feat: raise default tool-calling iteration limit from 90 to 500

The default max_iterations/agent.max_turns budget was set when long
agentic runs were rare; complex tasks now routinely exceed 90 tool
calls. Raise the default to 500 across every surface that hardcodes
the fallback: AIAgent constructor, DEFAULT_CONFIG, CLI resolution
chain, gateway env bridge, cron scheduler, and TUI gateway. Explicit
user config values are unaffected (deep-merge preserves them; no
_config_version bump needed).

Docs (en + zh-Hans), CLI help text, tips, and pinned tests updated
to match.

AGENTS.md

@@ -325,7 +325,7 @@ class AIAgent:
         provider: str = None,
         api_mode: str = None,              # "chat_completions" | "codex_responses" | ...
         model: str = "",                   # empty → resolved from config/provider later
-        max_iterations: int = 90,          # tool-calling iterations (shared with subagents)
+        max_iterations: int = 500,         # tool-calling iterations (shared with subagents)
         enabled_toolsets: list = None,
         disabled_toolsets: list = None,
         quiet_mode: bool = False,

agent/agent_init.py

@@ -455,7 +455,7 @@ def init_agent(
     command: str = None,
     args: list[str] | None = None,
     model: str = "",
-    max_iterations: int = 90,  # Default tool-calling iterations (shared with subagents)
+    max_iterations: int = 500,  # Default tool-calling iterations (shared with subagents)
     tool_delay: float = 1.0,
     enabled_toolsets: List[str] = None,
     disabled_toolsets: List[str] = None,
@@ -529,7 +529,7 @@ def init_agent(
         requested_provider (str): Original provider identity before runtime canonicalization
         api_mode (str): API mode override: "chat_completions" or "codex_responses"
         model (str): Model name to use (default: "anthropic/claude-opus-4.6")
-        max_iterations (int): Maximum number of tool calling iterations (default: 90)
+        max_iterations (int): Maximum number of tool calling iterations (default: 500)
         tool_delay (float): Delay between tool calls in seconds (default: 1.0)
         enabled_toolsets (List[str]): Only enable tools from these toolsets (optional)
         disabled_toolsets (List[str]): Disable tools from these toolsets (optional)

agent/iteration_budget.py

@@ -2,7 +2,7 @@
 
 Extracted from ``run_agent.py``.  Each ``AIAgent`` instance (parent or
 subagent) holds an :class:`IterationBudget`; the parent's cap comes from
-``max_iterations`` (default 90), each subagent's cap comes from
+``max_iterations`` (default 500), each subagent's cap comes from
 ``delegation.max_iterations`` (default 50).
 
 ``run_agent`` re-exports ``IterationBudget`` so existing
@@ -18,7 +18,7 @@ class IterationBudget:
     """Thread-safe iteration counter for an agent.
 
     Each agent (parent or subagent) gets its own ``IterationBudget``.
-    The parent's budget is capped at ``max_iterations`` (default 90).
+    The parent's budget is capped at ``max_iterations`` (default 500).
     Each subagent gets an independent budget capped at
     ``delegation.max_iterations`` (default 50) — this means total
     iterations across parent + subagents can exceed the parent's cap.

cli.py

@@ -471,7 +471,7 @@ def load_cli_config() -> Dict[str, Any]:
             "min_tail_user_messages": 1,  # Real user messages guaranteed in the tail (1 = existing single anchor)
         },
         "agent": {
-            "max_turns": 90,  # Default max tool-calling iterations (shared with subagents)
+            "max_turns": 500,  # Default max tool-calling iterations (shared with subagents)
             "verbose": False,
             "system_prompt": "",
             "prefill_messages_file": "",
@@ -4099,7 +4099,7 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin, CLIBillingMixin):
             provider: Inference provider ("auto", "openrouter", "nous", "openai-codex", "zai", "kimi-coding", "minimax", "minimax-cn")
             api_key: API key (default: from environment)
             base_url: API base URL (default: OpenRouter)
-            max_turns: Maximum tool-calling iterations shared with subagents (default: 90)
+            max_turns: Maximum tool-calling iterations shared with subagents (default: 500)
             verbose: Enable verbose logging
             compact: Use compact display mode
             resume: Session ID to resume (restores conversation history from SQLite)
@@ -4272,9 +4272,9 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin, CLIBillingMixin):
             try:
                 self.max_turns = int(os.getenv("HERMES_MAX_ITERATIONS", ""))
             except (TypeError, ValueError):
-                self.max_turns = 90
+                self.max_turns = 500
         else:
-            self.max_turns = 90
+            self.max_turns = 500
         
