Skip to content

Use Claude Code, Codex, VSCode, Pi, and OpenCode (and 6 other harnesses) for free (1.3B+ free tokens) from your terminal, app, IDE, or phone, and now from the browser with native browser sessions (multi-harness + multi-model) like OpenClaw (voice supported + ToS friendly)

Python

Fix Auto Mode classifier responses when stream is omitted (#1098)

Ali Khokhar984c4c23b78d
diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md
index ddfb4272..153abfc1 100644
--- a/ARCHITECTURE.md
+++ b/ARCHITECTURE.md
@@ -396,7 +396,8 @@ response. Once FCC has finalized the failure, the response includes
without causing a second client retry loop. After the first chunk has escaped,
HTTP status is committed; Messages emits an Anthropic `event: error` and closes
without a synthetic `message_stop`; Responses emits `response.failed` with the
-original response ID. Non-streaming Messages aggregate internally and return
+original response ID. Messages are non-streaming unless the client explicitly
+sets `stream: true`. Non-streaming Messages aggregate internally and return
non-2xx JSON for any terminal stream error, discarding incomplete content rather
than presenting a partial success.
diff --git a/pyproject.toml b/pyproject.toml
index 9d19ca7f..30db8792 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
[project]
name = "free-claude-code"
-version = "4.3.0"
+version = "4.3.1"
description = "Local proxy connecting coding agents to OpenAI-compatible AI providers"
readme = "README.md"
requires-python = ">=3.14.0"
diff --git a/src/free_claude_code/api/handlers/messages.py b/src/free_claude_code/api/handlers/messages.py
index 36ea7e62..b5a86aeb 100644
--- a/src/free_claude_code/api/handlers/messages.py
+++ b/src/free_claude_code/api/handlers/messages.py
@@ -135,12 +135,12 @@ class MessagesHandler:
self,
result: _MessagesResult,
*,
- stream: bool | None,
+ stream: bool,
request_id: str,
) -> object:
if isinstance(result, _MessagesCompleteResult):
return result.response
- if stream is False:
+ if not stream:
# Non-streaming clients (e.g. Claude Code utility calls) need a
# complete JSON Message; the internal pipeline is always SSE, so
# serving that raw here breaks the client SDK's response parse.
diff --git a/src/free_claude_code/api/routes.py b/src/free_claude_code/api/routes.py
index 2b045b9f..d88fbaba 100644
--- a/src/free_claude_code/api/routes.py
+++ b/src/free_claude_code/api/routes.py
@@ -109,7 +109,7 @@ async def create_message(
services: ApiServices = Depends(get_services),
_auth=Depends(require_proxy_auth),
):
- """Create a message (streaming by default; stream=false gets aggregated JSON)."""
+ """Create a message (JSON by default; stream=true returns Anthropic SSE)."""
return await _create_messages_response(
services,
request_data,
diff --git a/src/free_claude_code/core/anthropic/models.py b/src/free_claude_code/core/anthropic/models.py
index 105c12a4..a0f639da 100644
--- a/src/free_claude_code/core/anthropic/models.py
+++ b/src/free_claude_code/core/anthropic/models.py
@@ -183,7 +183,7 @@ class MessagesRequest(BaseModel):
messages: list[Message]
system: str | list[SystemContent] | None = None
stop_sequences: list[str] | None = None
- stream: bool | None = True
+ stream: bool = False
temperature: float | None = None
top_p: float | None = None
top_k: int | None = None
diff --git a/src/free_claude_code/core/anthropic/sse_aggregation.py b/src/free_claude_code/core/anthropic/sse_aggregation.py
index 0827380c..9ddb336e 100644
--- a/src/free_claude_code/core/anthropic/sse_aggregation.py
+++ b/src/free_claude_code/core/anthropic/sse_aggregation.py
@@ -1,10 +1,10 @@
"""Fold an Anthropic Messages SSE stream into a single JSON Message body.
-Used to honor client requests with ``stream: false``. The internal pipeline
-is always SSE (provider execution always yields a stream), so
-callers that need a non-streaming response consume the stream here and get
-back the same shape the real Anthropic API returns for a non-streaming
-``messages.create()`` call.
+Used when client requests omit ``stream`` or set ``stream: false``. The internal
+pipeline is always SSE (provider execution always yields a stream), so callers
+that need a non-streaming response consume the stream here and get back the same
+shape the real Anthropic API returns for a non-streaming ``messages.create()``
+call.
"""
import json
diff --git a/tests/api/test_api.py b/tests/api/test_api.py
index 482f0db3..6902d84a 100644
--- a/tests/api/test_api.py
