Spaces:
Running
Running
File size: 13,343 Bytes
a9e27c0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 | """Every ingress/egress pair survives a tool turn with protected reasoning."""
import asyncio
import json
from copy import deepcopy
from unittest.mock import patch
import httpx
import httpx2
import pytest
from fastapi.testclient import TestClient
from openai import AsyncOpenAI
from free_claude_code.config.provider_catalog import GEMINI_DEFAULT_BASE
from free_claude_code.config.settings import Settings
from free_claude_code.core.anthropic import aggregate_anthropic_sse_to_message
from free_claude_code.core.anthropic.stream_contracts import parse_sse_text
from free_claude_code.core.chat_completions.response import ChatResponseAccumulator
from free_claude_code.core.gemini.response import GeminiResponseAccumulator
from free_claude_code.providers.gemini import GeminiProvider
from free_claude_code.providers.open_router import OpenRouterProvider
from tests.api.support import create_test_app, provider_manager_for_app
from tests.providers.support import immediate_admission, make_provider_config
from tests.providers.test_custom_provider import definition, provider
from tests.providers.test_history_transports import _native
from tests.providers.test_native_tool_arguments import tool_events
SCHEMA = {"type": "object", "properties": {"path": {"type": "string"}}}
GEMINI_PARTS = [
{"text": "Find 17.", "thought": True},
{
"functionCall": {"id": "call_probe", "name": "read", "args": {"path": "kept"}},
"thoughtSignature": "native-google-signature",
},
]
@pytest.mark.parametrize("tool_type", [None, "custom", "web_search_20250305"])
def test_messages_custom_tool_type_is_translated_to_native_gemini(tool_type):
from tests.providers.test_gemini import _sse
bodies = []
def reply(request):
bodies.append(json.loads(request.content))
return _sse([{"text": "hello"}])
tool = {"name": "read", "description": "Read a file", "input_schema": SCHEMA}
if tool_type is not None:
tool["type"] = tool_type
with TestClient(matrix_app("gemini", reply)) as client:
response = client.post(
"/v1/messages",
json={
"model": "public",
"messages": [{"role": "user", "content": "hello"}],
"max_tokens": 128,
"tools": [tool],
},
)
if tool_type == "web_search_20250305":
assert response.status_code == 400
assert not bodies
else:
assert response.status_code == 200, response.text
assert bodies[0]["tools"] == [
{
"functionDeclarations": [
{
"name": "read",
"description": "Read a file",
"parametersJsonSchema": SCHEMA,
}
]
}
]
def source_events(protocol):
if protocol == "gemini":
return [
{
"responseId": "native-response",
"modelVersion": "upstream",
"candidates": [
{
"index": 0,
"content": {"role": "model", "parts": GEMINI_PARTS},
"finishReason": "STOP",
}
],
}
]
events = tool_events(protocol, '{"path":"kept"}')
native = _native(protocol)
if protocol == "chat":
return [
{
**events[0],
"choices": [{"index": 0, "delta": native, "finish_reason": None}],
},
*events,
]
if protocol == "messages":
tail = deepcopy(events[1:])
for event in tail:
if "index" in event:
event["index"] += 1
return [
events[0],
{"type": "content_block_start", "index": 0, "content_block": native},
{"type": "content_block_stop", "index": 0},
*tail,
]
for event in events:
if "output_index" in event:
event["output_index"] += 1
events[-1]["response"]["output"].insert(0, native)
events.insert(
1, {"type": "response.output_item.done", "output_index": 0, "item": native}
)
return [{**event, "sequence_number": index} for index, event in enumerate(events)]
def matrix_app(protocol, reply, *, max_attempts=2, model_fallbacks=None):
if protocol == "gemini":
instance = GeminiProvider(
make_provider_config(api_key="upstream-key", base_url=GEMINI_DEFAULT_BASE),
admission=immediate_admission(max_attempts=max_attempts),
transport=httpx.MockTransport(reply),
)
provider_id = "gemini"
settings = Settings(MODEL="gemini/models/backend")
elif protocol == "chat":
client = AsyncOpenAI(
api_key="upstream-key",
base_url="https://upstream.invalid/v1",
max_retries=0,
http_client=httpx2.AsyncClient(transport=httpx2.MockTransport(reply)),
)
with patch(
"free_claude_code.providers.openai_chat.client.AsyncOpenAI",
return_value=client,
):
instance = OpenRouterProvider(
make_provider_config(
api_key="upstream-key", base_url="https://upstream.invalid/v1"
),
admission=immediate_admission(max_attempts=max_attempts),
)
provider_id = "open_router"
settings = Settings(
MODEL="open_router/backend", MODEL_FALLBACKS=model_fallbacks
)
else:
configured = definition(
api_format="anthropic_messages"
if protocol == "messages"
else "openai_responses",
api_keys=["upstream-key"],
)
instance = provider(configured, openai_handler=reply, messages_handler=reply)
provider_id = configured.provider_id
settings = Settings(
MODEL=f"{provider_id}/backend", custom_providers=(configured,)
)
return create_test_app(settings, providers={provider_id: instance})
def ingress_request(wire, streaming, history=None):
tools = (
[{"name": "read", "input_schema": SCHEMA}]
if wire == "messages"
else [
{"type": "function", "function": {"name": "read", "parameters": SCHEMA}}
if wire == "chat"
else {"type": "function", "name": "read", "parameters": SCHEMA}
