fix(providers): validate stream idle timeout config

2026-06-17 00:47:44 +08:00
parent 7bec0f6e01
commit 846410f936
6 changed files with 190 additions and 10 deletions
@@ -3,7 +3,6 @@
 from __future__ import annotations

 import asyncio
-import os
 import re
 import secrets
 import string
@@ -14,6 +13,7 @@ from nanobot.providers.base import (
    LLMProvider,
    LLMResponse,
    ToolCallRequest,
+    resolve_stream_idle_timeout_s,
    tool_arguments_object_for_replay,
 )

@@ -613,7 +613,7 @@ class AnthropicProvider(LLMProvider):
            messages, tools, model, max_tokens, temperature,
            reasoning_effort, tool_choice,
        )
-        idle_timeout_s = int(os.environ.get("NANOBOT_STREAM_IDLE_TIMEOUT_S", "90"))
+        idle_timeout_s = resolve_stream_idle_timeout_s()
        try:
            async with self._client.messages.stream(**kwargs) as stream:
                if on_content_delta or on_thinking_delta or on_tool_call_delta:
@@ -682,7 +682,7 @@ class AnthropicProvider(LLMProvider):
            return LLMResponse(
                content=(
                    f"Error calling LLM: stream stalled for more than "
-                    f"{idle_timeout_s} seconds"
+                    f"{idle_timeout_s:g} seconds"
                ),
                finish_reason="error",
                error_kind="timeout",
@@ -2,6 +2,7 @@

 import asyncio
 import json
+import os
 import re
 from abc import ABC, abstractmethod
 from collections.abc import Awaitable, Callable
@@ -16,6 +17,34 @@ from loguru import logger

 from nanobot.utils.helpers import image_placeholder_text

+STREAM_IDLE_TIMEOUT_ENV = "NANOBOT_STREAM_IDLE_TIMEOUT_S"
+DEFAULT_STREAM_IDLE_TIMEOUT_S = 90.0
+MAX_STREAM_IDLE_TIMEOUT_S = 3600.0
+
+
+def resolve_stream_idle_timeout_s(
+    *,
+    env_value: str | None = None,
+    default: float = DEFAULT_STREAM_IDLE_TIMEOUT_S,
+    maximum: float = MAX_STREAM_IDLE_TIMEOUT_S,
+) -> float:
+    """Return a safe streaming idle timeout from env/config text."""
+    raw = os.environ.get(STREAM_IDLE_TIMEOUT_ENV) if env_value is None else env_value
+    if raw is None or not raw.strip():
+        return default
+    try:
+        value = float(raw)
+    except (TypeError, ValueError):
+        logger.warning("Ignoring invalid {}={!r}; using {}", STREAM_IDLE_TIMEOUT_ENV, raw, default)
+        return default
+    if value <= 0:
+        logger.warning("Ignoring non-positive {}={!r}; using {}", STREAM_IDLE_TIMEOUT_ENV, raw, default)
+        return default
+    if value > maximum:
+        logger.warning("Clamping {}={!r} to {}", STREAM_IDLE_TIMEOUT_ENV, raw, maximum)
+        return maximum
+    return value
+

@dataclass
 class ToolCallRequest:
@@ -15,6 +15,7 @@ from nanobot.providers.base import (
    LLMResponse,
    ToolCallRequest,
    parse_tool_arguments,
+    resolve_stream_idle_timeout_s,
    tool_arguments_object_for_replay,
 )

@@ -701,7 +702,7 @@ class BedrockProvider(LLMProvider):
        on_tool_call_delta: Callable[[dict[str, Any]], Awaitable[None]] | None = None,
    ) -> LLMResponse:
        _ = on_thinking_delta, on_tool_call_delta
-        idle_timeout_s = int(os.environ.get("NANOBOT_STREAM_IDLE_TIMEOUT_S", "90"))
+        idle_timeout_s = resolve_stream_idle_timeout_s()
        content_parts: list[str] = []
        reasoning_parts: list[str] = []
        thinking_blocks: list[dict[str, Any]] = []
@@ -742,7 +743,7 @@ class BedrockProvider(LLMProvider):
            return LLMResponse(
                content=(
                    f"Error calling LLM: stream stalled for more than "
-                    f"{idle_timeout_s} seconds"
+                    f"{idle_timeout_s:g} seconds"
                ),
                finish_reason="error",
                error_kind="timeout",
@@ -5,7 +5,6 @@ from __future__ import annotations
 import asyncio
 import hashlib
 import json
-import os
 from collections.abc import Awaitable, Callable
 from typing import Any

