博客里记载的 9 个 Bug，修了几个？一次源码级的全面审计

# ── Detect thinking-budget exhaustion ──────────────
# When the model spends ALL output tokens on reasoning
# and has none left for the response, continuation
# retries are pointless.  Detect this early and give a
# targeted error instead of wasting 3 API calls.
_has_think_tags = bool(
    _trunc_content and re.search(
        r'<(?:think|thinking|reasoning|REASONING_SCRATCHPAD)[^>]*>',
        _trunc_content,
        re.IGNORECASE,
    )
)
_thinking_exhausted = (
    not _trunc_has_tool_calls
    and _has_think_tags
    and (
        (_trunc_content is not None and not agent._has_content_after_think_block(_trunc_content))
        or _trunc_content is None
    )
)

⚠️ **Thinking Budget Exhausted**

The model used all its output tokens on reasoning
and had none left for the actual response.

To fix this:
→ Lower reasoning effort: `/thinkon low` or `/thinkon minimal`
→ Or switch to a larger/non-reasoning model with `/model`

from hermes_cli.config import get_env_value
for env_var in pconfig.api_key_env_vars:
    val = (get_env_value(env_var) or "").strip()
    if has_usable_secret(val):
        return val, env_var

# Fallback: try credential pool
try:
    from agent.credential_pool import load_pool
    pool = load_pool(provider_id)
    if pool and pool.has_credentials():
        entry = pool.peek()
        if entry:
            key = getattr(entry, "access_token", "") or getattr(entry, "runtime_api_key", "")
            if has_usable_secret(str(key).strip()):
                return str(key).strip(), f"credential_pool:{provider_id}"
except Exception:
    pass

def get_env_value(name, default=None):
    try:
        from hermes_cli.config import get_env_value as _get_env_value
    except ImportError:
        return os.getenv(name, default)
    value = _get_env_value(name)
    return value if value is not None else (default if default is not None else "")

GROQ_BASE_URL = os.getenv("GROQ_BASE_URL", "https://api.groq.com/openai/v1")
OPENAI_BASE_URL = os.getenv("STT_OPENAI_BASE_URL", "https://api.openai.com/v1")
XAI_STT_BASE_URL = os.getenv("XAI_STT_BASE_URL", "https://api.x.ai/v1")
ELEVENLABS_STT_BASE_URL = os.getenv("ELEVENLABS_STT_BASE_URL", "https://api.elevenlabs.io/v1")

def _clarify_callback_sync(question: str, choices) -> str:
    from tools import clarify_gateway as _clarify_mod
    clarify_id = _uuid.uuid4().hex[:10]
    _clarify_mod.register(clarify_id=clarify_id, session_key=session_key or "",
                          question=question, choices=list(choices) if choices else None)
    # 暂停 typing indicator
    try:
        _status_adapter.pause_typing_for_chat(_status_chat_id)
    except Exception:
        pass
    # 发送到平台
    fut = safe_schedule_threadsafe(
        _status_adapter.send_clarify(chat_id=_status_chat_id, question=question, ...),
        _loop_for_step, ...)
    # 等待用户回复
    timeout = _clarify_mod.get_clarify_timeout()
    response = _clarify_mod.wait_for_response(clarify_id, timeout=float(timeout))
    if response is None or response == "":
        return f"[user did not respond within {int(timeout / 60)}m]"
    return response

agent.clarify_callback = _clarify_callback_sync

resolved_provider, resolved_model, resolved_base_url, resolved_api_key, resolved_api_mode = _resolve_task_provider_model(
    task, provider, model, base_url, api_key)
# ...
effective_provider, client, final_model = resolve_vision_provider_client(
    provider=resolved_provider if resolved_provider != "auto" else provider,
    model=resolved_model or model,
    base_url=resolved_base_url or base_url,
    api_key=resolved_api_key or api_key,  # ← 这个 or fallback 是关键
    async_mode=False,
)

# OAuth device code flow constants (same client ID as opencode/Copilot CLI)
COPILOT_OAUTH_CLIENT_ID = "Ov23li8tweQw6odWQebz"

def _list_titled_sessions() -> list[dict]:
    user_source = source.platform.value if source.platform else None
    sessions = self._session_db.list_sessions_rich(source=user_source, limit=10)
    return [s for s in sessions if s.get("title")][:10]

# agent/transports/chat_completions.py
if params.get("supports_reasoning", False):
    extra_body["reasoning"] = {"enabled": True, "effort": "medium"}