+++ b/tests/api/test_api.py
@@ -128,6 +128,36 @@ def test_create_message_stream(client: TestClient):
assert _stream_response_calls[0][1]["request_id"] == response.headers["request-id"]
+def test_auto_mode_classifier_without_stream_returns_json(client: TestClient):
+ """Claude side queries omit stream and require one complete Message object."""
+ _stream_response_calls.clear()
+ payload = {
+ "model": "claude-opus-4-8",
+ "max_tokens": 64,
+ "system": (
+ "You are a security monitor. Respond with <block>yes</block> "
+ "or <block>no</block>."
+ ),
+ "messages": [
+ {
+ "role": "user",
+ "content": "<transcript>\nBash curl example.com\n</transcript>",
+ }
+ ],
+ }
+
+ response = client.post("/v1/messages?beta=true", json=payload)
+
+ assert response.status_code == 200
+ assert response.headers["content-type"].startswith("application/json")
+ body = response.json()
+ assert body["type"] == "message"
+ assert body["usage"] == {"input_tokens": 0, "output_tokens": 0}
+ routed_request = _stream_response_calls[0][0][0]
+ assert routed_request.stream is False
+ assert _stream_response_calls[0][1]["thinking_enabled"] is False
+
+
def test_create_message_ingress_error_has_request_id_without_terminal_header(
client: TestClient,
):
diff --git a/tests/api/test_api_handlers.py b/tests/api/test_api_handlers.py
index 4b636bce..8db13369 100644
--- a/tests/api/test_api_handlers.py
+++ b/tests/api/test_api_handlers.py
@@ -103,6 +103,7 @@ async def test_messages_handler_passes_routed_request_and_stream_metadata() -> N
request = MessagesRequest(
model="nvidia_nim/test-model",
max_tokens=100,
+ stream=True,
messages=[Message(role="user", content="hi")],
)
@@ -373,6 +374,7 @@ async def test_messages_handler_forces_no_thinking_for_safety_classifier() -> No
request = MessagesRequest(
model="nvidia_nim/test-model",
max_tokens=100,
+ stream=True,
system=_CLASSIFIER_SYSTEM,
messages=[Message(role="user", content=_CLASSIFIER_USER)],
)
@@ -406,6 +408,7 @@ async def test_messages_handler_preserves_thinking_for_non_classifier() -> None:
request = MessagesRequest(
model="nvidia_nim/test-model",
max_tokens=100,
+ stream=True,
system="Explain XML formats.",
messages=[
Message(
@@ -441,6 +444,7 @@ async def test_messages_handler_keeps_existing_no_thinking_for_classifier() -> N
request = MessagesRequest(
model="claude-3-freecc-no-thinking/nvidia_nim/test-model",
max_tokens=100,
+ stream=True,
system=_CLASSIFIER_SYSTEM,
messages=[Message(role="user", content=_CLASSIFIER_USER)],
)
diff --git a/tests/api/test_web_server_tools.py b/tests/api/test_web_server_tools.py
index d6604400..8e93f84c 100644
--- a/tests/api/test_web_server_tools.py
+++ b/tests/api/test_web_server_tools.py
@@ -424,6 +424,7 @@ async def test_service_streams_forced_web_search_by_default(monkeypatch):
request = MessagesRequest(
model="claude-haiku-4-5-20251001",
max_tokens=100,
+ stream=True,
messages=[Message(role="user", content="Search for DeepSeek V4")],
tools=[Tool(name="web_search", type="web_search_20250305")],
tool_choice={"type": "tool", "name": "web_search"},
diff --git a/tests/core/anthropic/test_models.py b/tests/core/anthropic/test_models.py
index 10ff809c..01e26751 100644
--- a/tests/core/anthropic/test_models.py
+++ b/tests/core/anthropic/test_models.py
@@ -1,4 +1,5 @@
import pytest
+from pydantic import ValidationError
from free_claude_code.core.anthropic.conversion import (
OpenAIConversionError,
@@ -21,6 +22,19 @@ def test_messages_request_parses_without_model_mapping_side_effects():
)
assert request.model == "claude-3-opus"
+ assert request.stream is False
+
+
+def test_messages_request_rejects_null_stream() -> None:
+ with pytest.raises(ValidationError):
+ MessagesRequest.model_validate(
+ {
+ "model": "claude-3-opus",
+ "max_tokens": 100,
+ "messages": [{"role": "user", "content": "hello"}],
+ "stream": None,
+ }
+ )
def test_messages_request_normalizes_system_role_messages():
@@ -292,7 +306,7 @@ def test_messages_request_dump_preserves_public_defaults_and_excludes_internal_f
assert dumped == {
"model": "model",
"messages": [{"role": "user", "content": "hello"}],
- "stream": True,
+ "stream": False,
"thinking": {"enabled": True, "type": "adaptive"},
"client_extension": {"enabled": True},
}
diff --git a/uv.lock b/uv.lock
index 9d385036..15793dc5 100644
--- a/uv.lock
+++ b/uv.lock
@@ -561,7 +561,7 @@ wheels = [
[[package]]
name = "free-claude-code"
-version = "4.3.0"
+version = "4.3.1"
source = { editable = "." }
dependencies = [
{ name = "aiohttp" },