         # Parse and validate toolsets
         self.enabled_toolsets = toolsets
@@ -13128,8 +13128,8 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin, CLIBillingMixin):
             # Notify when iteration budget was hit
             if result and not result.get("completed") and not result.get("interrupted"):
                 _api_calls = result.get("api_calls", 0)
-                if _api_calls >= getattr(self.agent, "max_iterations", 90):
-                    _max_iter = getattr(self.agent, "max_iterations", 90)
+                if _api_calls >= getattr(self.agent, "max_iterations", 500):
+                    _max_iter = getattr(self.agent, "max_iterations", 500)
                     _cprint(
                         f"\n{_DIM}⚠ Iteration budget reached "
                         f"({_api_calls}/{_max_iter}) — "

cron/scheduler.py

@@ -3254,7 +3254,7 @@ def run_job(
                     prefill_messages = None
 
         # Max iterations
-        max_iterations = _cfg.get("agent", {}).get("max_turns") or _cfg.get("max_turns") or 90
+        max_iterations = _cfg.get("agent", {}).get("max_turns") or _cfg.get("max_turns") or 500
 
         # Provider routing
         pr = _cfg.get("provider_routing") or {}

gateway/run.py

@@ -1667,9 +1667,9 @@ def _current_max_iterations() -> int:
     """Return the current per-turn iteration budget after runtime env refresh."""
     _reload_runtime_env_preserving_config_authority()
     try:
-        return int(os.getenv("HERMES_MAX_ITERATIONS", "90"))
+        return int(os.getenv("HERMES_MAX_ITERATIONS", "500"))
     except (TypeError, ValueError):
-        return 90
+        return 500
 
 
 from contextlib import contextmanager as _contextmanager
@@ -7877,10 +7877,10 @@ class GatewayRunner(GatewayAuthorizationMixin, GatewayKanbanWatchersMixin, Gatew
         # config.yaml → env bridge did the right thing at a glance (instead
         # of silently running at a stale .env value for weeks).
         try:
-            _effective_max_iter = int(os.getenv("HERMES_MAX_ITERATIONS", "90"))
+            _effective_max_iter = int(os.getenv("HERMES_MAX_ITERATIONS", "500"))
             logger.info(
                 "Agent budget: max_iterations=%d (agent.max_turns from config.yaml, "
-                "or HERMES_MAX_ITERATIONS from .env, or default 90)",
+                "or HERMES_MAX_ITERATIONS from .env, or default 500)",
                 _effective_max_iter,
             )
         except Exception:

hermes_cli/_parser.py

@@ -382,7 +382,7 @@ def build_top_level_parser():
         type=int,
         default=None,
         metavar="N",
-        help="Maximum tool-calling iterations per conversation turn (default: 90, or agent.max_turns in config)",
+        help="Maximum tool-calling iterations per conversation turn (default: 500, or agent.max_turns in config)",
     )
     _inherited_flag(
         chat_parser,

hermes_cli/config.py

@@ -945,7 +945,7 @@ DEFAULT_CONFIG = {
     # pressure. Reopening one re-resumes it from disk. 0/null disables.
     "max_live_sessions": 16,
     "agent": {
-        "max_turns": 90,
+        "max_turns": 500,
         # Inactivity timeout for gateway agent execution (seconds).
         # The agent can run indefinitely as long as it's actively calling
         # tools or receiving API responses.  Only fires when the agent has