@@ -13,7 +12,12 @@ import httpx
 from loguru import logger
 from oauth_cli_kit import get_token as get_codex_token

-from nanobot.providers.base import LLMProvider, LLMResponse, ToolCallRequest
+from nanobot.providers.base import (
+    LLMProvider,
+    LLMResponse,
+    ToolCallRequest,
+    resolve_stream_idle_timeout_s,
+)
 from nanobot.providers.openai_responses import (
    consume_sse_with_reasoning,
    convert_messages,
@@ -199,7 +203,7 @@ async def _request_codex(
    on_thinking_delta: Callable[[str], Awaitable[None]] | None = None,
    on_tool_call_delta: Callable[[dict[str, Any]], Awaitable[None]] | None = None,
 ) -> tuple[str, list[ToolCallRequest], str, dict[str, int], str | None]:
-    idle_timeout_s = int(os.environ.get("NANOBOT_STREAM_IDLE_TIMEOUT_S", "90"))
+    idle_timeout_s = resolve_stream_idle_timeout_s()
    async with httpx.AsyncClient(timeout=idle_timeout_s, verify=verify) as client:
        async with client.stream("POST", url, headers=headers, json=body) as response:
            if response.status_code != 200:
@@ -25,6 +25,7 @@ from nanobot.providers.base import (
    LLMResponse,
    ToolCallRequest,
    parse_tool_arguments,
+    resolve_stream_idle_timeout_s,
    tool_arguments_json_for_replay,
 )
 from nanobot.providers.openai_responses import (
@@ -1386,7 +1387,7 @@ class OpenAICompatProvider(LLMProvider):
        on_tool_call_delta: Callable[[dict[str, Any]], Awaitable[None]] | None = None,
    ) -> LLMResponse:
        await self._ensure_client()
-        idle_timeout_s = int(os.environ.get("NANOBOT_STREAM_IDLE_TIMEOUT_S", "90"))
+        idle_timeout_s = resolve_stream_idle_timeout_s()
        try:
            if self._should_use_responses_api(model, reasoning_effort):
                try:
@@ -1503,7 +1504,7 @@ class OpenAICompatProvider(LLMProvider):
            return LLMResponse(
                content=(
                    f"Error calling LLM: stream stalled for more than "
-                    f"{idle_timeout_s} seconds"
+                    f"{idle_timeout_s:g} seconds"
                ),
                finish_reason="error",
                error_kind="timeout",
@@ -0,0 +1,145 @@
+from __future__ import annotations
+
+from types import SimpleNamespace
+from typing import Any
+from unittest.mock import AsyncMock, MagicMock
+
+import httpx
+import pytest
+
+import nanobot.providers.openai_codex_provider as codex_provider
+from nanobot.providers.anthropic_provider import AnthropicProvider
+from nanobot.providers.base import (
+    DEFAULT_STREAM_IDLE_TIMEOUT_S,
+    MAX_STREAM_IDLE_TIMEOUT_S,
+    resolve_stream_idle_timeout_s,
+)
+from nanobot.providers.bedrock_provider import BedrockProvider
+from nanobot.providers.openai_compat_provider import OpenAICompatProvider
+
+
+class _AsyncStream:
+    def __init__(self, chunks: list[Any]) -> None:
+        self._chunks = chunks
+        self._idx = 0
+
+    def __aiter__(self) -> _AsyncStream:
+        return self
+
+    async def __anext__(self) -> Any:
+        if self._idx >= len(self._chunks):
+            raise StopAsyncIteration
+        chunk = self._chunks[self._idx]
+        self._idx += 1
+        return chunk
+
+
+class _AnthropicStream(_AsyncStream):
+    def __init__(self, chunks: list[Any]) -> None:
+        super().__init__(chunks)
+        self.get_final_message = AsyncMock(return_value=SimpleNamespace(
+            content=[SimpleNamespace(type="text", text="ok")],
+            stop_reason="end_turn",
+            usage=SimpleNamespace(input_tokens=1, output_tokens=1),
+        ))
+
+    async def __aenter__(self) -> _AnthropicStream:
+        return self
