with skew; refresh-token rotation
  synced both ways between the singleton store and the credential
  pool so multi-process / multi-profile setups don't tear each other's
  refresh tokens.
* Reactive 401 retry: on a 401 from the xAI Responses API, the agent
  refreshes the token, swaps it back into `self.api_key`, and retries
  the call once. Guarded against silent account swaps when the active
  key was sourced from a different (manual) pool entry.
* Auxiliary tasks (curator, vision, embeddings, etc.) route through a
  dedicated xAI Responses-mode auxiliary client instead of falling back
  to OpenRouter billing.
* Direct HTTP tools (`tools/xai_http.py`, transcription, TTS, image-gen
  plugin) resolve credentials through a unified runtime → singleton →
  env-var fallback chain so xai-oauth users get them for free.
* `hermes auth add xai-oauth` and `hermes auth remove xai-oauth N` are
  wired through the standard auth-commands surface; remove cleans up
  the singleton loopback_pkce entry so it doesn't silently reinstate.
* `hermes model` provider picker shows
  "xAI Grok OAuth (SuperGrok Subscription)" and the model-flow falls
  back to pool credentials when the singleton is missing.

Hardening
---------
* Discovery and refresh responses validate the returned
  `token_endpoint` host against the same `*.x.ai` allowlist as the
  authorization endpoint, blocking MITM persistence of a hostile
  endpoint.
* Discovery / refresh / token-exchange `response.json()` calls are
  wrapped to raise typed `AuthError` on malformed bodies (captive
  portals, proxy error pages) instead of leaking JSONDecodeError
  tracebacks.
* `prompt_cache_key` is routed through `extra_body` on the codex
  transport (sending it as a top-level kwarg trips xAI's SDK with a
  TypeError).
* Credential-pool sync-back preserves `active_provider` so refreshing
  an OAuth entry doesn't silently flip the active provider out from
  under the running agent.

Testing
-------
* New `tests/hermes_cli/test_auth_xai_oauth_provider.py` (~63 tests)
  covers JWT expiry, OAuth URL params (plan + referrer), CORS origins,
  redirect URI validation, singleton↔pool sync, concurrency races,
  refresh error paths, runtime resolution, and malformed-JSON guards.
* Extended `test_credential_pool.py`, `test_codex_transport.py`, and
  `test_run_agent_codex_responses.py` cover the pool sync-back,
  `extra_body` routing, and 401 reactive refresh paths.
* 165 tests passing on this branch via `scripts/run_tests.sh`.
This commit is contained in:
Jaaneek 2026-05-15 16:10:38 +01:00 committed by Teknium
parent 9fb40e6a3d
commit b62c997973
27 changed files with 3843 additions and 131 deletions