hermes_cli/tips.py

@@ -69,7 +69,7 @@ TIPS = [
     "hermes chat -t web,terminal enables only specific toolsets for a focused session.",
     "hermes chat -s github-pr-workflow preloads a skill at launch.",
     "hermes chat -q \"query\" runs a single non-interactive query and exits.",
-    "hermes chat --max-turns 200 overrides the default 90-iteration limit per turn.",
+    "hermes chat --max-turns 1000 overrides the default 500-iteration limit per turn.",
     "hermes chat --checkpoints enables filesystem snapshots before every destructive file change.",
     "hermes --yolo bypasses all dangerous command approval prompts for the entire session.",
     "hermes chat --source telegram tags the session for filtering in hermes sessions list.",
@@ -112,7 +112,7 @@ TIPS = [
     "Set display.busy_input_mode: queue to queue messages instead of interrupting the agent, or steer to inject them mid-run via /steer.",
     "Set display.resume_display: minimal to skip the full conversation recap on session resume.",
     "Set compression.threshold: 0.50 to control when auto-compression fires (default: 50% of context).",
-    "Set agent.max_turns: 200 to let the agent take more tool-calling steps per turn.",
+    "Set agent.max_turns: 1000 to let the agent take more tool-calling steps per turn.",
     "Set file_read_max_chars: 200000 to increase the max content per read_file call.",
     "Set approvals.mode: smart to let an LLM auto-approve safe commands and auto-deny dangerous ones.",
     "Set fallback_model in config.yaml to automatically fail over to a backup provider.",

run_agent.py

@@ -437,7 +437,7 @@ class AIAgent:
         command: str = None,
         args: list[str] | None = None,
         model: str = "",
-        max_iterations: int = 90,  # Default tool-calling iterations (shared with subagents)
+        max_iterations: int = 500,  # Default tool-calling iterations (shared with subagents)
         tool_delay: float = 1.0,
         enabled_toolsets: List[str] = None,
         disabled_toolsets: List[str] = None,

tests/cli/test_cli_init.py

@@ -60,7 +60,7 @@ class TestMaxTurnsResolution:
     def test_default_max_turns_is_integer(self):
         cli = _make_cli()
         assert isinstance(cli.max_turns, int)
-        assert cli.max_turns == 90
+        assert cli.max_turns == 500
 
     def test_explicit_max_turns_honored(self):
         cli = _make_cli(max_turns=25)
@@ -69,7 +69,7 @@ class TestMaxTurnsResolution:
     def test_none_max_turns_gets_default(self):
         cli = _make_cli(max_turns=None)
         assert isinstance(cli.max_turns, int)
-        assert cli.max_turns == 90
+        assert cli.max_turns == 500
 
     def test_env_var_max_turns(self):
         """Env var is used when config file doesn't set max_turns."""
@@ -79,7 +79,7 @@ class TestMaxTurnsResolution:
     def test_invalid_env_var_max_turns_falls_back_to_default(self):
         """Invalid env values should not crash CLI init."""
         cli_obj = _make_cli(env_overrides={"HERMES_MAX_ITERATIONS": "not-a-number"})
-        assert cli_obj.max_turns == 90
+        assert cli_obj.max_turns == 500
 
     def test_legacy_root_max_turns_is_used_when_agent_key_exists_without_value(self):
         cli_obj = _make_cli(config_overrides={"agent": {}, "max_turns": 77})
@@ -88,7 +88,7 @@ class TestMaxTurnsResolution:
     def test_max_turns_never_none_for_agent(self):
         """The value passed to AIAgent must never be None (causes TypeError in run_conversation)."""
         cli = _make_cli()
-        assert isinstance(cli.max_turns, int) and cli.max_turns == 90
+        assert isinstance(cli.max_turns, int) and cli.max_turns == 500
 
 
 class TestVerboseAndToolProgress:

tests/test_tui_gateway_server.py

@@ -11656,22 +11656,22 @@ def test_make_agent_waits_for_shared_mcp_discovery(monkeypatch):
 
 def test_make_agent_nested_max_turns_takes_priority(monkeypatch):
     _setup_make_agent_mocks(
-        monkeypatch, {"agent": {"max_turns": 500}, "max_turns": 100}
+        monkeypatch, {"agent": {"max_turns": 400}, "max_turns": 100}
     )
 
     with patch("run_agent.AIAgent") as mock_agent:
         server._make_agent("sid1", "key1")
 
-    assert mock_agent.call_args.kwargs["max_iterations"] == 500
+    assert mock_agent.call_args.kwargs["max_iterations"] == 400
 
 
-def test_make_agent_defaults_to_90(monkeypatch):
+def test_make_agent_defaults_to_500(monkeypatch):
     _setup_make_agent_mocks(monkeypatch, {})
 
     with patch("run_agent.AIAgent") as mock_agent:
         server._make_agent("sid1", "key1")
 