+
+    async def __aexit__(self, *_exc: object) -> None:
+        pass
+
+
+class _BedrockClient:
+    def converse_stream(self, **_kwargs: Any) -> dict[str, Any]:
+        return {"stream": iter([
+            {"contentBlockDelta": {"contentBlockIndex": 0, "delta": {"text": "ok"}}},
+            {"messageStop": {"stopReason": "end_turn"}},
+        ])}
+
+
+def test_stream_idle_timeout_parser_rejects_invalid_values() -> None:
+    assert resolve_stream_idle_timeout_s(env_value="abc") == DEFAULT_STREAM_IDLE_TIMEOUT_S
+    assert resolve_stream_idle_timeout_s(env_value="-1") == DEFAULT_STREAM_IDLE_TIMEOUT_S
+    assert resolve_stream_idle_timeout_s(env_value="0") == DEFAULT_STREAM_IDLE_TIMEOUT_S
+
+
+def test_stream_idle_timeout_parser_accepts_and_clamps_numeric_values() -> None:
+    assert resolve_stream_idle_timeout_s(env_value="1.5") == 1.5
+    assert resolve_stream_idle_timeout_s(env_value="7200") == MAX_STREAM_IDLE_TIMEOUT_S
+
+
+@pytest.mark.asyncio
+async def test_openai_compat_stream_ignores_invalid_idle_timeout_env(monkeypatch) -> None:
+    monkeypatch.setenv("NANOBOT_STREAM_IDLE_TIMEOUT_S", "abc")
+    provider = OpenAICompatProvider(api_key="sk-test", api_base="https://example.com/v1")
+
+    chunk = SimpleNamespace(
+        choices=[SimpleNamespace(
+            delta=SimpleNamespace(
+                content="ok",
+                reasoning_content=None,
+                reasoning=None,
+                tool_calls=None,
+                function_call=None,
+            ),
+            finish_reason="stop",
+        )],
+        usage=None,
+    )
+    provider._client = SimpleNamespace(
+        chat=SimpleNamespace(completions=SimpleNamespace(
+            create=AsyncMock(return_value=_AsyncStream([chunk])),
+        )),
+    )
+
+    result = await provider.chat_stream(messages=[{"role": "user", "content": "hi"}])
+
+    assert result.content == "ok"
+
+
+@pytest.mark.asyncio
+async def test_anthropic_stream_ignores_invalid_idle_timeout_env(monkeypatch) -> None:
+    monkeypatch.setenv("NANOBOT_STREAM_IDLE_TIMEOUT_S", "abc")
+    provider = AnthropicProvider(api_key="sk-test")
+    provider._client = MagicMock()
+    provider._client.messages.stream = MagicMock(return_value=_AnthropicStream([]))
+
+    result = await provider.chat_stream(messages=[{"role": "user", "content": "hi"}])
+
+    assert result.content == "ok"
+
+
+@pytest.mark.asyncio
+async def test_bedrock_stream_ignores_invalid_idle_timeout_env(monkeypatch) -> None:
+    monkeypatch.setenv("NANOBOT_STREAM_IDLE_TIMEOUT_S", "abc")
+    provider = BedrockProvider(region="us-east-1", client=_BedrockClient())
+
+    result = await provider.chat_stream(messages=[{"role": "user", "content": "hi"}])
+
+    assert result.content == "ok"
+
+
+@pytest.mark.asyncio
+async def test_codex_stream_ignores_invalid_idle_timeout_env(monkeypatch) -> None:
+    monkeypatch.setenv("NANOBOT_STREAM_IDLE_TIMEOUT_S", "abc")
+    original_client = httpx.AsyncClient
+    seen: dict[str, float] = {}
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        return httpx.Response(200, request=request)
+
+    def fake_client(*, timeout: float, verify: bool) -> httpx.AsyncClient:
+        seen["timeout"] = timeout
+        return original_client(transport=httpx.MockTransport(handler), timeout=timeout)
+
+    monkeypatch.setattr(codex_provider.httpx, "AsyncClient", fake_client)
+
+    await codex_provider._request_codex(
+        "https://codex.example/responses",
+        {},
+        {"input": []},
+        verify=True,
+    )
+
+    assert seen["timeout"] == DEFAULT_STREAM_IDLE_TIMEOUT_S