]
)
if wire == "gemini":
method = "streamGenerateContent?alt=sse" if streaming else "generateContent"
return f"/v1beta/models/public:{method}", {
"contents": history or [{"role": "user", "parts": [{"text": "hello"}]}],
"tools": [
{
"functionDeclarations": [
{"name": "read", "parametersJsonSchema": SCHEMA}
]
}
],
"generationConfig": {"thinkingConfig": {"includeThoughts": True}},
}
body = {
"model": "public",
"stream": streaming,
"tools": tools,
"input" if wire == "responses" else "messages": history
or [{"role": "user", "content": "hello"}],
}
if wire == "messages":
body["max_tokens"] = 100
return "/v1/chat/completions" if wire == "chat" else f"/v1/{wire}", body
def response_value(response, wire, streaming):
assert response.status_code == 200, response.text
if not streaming:
return response.json()
events = parse_sse_text(response.text)
if wire == "responses":
assert events[-1].event == "response.completed", response.text
return events[-1].data["response"]
if wire == "messages":
async def stream():
yield response.text
message, error, stopped = asyncio.run(
aggregate_anthropic_sse_to_message(stream())
)
assert error is None and stopped, response.text
return message
accumulator = (
ChatResponseAccumulator() if wire == "chat" else GeminiResponseAccumulator()
)
for event in events:
accumulator.feed(event.data)
if wire == "chat":
assert response.text.endswith("data: [DONE]\n\n"), response.text
assert isinstance(accumulator, ChatResponseAccumulator)
accumulator.finish()
return accumulator.result()
def reply_history(value, wire):
initial = {"role": "user", "content": "hello"}
if wire == "messages":
call = next(block for block in value["content"] if block["type"] == "tool_use")
assert call["input"] == {"path": "kept"}
return [
initial,
{"role": "assistant", "content": value["content"]},
{
"role": "user",
"content": [
{
"type": "tool_result",
"tool_use_id": call["id"],
"content": "tool-value",
}
],
},
]
if wire == "responses":
call = next(item for item in value["output"] if item["type"] == "function_call")
assert json.loads(call["arguments"]) == {"path": "kept"}
return [
initial,
*value["output"],
{
"type": "function_call_output",
"call_id": call["call_id"],
"output": "tool-value",
},
]
if wire == "chat":
message = value["choices"][0]["message"]
call = message["tool_calls"][0]
assert json.loads(call["function"]["arguments"]) == {"path": "kept"}
return [
initial,
message,
{"role": "tool", "tool_call_id": call["id"], "content": "tool-value"},
]
content = value["candidates"][0]["content"]
call = next(
part["functionCall"] for part in content["parts"] if "functionCall" in part
)
assert call["args"] == {"path": "kept"}
return [
{"role": "user", "parts": [{"text": "hello"}]},
content,
{
"role": "user",
"parts": [
{
"functionResponse": {
"id": call["id"],
"name": "read",
"response": {"output": "tool-value"},
}
}
],
},
]
@pytest.mark.parametrize("protocol", ["chat", "messages", "responses", "gemini"])
@pytest.mark.parametrize(
("wire", "streaming"),
[
(wire, stream)
for wire in ("messages", "responses", "chat", "gemini")
for stream in (False, True)
if wire != "responses" or stream
],
)
def test_all_sixteen_paths_replay_native_state_and_tool_results(
protocol, wire, streaming
):
bodies = []
def reply(request):
bodies.append(json.loads(request.content))
events = source_events(protocol)
raw = "".join(
(f"event: {event['type']}\n" if protocol == "messages" else "")
+ f"data: {json.dumps(event)}\n\n"
for event in events
)
response = (
httpx.Response if protocol in {"messages", "gemini"} else httpx2.Response
)
return response(200, headers={"content-type": "text/event-stream"}, text=raw)
app = matrix_app(protocol, reply)
with TestClient(app) as client:
path, body = ingress_request(wire, streaming)
first = response_value(client.post(path, json=body), wire, streaming)
assert "Find 17." in json.dumps(first)
path, body = ingress_request(wire, streaming, reply_history(first, wire))
response_value(client.post(path, json=body), wire, streaming)
assert provider_manager_for_app(app)._current.active_leases == 0
assert len(bodies) == 2
assert "tool-value" in json.dumps(bodies[1])
if protocol == "gemini":
assert bodies[1]["contents"][1] == {"role": "model", "parts": GEMINI_PARTS}
elif protocol == "messages":
replayed = next(
block
for block in bodies[1]["messages"][1]["content"]
if block.get("type") == "thinking"
)
expected = _native(protocol)
# The existing Messages -> Responses codec retains signed fields, but
# has never retained arbitrary block extensions. New paths retain both.
if wire == "responses":
expected.pop("extension")
assert replayed == expected
elif protocol == "responses":
assert _native(protocol) in bodies[1]["input"]
else:
assistant = next(
message
for message in bodies[1]["messages"]
if message["role"] == "assistant"
)
if wire in {"chat", "gemini"}:
assert (
assistant["reasoning_details"] == _native(protocol)["reasoning_details"]
)
else:
# Existing Messages/Responses presentations expose Chat's opaque
# bytes separately from its readable scalar representation.
assert any(
detail.get("data") == "opaque-original"
for detail in assistant["reasoning_details"]
)
assert json.dumps(assistant).count("Find 17.") == 1
|