-    assert mock_agent.call_args.kwargs["max_iterations"] == 90
+    assert mock_agent.call_args.kwargs["max_iterations"] == 500
 
 
 def test_make_agent_uses_session_runtime_overrides(monkeypatch):

tui_gateway/server.py

@@ -5751,7 +5751,7 @@ def _make_agent(
     _pr = _load_provider_routing()
     return AIAgent(
         model=model,
-        max_iterations=_cfg_max_turns(cfg, 90),
+        max_iterations=_cfg_max_turns(cfg, 500),
         provider=runtime.get("provider"),
         base_url=runtime.get("base_url"),
         api_key=runtime.get("api_key"),
@@ -17560,7 +17560,7 @@ def _(rid, params: dict) -> dict:
             {
                 "title": "Agent",
                 "rows": [
-                    ["Max Turns", str(_cfg_max_turns(cfg, 90))],
+                    ["Max Turns", str(_cfg_max_turns(cfg, 500))],
                     ["Toolsets", ", ".join(cfg.get("enabled_toolsets", [])) or "all"],
                     ["Verbose", str(cfg.get("verbose", False))],
                 ],

website/docs/developer-guide/agent-loop.md

@@ -181,7 +181,7 @@ These tools modify agent state directly and return synthetic tool results withou
 
 The agent tracks iterations via `IterationBudget`:
 
-- Default: 90 iterations (configurable via `agent.max_turns`)
+- Default: 500 iterations (configurable via `agent.max_turns`)
 - Each agent gets its own budget. Subagents get independent budgets capped at `delegation.max_iterations` (default 50) — total iterations across parent + subagents can exceed the parent's cap
 - At 100%, the agent stops and returns a summary of work done
 

website/docs/guides/python-library.md

@@ -309,7 +309,7 @@ print(review)
 | `disabled_toolsets` | `List[str]` | `None` | Blacklist specific toolsets |
 | `save_trajectories` | `bool` | `False` | Save conversations to JSONL |
 | `ephemeral_system_prompt` | `str` | `None` | Custom system prompt (not saved to trajectories) |
-| `max_iterations` | `int` | `90` | Max tool-calling iterations per conversation |
+| `max_iterations` | `int` | `500` | Max tool-calling iterations per conversation |
 | `skip_context_files` | `bool` | `False` | Skip loading AGENTS.md files |
 | `skip_memory` | `bool` | `False` | Disable persistent memory read/write |
 | `api_key` | `str` | `None` | API key (falls back to env vars) |
@@ -329,5 +329,5 @@ print(review)
 :::warning
 - **Thread safety**: Create one `AIAgent` per thread or task. Never share an instance across concurrent calls.
 - **Resource cleanup**: The agent automatically cleans up resources (terminal sessions, browser instances) when a conversation ends. If you're running in a long-lived process, ensure each conversation completes normally.
-- **Iteration limits**: The default `max_iterations=90` is generous. For simple Q&A use cases, consider lowering it (e.g., `max_iterations=10`) to prevent runaway tool-calling loops and control costs.
+- **Iteration limits**: The default `max_iterations=500` is generous. For simple Q&A use cases, consider lowering it (e.g., `max_iterations=10`) to prevent runaway tool-calling loops and control costs.
 :::

website/docs/reference/cli-commands.md

@@ -120,7 +120,7 @@ Common options:
 | `--ignore-rules` | Skip auto-injection of `AGENTS.md`, `SOUL.md`, `.cursorrules`, persistent memory, and preloaded skills. Combine with `--ignore-user-config` for a fully isolated run. |
 | `--safe-mode` | Troubleshooting mode: disable ALL customizations — user config, rules/memory injection, plugins, shell hooks, and MCP servers (implies `--ignore-user-config` and `--ignore-rules`). Use to isolate whether a problem comes from your setup or from Hermes itself. |
 | `--source <tag>` | Session source tag for filtering (default: `cli`). Use `tool` for third-party integrations that should not appear in user session lists. |
-| `--max-turns <N>` | Maximum tool-calling iterations per conversation turn (default: 90, or `agent.max_turns` in config). |
+| `--max-turns <N>` | Maximum tool-calling iterations per conversation turn (default: 500, or `agent.max_turns` in config). |
 
