Make the first RAG slice executable without risking production data

The Stage 1 foundation now proves provider contracts with mocks and validates PostgreSQL/pgvector ingestion, approval binding, retrieval, reranking, and idempotency using only synthetic data. Live Bailian validation remains gated on rotating the exposed key.

Constraint: The key shown in chat is compromised and cannot be used or committed

Rejected: Mark Stage 1 complete from mock and offline results | real three-model smoke is still required

Confidence: high

Scope-risk: moderate

Reversibility: clean

Directive: Do not enable real-data ingestion until Stage 3 cloud approval and outbound manifest controls are enforced end to end

Tested: make verify; 41 pytest tests; strict mypy; Ruff; Compose config; pinned image build; empty-volume migration; role denial; two idempotent 20-vector seeds; database restart persistence

Not-tested: Live Bailian calls require a newly rotated key; React product UI is not implemented

backend/tests/contract/test_bailian_chat.py

@@ -0,0 +1,175 @@
+from __future__ import annotations
+
+import json
+
+import httpx
+import pytest
+
+from app.adapters.bailian.chat import BailianChatAdapter
+from app.ports.model_providers import ChatMessage, ModelProviderError
+
+BASE_URL = "https://workspace.cn-beijing.maas.aliyuncs.com/compatible-mode/v1"
+API_KEY = "sk-test-chat-secret"
+
+
+@pytest.mark.asyncio
+async def test_chat_completion_disables_thinking_and_search() -> None:
+    messages = [ChatMessage(role="user", content="Give a grounded answer")]
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        assert request.url.path == "/compatible-mode/v1/chat/completions"
+        assert request.headers["Authorization"] == f"Bearer {API_KEY}"
+        payload = json.loads(request.content)
+        assert payload["model"] == "deepseek-v4-flash"
+        assert payload["messages"] == [{"role": "user", "content": "Give a grounded answer"}]
+        assert payload["enable_thinking"] is False
+        assert payload["enable_search"] is False
+        assert payload["stream"] is False
+        return httpx.Response(
+            200,
+            json={
+                "id": "req_chat_1",
+                "model": "deepseek-v4-flash",
+                "choices": [
+                    {
+                        "message": {"role": "assistant", "content": "grounded"},
+                        "finish_reason": "stop",
+                    }
+                ],
+                "usage": {
+                    "prompt_tokens": 5,
+                    "completion_tokens": 1,
+                    "total_tokens": 6,
+                },
+            },
+        )
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as client:
+        adapter = BailianChatAdapter(
+            api_key=API_KEY,
+            base_url=BASE_URL,
+            http_client=client,
+        )
+        result = await adapter.complete(messages, max_tokens=32)
+
+    assert result.content == "grounded"
+    assert result.finish_reason == "stop"
+    assert result.usage.total_tokens == 6
+
+
+@pytest.mark.asyncio
+async def test_chat_stream_disables_thinking_and_search_and_parses_sse() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        payload = json.loads(request.content)
+        assert payload["enable_thinking"] is False
+        assert payload["enable_search"] is False
+        assert payload["stream"] is True
+        assert payload["stream_options"] == {"include_usage": True}
+        chunks = (
+            'data: {"id":"req_stream","model":"deepseek-v4-flash",'
+            '"choices":[{"delta":{"content":"矿"},"finish_reason":null}]}\n\n'
+            'data: {"id":"req_stream","model":"deepseek-v4-flash",'
+            '"choices":[{"delta":{"content":"体"},"finish_reason":"stop"}]}\n\n'
+            'data: {"id":"req_stream","model":"deepseek-v4-flash",'
+            '"choices":[],"usage":{"prompt_tokens":4,"completion_tokens":2,'
+            '"total_tokens":6}}\n\n'
+            "data: [DONE]\n\n"
+        )
+        return httpx.Response(
+            200,
+            headers={"content-type": "text/event-stream"},
+            content=chunks.encode(),
+        )
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as client:
+        adapter = BailianChatAdapter(
+            api_key=API_KEY,
+            base_url=BASE_URL,
+            http_client=client,
+        )
+        events = [
+            event
+            async for event in adapter.stream(
+                [ChatMessage(role="user", content="question")],
+                max_tokens=32,
+            )
+        ]
+
+    assert "".join(event.delta for event in events) == "矿体"
+    assert events[1].finish_reason == "stop"
+    assert events[2].usage.total_tokens == 6
+
+
+@pytest.mark.asyncio
+async def test_chat_error_does_not_expose_key_or_message_body() -> None:
+    sensitive_text = "CONFIDENTIAL_PROMPT_BODY"
+
+    def handler(_: httpx.Request) -> httpx.Response:
+        return httpx.Response(
+            401,
+            json={
+                "error": {
+                    "code": API_KEY,
+                    "message": f"{sensitive_text}: {API_KEY}",
+                }
+            },
+        )
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as client:
+        adapter = BailianChatAdapter(
+            api_key=API_KEY,
+            base_url=BASE_URL,
+            http_client=client,
+        )
+        with pytest.raises(ModelProviderError) as exc_info:
+            await adapter.complete(
+                [ChatMessage(role="user", content=sensitive_text)],
+                max_tokens=8,
+            )
+
+    rendered = f"{exc_info.value!s} {exc_info.value!r}"
+    assert API_KEY not in rendered
+    assert sensitive_text not in rendered
+
+
+@pytest.mark.asyncio
+async def test_chat_stream_honors_retry_after_before_any_token() -> None:
+    calls = 0
+
+    def handler(_: httpx.Request) -> httpx.Response:
+        nonlocal calls
+        calls += 1
+        if calls == 1:
+            return httpx.Response(
+                429,
+                headers={"Retry-After": "0"},
+                json={"error": {"code": "rate_limited"}},
+            )
+        return httpx.Response(
+            200,
+            headers={"content-type": "text/event-stream"},
+            content=(
+                'data: {"model":"deepseek-v4-flash",'
+                '"choices":[{"delta":{"content":"正常"},"finish_reason":"stop"}]}\n\n'
+                "data: [DONE]\n\n"
+            ).encode(),
+        )
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as client:
+        adapter = BailianChatAdapter(
+            api_key=API_KEY,
+            base_url=BASE_URL,
+            http_client=client,
+            max_retries=1,
+            retry_base_seconds=0,
+        )
+        events = [
+            event
+            async for event in adapter.stream(
+                [ChatMessage(role="user", content="question")],
+                max_tokens=8,
+            )
+        ]
+
+    assert "".join(event.delta for event in events) == "正常"
+    assert calls == 2