feat(learning-harness): chat endpoint independent of embeddings, with the DeepSeek adapter

Task 1.2. Detectors, the classifier's model tier, the proposer and the judge
resolve one chat endpoint: an explicit override first, then the scripted client
flag, then an endpoint named in the config, then DeepSeek, then OpenRouter.
Prove-order retrieval keeps its own embedding endpoint untouched, because the
chat provider we default to offers none.

The adapter handles the provider's documented quirks: thinking is on by default
there and silently ignores temperature, so it is disabled explicitly; the
reasoning field is kept on the reply for the caller to replay across tool turns;
JSON mode has no schema and occasionally returns empty content, so one retry is
built in, never a loop; token log-probabilities are requested only when the
caller wants them, and their mean is what lets the classifier abstain.

Cost comes from the provider's own usage fields, cache hits priced separately,
and a model with no recorded price reports no cost rather than a made-up one.
The endpoint record deliberately holds no key, so it is safe to journal.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01T7XD7x5KonMdmGRkp63EQG

src/openultrasast/provider/openrouter.py

@@ -68,10 +68,29 @@ class OpenRouterChatClient:
         messages: Sequence[Mapping[str, object]],
         tools: Sequence[Mapping[str, object]] | None = None,
         timeout_seconds: int = 60,
+        extra_body: Mapping[str, object] | None = None,
     ) -> dict[str, object]:
+        """The assistant message of the reply, provider fields included (a reasoning field is kept for the caller to replay)."""
+        payload = self.complete_chat_raw(
+            model=model, messages=messages, tools=tools, timeout_seconds=timeout_seconds, extra_body=extra_body
+        )
+        return _extract_message(payload)
+
+    def complete_chat_raw(
+        self,
+        *,
+        model: str,
+        messages: Sequence[Mapping[str, object]],
+        tools: Sequence[Mapping[str, object]] | None = None,
+        timeout_seconds: int = 60,
+        extra_body: Mapping[str, object] | None = None,
+    ) -> dict[str, object]:
+        """The whole response payload, so a caller can read `usage` for cost accounting."""
         body: dict[str, object] = {"model": model, "messages": list(messages), "temperature": 0}
         if tools:
             body["tools"] = list(tools)
+        for key, value in (extra_body or {}).items():
+            body[key] = value
         payload = json.dumps(body).encode()
         request = urllib.request.Request(
             f"{self.base_url.rstrip('/')}/chat/completions",
@@ -88,10 +107,9 @@ class OpenRouterChatClient:
                 return cast(dict[str, object], json.loads(response.read().decode()))
 
         try:
-            response_payload = call_with_retry(_do, attempts=self.max_attempts, base_delay=self.retry_base_delay)
+            return call_with_retry(_do, attempts=self.max_attempts, base_delay=self.retry_base_delay)
         except (urllib.error.URLError, TimeoutError, json.JSONDecodeError) as exc:
             raise OpenRouterError(f"OpenRouter chat request failed: {exc}") from exc
-        return _extract_message(response_payload)
 
     def complete_json(self, *, model: str, messages: list[dict[str, str]], timeout_seconds: int = 60) -> object:
         message = self.complete_chat(model=model, messages=messages, timeout_seconds=timeout_seconds)