 Examples:
 

website/docs/reference/environment-variables.md

@@ -733,7 +733,7 @@ Advanced per-platform knobs for throttling the outbound message batcher. Most us
 
 | Variable | Description |
 |----------|-------------|
-| `HERMES_MAX_ITERATIONS` | Max tool-calling iterations per conversation (default: 90) |
+| `HERMES_MAX_ITERATIONS` | Max tool-calling iterations per conversation (default: 500) |
 | `HERMES_INFERENCE_MODEL` | Override model name at process level (takes priority over `config.yaml` for the session). Also settable via `-m`/`--model` flag. |
 | `HERMES_YOLO_MODE` | Set to `1` to bypass dangerous-command approval prompts. Equivalent to `--yolo`. |
 | `HERMES_ACCEPT_HOOKS` | Auto-approve any unseen shell hooks declared in `config.yaml` without a TTY prompt. Equivalent to `--accept-hooks` or `hooks_auto_accept: true`. |

website/docs/user-guide/configuration.md

@@ -854,13 +854,13 @@ See [Memory Providers](/user-guide/features/memory-providers) for the analogous
 
 ## Iteration Budget
 
-When the agent is working on a complex task with many tool calls, it can burn through its iteration budget (default: 90 turns). Hermes does **not** inject mid-task pressure warnings — earlier builds warned the model at 70%/90% budget, which caused models to abandon complex tasks prematurely and was removed in April 2026.
+When the agent is working on a complex task with many tool calls, it can burn through its iteration budget (default: 500 turns). Hermes does **not** inject mid-task pressure warnings — earlier builds warned the model at 70%/90% budget, which caused models to abandon complex tasks prematurely and was removed in April 2026.
 
-Instead, when the budget is actually exhausted (90/90), Hermes injects one message asking the model to wrap up and allows a single **grace call** so it can deliver a final response. If that grace call still doesn't produce text, the agent is asked to summarise what it accomplished.
+Instead, when the budget is actually exhausted (500/500), Hermes injects one message asking the model to wrap up and allows a single **grace call** so it can deliver a final response. If that grace call still doesn't produce text, the agent is asked to summarise what it accomplished.
 
 ```yaml
 agent:
-  max_turns: 90                # Max iterations per conversation turn (default: 90)
+  max_turns: 500               # Max iterations per conversation turn (default: 500)
   api_max_retries: 3           # Retries per provider before fallback engages (default: 3)
 ```
 

website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/developer-guide/agent-loop.md

@@ -181,7 +181,7 @@ for each tool_call in response.tool_calls:
 
 agent 通过 `IterationBudget` 追踪迭代次数：
 
-- 默认：90 次迭代（可通过 `agent.max_turns` 配置）
+- 默认：500 次迭代（可通过 `agent.max_turns` 配置）
 - 每个 agent 拥有独立预算。子 agent 获得独立预算，上限为 `delegation.max_iterations`（默认 50）——父 agent 与子 agent 的总迭代次数可超过父 agent 的上限
 - 达到 100% 时，agent 停止并返回已完成工作的摘要
 

website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/guides/python-library.md

@@ -309,7 +309,7 @@ print(review)
 | `disabled_toolsets` | `List[str]` | `None` | 黑名单指定工具集 |
 | `save_trajectories` | `bool` | `False` | 将对话保存为 JSONL |
 | `ephemeral_system_prompt` | `str` | `None` | 自定义系统 prompt（不保存到轨迹文件） |
-| `max_iterations` | `int` | `90` | 每次对话的最大工具调用迭代次数 |
+| `max_iterations` | `int` | `500` | 每次对话的最大工具调用迭代次数 |
 | `skip_context_files` | `bool` | `False` | 跳过加载 AGENTS.md 文件 |
 | `skip_memory` | `bool` | `False` | 禁用持久化内存的读写 |
 | `api_key` | `str` | `None` | API 密钥（回退到环境变量） |
@@ -329,5 +329,5 @@ print(review)
 :::warning
 - **线程安全**：每个线程或任务创建一个 `AIAgent` 实例。切勿在并发调用中共享同一实例。
 - **资源清理**：agent 在对话结束时会自动清理资源（终端会话、浏览器实例）。若在长期运行的进程中使用，请确保每次对话正常结束。
-- **迭代限制**：默认的 `max_iterations=90` 较为宽松。对于简单的问答场景，建议适当降低该值（如 `max_iterations=10`），以防止工具调用循环失控并控制成本。
+- **迭代限制**：默认的 `max_iterations=500` 较为宽松。对于简单的问答场景，建议适当降低该值（如 `max_iterations=10`），以防止工具调用循环失控并控制成本。
 :::

