Support codex responses API

src/aituner/http_client.py

@@ -147,17 +147,25 @@ def chat_completion(
     base_url: str,
     api_key_env: str | None,
     provider: str = "custom",
+    wire_api: str = "chat.completions",
     model: str,
     messages: list[dict[str, Any]],
     timeout_s: float,
     system_prompt: str = "",
 ) -> dict[str, Any]:
-    payload: dict[str, Any] = {"model": model, "messages": messages}
-    if system_prompt:
-        payload["messages"] = [{"role": "system", "content": system_prompt}, *messages]
+    if wire_api == "responses":
+        payload = {"model": model, "input": messages}
+        if system_prompt:
+            payload["instructions"] = system_prompt
+        path = "/v1/responses"
+    else:
+        payload = {"model": model, "messages": messages}
+        if system_prompt:
+            payload["messages"] = [{"role": "system", "content": system_prompt}, *messages]
+        path = "/v1/chat/completions"
     data = json.dumps(payload).encode("utf-8")
     request = urllib.request.Request(
-        url=_openai_url(base_url, "/v1/chat/completions"),
+        url=_openai_url(base_url, path),
         headers=_auth_headers(api_key_env, provider),
         data=data,
         method="POST",
@@ -167,7 +175,7 @@ def chat_completion(
             return json.loads(response.read().decode("utf-8"))
     except urllib.error.HTTPError as exc:
         detail = exc.read().decode("utf-8", errors="replace")
-        raise HttpClientError(f"chat_completion failed: {exc.code} {detail}") from exc
+        raise HttpClientError(f"llm_completion failed: {exc.code} {detail}") from exc
 
 
 @dataclass(frozen=True)

src/aituner/llm.py

@@ -183,6 +183,45 @@ def parse_proposal_text(text: str, study: StudySpec) -> Proposal:
     return validate_proposal(proposal, study)
 
 
+def _extract_response_text(response: dict[str, Any]) -> str:
+    output_text = response.get("output_text")
+    if isinstance(output_text, str) and output_text:
+        return output_text
+    choices = response.get("choices")
+    if isinstance(choices, list) and choices:
+        message = choices[0].get("message", {})
+        if isinstance(message, dict):
+            content = message.get("content")
+            if isinstance(content, str):
+                return content
+            if isinstance(content, list):
+                text = "".join(
+                    item.get("text", "")
+                    for item in content
+                    if isinstance(item, dict) and isinstance(item.get("text"), str)
+                )
+                if text:
+                    return text
+    output = response.get("output")
+    if isinstance(output, list):
+        parts: list[str] = []
+        for item in output:
+            if not isinstance(item, dict):
+                continue
+            content = item.get("content")
+            if not isinstance(content, list):
+                continue
+            for block in content:
+                if not isinstance(block, dict):
+                    continue
+                text = block.get("text")
+                if isinstance(text, str) and text:
+                    parts.append(text)
+        if parts:
+            return "".join(parts)
+    raise RuntimeError("LLM response content is empty")
+
+
 def call_llm_for_proposal(
     *,
     policy: LLMPolicySpec,
@@ -194,24 +233,10 @@ def call_llm_for_proposal(
         base_url=policy.endpoint.base_url,
         api_key_env=policy.endpoint.api_key_env,
         provider=policy.endpoint.provider,
+        wire_api=policy.endpoint.wire_api,
         model=policy.endpoint.model,
         messages=[{"role": "user", "content": prompt}],
         timeout_s=policy.endpoint.timeout_s,
         system_prompt=policy.system_prompt,
     )
-    choices = response.get("choices")
-    if not isinstance(choices, list) or not choices:
-        raise RuntimeError("LLM response does not contain choices")
-    message = choices[0].get("message", {})
-    if not isinstance(message, dict):
-        raise RuntimeError("LLM response does not contain a valid message")
-    content = message.get("content")
-    if isinstance(content, str):
-        return content
-    if isinstance(content, list):
-        return "".join(
-            item.get("text", "")
-            for item in content
-            if isinstance(item, dict) and isinstance(item.get("text"), str)
-        )
-    raise RuntimeError("LLM response content is empty")
+    return _extract_response_text(response)

src/aituner/spec.py

@@ -57,10 +57,10 @@ def _coerce_str_list(value: Any, *, context: str) -> list[str]:
     return result
 