View file

@ -1275,7 +1275,7 @@ class AIAgent:
self.api_mode = api_mode
elif self.provider == "openai-codex":
self.api_mode = "codex_responses"
elif self.provider == "xai":
elif self.provider in {"xai", "xai-oauth"}:
self.api_mode = "codex_responses"
elif (provider_name is None) and (
self._base_url_hostname == "chatgpt.com"
@ -7139,15 +7139,60 @@ class AIAgent:
raise RuntimeError("Responses create(stream=True) fallback did not emit a terminal response.")
def _try_refresh_codex_client_credentials(self, *, force: bool = True) -> bool:
if self.api_mode != "codex_responses" or self.provider != "openai-codex":
if self.api_mode != "codex_responses" or self.provider not in {"openai-codex", "xai-oauth"}:
return False
# Guard against silent account swap.
#
# When an agent is using a non-singleton credential — e.g. a manual
# pool entry (``hermes auth add xai-oauth``) whose tokens belong to
# a different account than the loopback_pkce singleton, or an agent
# constructed with an explicit ``api_key=`` arg — force-refreshing
# the singleton here and adopting its tokens silently re-routes the
# rest of the conversation onto the singleton's account. The
# credential pool's reactive recovery (``_recover_with_credential_pool``)
# is the right channel for that case; this path is the
# singleton-only fallback used when the pool can't recover, and
# MUST only fire when the agent really is on singleton tokens.
try:
if self.provider == "openai-codex":
from hermes_cli.auth import resolve_codex_runtime_credentials
singleton_now = resolve_codex_runtime_credentials(
refresh_if_expiring=False,
)
else:
from hermes_cli.auth import resolve_xai_oauth_runtime_credentials
singleton_now = resolve_xai_oauth_runtime_credentials(
refresh_if_expiring=False,
)
except Exception as exc:
logger.debug("%s singleton read failed: %s", self.provider, exc)
return False
singleton_key = str(singleton_now.get("api_key") or "").strip()
active_key = str(self.api_key or "").strip()
if singleton_key and active_key and singleton_key != active_key:
logger.debug(
"%s singleton tokens differ from the active api_key; "
"skipping singleton force-refresh to avoid silent account swap. "
"Reactive credential rotation should go through the pool.",
self.provider,
)
return False
try:
from hermes_cli.auth import resolve_codex_runtime_credentials
if self.provider == "openai-codex":
from hermes_cli.auth import resolve_codex_runtime_credentials
creds = resolve_codex_runtime_credentials(force_refresh=force)
creds = resolve_codex_runtime_credentials(force_refresh=force)
else:
from hermes_cli.auth import resolve_xai_oauth_runtime_credentials
creds = resolve_xai_oauth_runtime_credentials(force_refresh=force)
except Exception as exc:
logger.debug("Codex credential refresh failed: %s", exc)
logger.debug("%s credential refresh failed: %s", self.provider, exc)
return False
api_key = creds.get("api_key")
@ -7162,7 +7207,7 @@ class AIAgent:
self._client_kwargs["api_key"] = self.api_key
self._client_kwargs["base_url"] = self.base_url
if not self._replace_primary_openai_client(reason="codex_credential_refresh"):
if not self._replace_primary_openai_client(reason=f"{self.provider}_credential_refresh"):
return False
return True
@ -9631,7 +9676,7 @@ class AIAgent:
and "/backend-api/codex" in self._base_url_lower
)
)
is_xai_responses = self.provider == "xai" or self._base_url_hostname == "api.x.ai"
is_xai_responses = self.provider in {"xai", "xai-oauth"} or self._base_url_hostname == "api.x.ai"
_msgs_for_codex = self._prepare_messages_for_non_vision_model(api_messages)
return _ct.build_kwargs(
model=self.model,
@ -13700,13 +13745,14 @@ class AIAgent:
if (
self.api_mode == "codex_responses"
and self.provider == "openai-codex"
and self.provider in {"openai-codex", "xai-oauth"}
and status_code == 401
and not codex_auth_retry_attempted
):
codex_auth_retry_attempted = True
if self._try_refresh_codex_client_credentials(force=True):
self._vprint(f"{self.log_prefix}🔐 Codex auth refreshed after 401. Retrying request...")
_label = "xAI OAuth" if self.provider == "xai-oauth" else "Codex"
self._vprint(f"{self.log_prefix}🔐 {_label} auth refreshed after 401. Retrying request...")
continue
if (
self.api_mode == "chat_completions"
@ -14346,11 +14392,15 @@ class AIAgent:
self._vprint(f"{self.log_prefix} 🌐 Endpoint: {_base}", force=True)
# Actionable guidance for common auth errors
if classified.is_auth or classified.reason == FailoverReason.billing:
if _provider == "openai-codex" and status_code == 401:
self._vprint(f"{self.log_prefix} 💡 Codex OAuth token was rejected (HTTP 401). Your token may have been", force=True)
self._vprint(f"{self.log_prefix} refreshed by another client (Codex CLI, VS Code). To fix:", force=True)
self._vprint(f"{self.log_prefix} 1. Run `codex` in your terminal to generate fresh tokens.", force=True)
self._vprint(f"{self.log_prefix} 2. Then run `hermes auth` to re-authenticate.", force=True)
if _provider in {"openai-codex", "xai-oauth"} and status_code == 401:
if _provider == "openai-codex":
self._vprint(f"{self.log_prefix} 💡 Codex OAuth token was rejected (HTTP 401). Your token may have been", force=True)
self._vprint(f"{self.log_prefix} refreshed by another client (Codex CLI, VS Code). To fix:", force=True)
self._vprint(f"{self.log_prefix} 1. Run `codex` in your terminal to generate fresh tokens.", force=True)
self._vprint(f"{self.log_prefix} 2. Then run `hermes auth` to re-authenticate.", force=True)
else:
self._vprint(f"{self.log_prefix} 💡 xAI OAuth token was rejected (HTTP 401). To fix:", force=True)
self._vprint(f"{self.log_prefix} re-authenticate with xAI Grok OAuth (SuperGrok Subscription) from `hermes model`.", force=True)
else:
self._vprint(f"{self.log_prefix} 💡 Your API key was rejected by the provider. Check:", force=True)
self._vprint(f"{self.log_prefix} • Is the key valid? Run: hermes setup", force=True)