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

@@ -108,7 +108,7 @@ hermes chat [options]
 | `--ignore-user-config` | 忽略 `~/.hermes/config.yaml`，使用内置默认值。`.env` 中的凭据仍会加载。适用于隔离的 CI 运行、可复现的 bug 报告和第三方集成。 |
 | `--ignore-rules` | 跳过 `AGENTS.md`、`SOUL.md`、`.cursorrules`、持久 memory 和预加载 skill 的自动注入。与 `--ignore-user-config` 组合可实现完全隔离的运行。 |
 | `--source <tag>` | 用于过滤的会话来源标签（默认：`cli`）。对于不应出现在用户会话列表中的第三方集成，使用 `tool`。 |
-| `--max-turns <N>` | 每个对话轮次的最大工具调用迭代次数（默认：90，或 config 中的 `agent.max_turns`）。 |
+| `--max-turns <N>` | 每个对话轮次的最大工具调用迭代次数（默认：500，或 config 中的 `agent.max_turns`）。 |
 
 示例：
 

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

@@ -526,7 +526,7 @@ Graph 事件（Teams 会议、日历、聊天等）的入站变更通知监听
 
 | 变量 | 描述 |
 |----------|-------------|
-| `HERMES_MAX_ITERATIONS` | 每次对话的最大工具调用迭代次数（默认：90） |
+| `HERMES_MAX_ITERATIONS` | 每次对话的最大工具调用迭代次数（默认：500） |
 | `HERMES_INFERENCE_MODEL` | 在进程级别覆盖模型名称（优先于本次会话的 `config.yaml`）。也可通过 `-m`/`--model` 标志设置。 |
 | `HERMES_YOLO_MODE` | 设为 `1` 可绕过危险命令审批提示。等同于 `--yolo`。 |
 | `HERMES_ACCEPT_HOOKS` | 无需 TTY 提示自动批准 `config.yaml` 中声明的任何未见过的 shell hook。等同于 `--accept-hooks` 或 `hooks_auto_accept: true`。 |

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

@@ -637,17 +637,17 @@ context:
 
 ## 迭代预算
 
-当 agent 在处理具有许多工具调用的复杂任务时，它可能会耗尽其迭代预算（默认：90 轮）。Hermes **不会**在任务中途注入压力警告 —— 早期版本会在预算达到 70%/90% 时警告模型，这会导致模型过早放弃复杂任务，该机制已于 2026 年 4 月移除。
+当 agent 在处理具有许多工具调用的复杂任务时，它可能会耗尽其迭代预算（默认：500 轮）。Hermes **不会**在任务中途注入压力警告 —— 早期版本会在预算达到 70%/90% 时警告模型，这会导致模型过早放弃复杂任务，该机制已于 2026 年 4 月移除。
 
-取而代之的是，当预算真正耗尽（90/90）时，Hermes 注入一条消息要求模型收尾，并允许一次**宽限调用**以便其给出最终响应。如果该宽限调用仍未产生文本，则会要求 agent 总结已完成的工作。
+取而代之的是，当预算真正耗尽（500/500）时，Hermes 注入一条消息要求模型收尾，并允许一次**宽限调用**以便其给出最终响应。如果该宽限调用仍未产生文本，则会要求 agent 总结已完成的工作。
 
 ```yaml
 agent:
-  max_turns: 90                # 每次对话轮次的最大迭代次数（默认：90）
+  max_turns: 500               # 每次对话轮次的最大迭代次数（默认：500）
   api_max_retries: 3           # 回退启动前每个 provider 的重试次数（默认：3）
 ```
 
-当迭代预算完全耗尽时，CLI 向用户显示通知：`⚠ Iteration budget reached (90/90) — response may be incomplete`。
+当迭代预算完全耗尽时，CLI 向用户显示通知：`⚠ Iteration budget reached (500/500) — response may be incomplete`。
 
 `agent.api_max_retries` 控制 Hermes 在回退 provider 切换启动**之前**对瞬时错误（速率限制、连接断开、5xx）重试 provider API 调用的次数。默认为 `3` —— 总共四次尝试。如果您配置了[回退 providers](/user-guide/features/fallback-providers) 并希望更快地故障转移，请将其降至 `0`，这样主 provider 上的第一个瞬时错误会立即切换到回退，而不是对不稳定的端点进行重试。