 
-def _resolve_codex_base_url() -> str:
+def _resolve_codex_endpoint() -> tuple[str, str]:
     override = os.environ.get("AITUNER_CODEX_BASE_URL", "").strip()
     if override:
-        return override
+        return override, "chat.completions"
     config_path = Path.home() / ".codex" / "config.toml"
     try:
         payload = tomllib.loads(config_path.read_text(encoding="utf-8"))
@@ -81,16 +81,19 @@ def _resolve_codex_base_url() -> str:
         if isinstance(selected, Mapping):
             base_url = str(selected.get("base_url") or "").strip()
             if base_url:
-                return base_url
+                wire_api = str(selected.get("wire_api") or "chat.completions").strip()
+                return base_url, wire_api
     if len(providers) == 1:
         only_provider = next(iter(providers.values()))
         if isinstance(only_provider, Mapping):
             base_url = str(only_provider.get("base_url") or "").strip()
             if base_url:
-                return base_url
+                wire_api = str(only_provider.get("wire_api") or "chat.completions").strip()
+                return base_url, wire_api
     root_base_url = str(payload.get("base_url") or "").strip()
     if root_base_url:
-        return root_base_url
+        wire_api = str(payload.get("wire_api") or "chat.completions").strip()
+        return root_base_url, wire_api
     raise SpecError(
         "Unable to resolve llm.endpoint.base_url for provider=codex from "
         f"{config_path}. Set llm.endpoint.base_url explicitly or set "
@@ -386,6 +389,7 @@ class LLMEndpointSpec:
     base_url: str
     model: str
     provider: str = "custom"
+    wire_api: str = "chat.completions"
     api_key_env: str = "OPENAI_API_KEY"
     timeout_s: float = 120.0
 
@@ -393,20 +397,29 @@ class LLMEndpointSpec:
     def from_dict(cls, data: Mapping[str, Any]) -> "LLMEndpointSpec":
         provider = str(data.get("provider") or "custom").strip().lower()
         base_url = str(data.get("base_url") or "").strip()
+        wire_api = str(data.get("wire_api") or "").strip()
         api_key_env = str(data.get("api_key_env") or "").strip()
         if provider == "codex":
-            if not base_url:
-                base_url = _resolve_codex_base_url()
+            if not base_url or not wire_api:
+                resolved_base_url, resolved_wire_api = _resolve_codex_endpoint()
+                if not base_url:
+                    base_url = resolved_base_url
+                if not wire_api:
+                    wire_api = resolved_wire_api
             if not api_key_env:
                 api_key_env = "OPENAI_API_KEY"
         elif provider == "bailian":
             if not base_url:
                 base_url = "https://dashscope.aliyuncs.com/compatible-mode/v1"
+            if not wire_api:
+                wire_api = "chat.completions"
             if not api_key_env:
                 api_key_env = "DASHSCOPE_API_KEY"
         elif provider == "custom":
             if not base_url:
                 raise SpecError("llm.endpoint.base_url must be a non-empty string.")
+            if not wire_api:
+                wire_api = "chat.completions"
             if not api_key_env:
                 api_key_env = "OPENAI_API_KEY"
         else:
@@ -415,6 +428,7 @@ class LLMEndpointSpec:
             base_url=_require_str(base_url, context="llm.endpoint.base_url"),
             model=_require_str(data.get("model"), context="llm.endpoint.model"),
             provider=provider,
+            wire_api=_require_str(wire_api, context="llm.endpoint.wire_api"),
             api_key_env=_require_str(api_key_env, context="llm.endpoint.api_key_env"),
             timeout_s=_require_float(
                 data.get("timeout_s", 120.0), context="llm.endpoint.timeout_s"

tests/test_core_flow.py

@@ -11,7 +11,7 @@ from unittest import mock
 from aituner.cli import main as cli_main
 from aituner.http_client import _auth_headers, _openai_url, _should_bypass_proxy
 from aituner.job import append_job, build_trial_job
-from aituner.llm import build_prompt, parse_proposal_text
+from aituner.llm import _extract_response_text, build_prompt, parse_proposal_text
 from aituner.search import ThresholdProbe, binary_search_max_feasible
 from aituner.slo import RequestOutcome, evaluate_request, summarize_evaluations
 from aituner.spec import (
@@ -242,6 +242,7 @@ class CoreFlowTests(unittest.TestCase):
                         "",
                         "[model_providers.ipads]",
                         'base_url = "http://codex.example/v1"',
+                        'wire_api = "responses"',
                     ]
                 ),
                 encoding="utf-8",
@@ -250,8 +251,24 @@ class CoreFlowTests(unittest.TestCase):
                 endpoint = LLMEndpointSpec.from_dict({"provider": "codex", "model": "gpt-5.4"})
             self.assertEqual(endpoint.provider, "codex")
             self.assertEqual(endpoint.base_url, "http://codex.example/v1")
+            self.assertEqual(endpoint.wire_api, "responses")
             self.assertEqual(endpoint.api_key_env, "OPENAI_API_KEY")
 
+    def test_extract_response_text_supports_responses_api_output(self) -> None:
+        text = _extract_response_text(
+            {
+                "output": [
+                    {
+                        "type": "message",
+                        "content": [
+                            {"type": "output_text", "text": '{"diagnosis":"ok"}'}
+                        ],
+                    }
+                ]
+            }
+        )
+        self.assertEqual(text, '{"diagnosis":"ok"}')
+
     def test_auth_headers_load_bailian_key_from_dotenv(self) -> None:
         with tempfile.TemporaryDirectory() as tmp:
             tmp_path = Path(tmp)