Merge remote-tracking branch 'origin/main' into pr-3379

nanobot/providers/openai_compat_provider.py

@@ -446,10 +446,11 @@ class OpenAICompatProvider(LLMProvider):
         reasoning_effort: str | None,
     ) -> bool:
         """Use Responses API only for direct OpenAI requests that benefit from it."""
-        if self._spec and self._spec.name != "openai":
-            return False
-        if not _is_direct_openai_base(self._effective_base):
+        if self._spec and self._spec.name not in ("openai", "github_copilot"):
             return False
+        if self._spec is None or self._spec.name != "github_copilot":
+            if not _is_direct_openai_base(self._effective_base):
+                return False
 
         model_name = (model or self.default_model).lower()
         wants = False
@@ -527,6 +528,8 @@ class OpenAICompatProvider(LLMProvider):
     ) -> dict[str, Any]:
         """Build a Responses API body for direct OpenAI requests."""
         model_name = model or self.default_model
+        if self._spec and self._spec.strip_model_prefix:
+            model_name = model_name.split("/")[-1]
         sanitized_messages = self._sanitize_messages(self._sanitize_empty_content(messages))
         instructions, input_items = convert_messages(sanitized_messages)
 
@@ -987,6 +990,11 @@ class OpenAICompatProvider(LLMProvider):
                     self._record_responses_success(model, reasoning_effort)
                     return result
                 except Exception as responses_error:
+                    if self._spec and self._spec.name == "github_copilot":
+                        # Copilot gateway exposes GPT-5/o-series only via /responses;
+                        # falling back to /chat/completions cannot succeed and would
+                        # hide the real error.
+                        raise
                     if not self._should_fallback_from_responses_error(responses_error):
                         raise
                     self._record_responses_failure(model, reasoning_effort)
@@ -1045,6 +1053,11 @@ class OpenAICompatProvider(LLMProvider):
                         reasoning_content=reasoning_content,
                     )
                 except Exception as responses_error:
+                    if self._spec and self._spec.name == "github_copilot":
+                        # Copilot gateway exposes GPT-5/o-series only via /responses;
+                        # falling back to /chat/completions cannot succeed and would
+                        # hide the real error.
+                        raise
                     if not self._should_fallback_from_responses_error(responses_error):
                         raise
                     self._record_responses_failure(model, reasoning_effort)

nanobot/providers/registry.py

@@ -223,6 +223,7 @@ PROVIDERS: tuple[ProviderSpec, ...] = (
         default_api_base="https://api.githubcopilot.com",
         strip_model_prefix=True,
         is_oauth=True,
+        supports_max_completion_tokens=True,
     ),
     # DeepSeek: OpenAI-compatible at api.deepseek.com
     ProviderSpec(

tests/cli/test_commands.py

@@ -421,13 +421,13 @@ async def test_github_copilot_provider_refreshes_client_api_key_before_chat():
     })
 
     with patch("nanobot.providers.openai_compat_provider.AsyncOpenAI", return_value=mock_client):
-        provider = GitHubCopilotProvider(default_model="github-copilot/gpt-5.1")
+        provider = GitHubCopilotProvider(default_model="github-copilot/gpt-4")
 
     provider._get_copilot_access_token = AsyncMock(return_value="copilot-access-token")
 
     response = await provider.chat(
         messages=[{"role": "user", "content": "hi"}],
-        model="github-copilot/gpt-5.1",
+        model="github-copilot/gpt-4",
         max_tokens=16,
         temperature=0.1,
     )

tests/providers/test_github_copilot_routing.py

@@ -0,0 +1,79 @@
+"""Regression tests for GitHub Copilot /responses routing.
+
+Covers the Copilot-specific branches added to route GPT-5 / o-series models
+through the /responses endpoint without falling back to /chat/completions.
+"""
+
+from __future__ import annotations
+
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+from nanobot.providers.openai_compat_provider import OpenAICompatProvider
+from nanobot.providers.registry import find_by_name
+
+
+def _make_copilot_provider() -> OpenAICompatProvider:
+    """Build a bare provider with the real github_copilot spec (no network)."""
+    p = OpenAICompatProvider.__new__(OpenAICompatProvider)
+    p.default_model = "github_copilot/gpt-5.4-mini"
+    p._spec = find_by_name("github_copilot")
+    p._effective_base = "https://api.githubcopilot.com"
+    p._responses_failures = {}
+    p._responses_tripped_at = {}
+    return p
+
+
+def test_should_use_responses_api_allows_github_copilot_non_openai_base():
+    """github_copilot bypasses the direct-OpenAI base check and still opts in for GPT-5."""
+    provider = _make_copilot_provider()
+    assert provider._should_use_responses_api("github_copilot/gpt-5.4-mini", None) is True
+    assert provider._should_use_responses_api("github_copilot/o3", None) is True
+
+
+def test_build_responses_body_strips_github_copilot_prefix():
+    """/responses body must send the bare model name; gateway rejects routing prefixes."""
+    provider = _make_copilot_provider()
+    body = provider._build_responses_body(
+        messages=[{"role": "user", "content": "hi"}],
+        tools=None,
+        model="github_copilot/gpt-5.4-mini",
+        max_tokens=16,
+        temperature=0.1,
+        reasoning_effort=None,
+        tool_choice=None,
+    )
+    assert body["model"] == "gpt-5.4-mini"
+
+
+@pytest.mark.asyncio
+async def test_github_copilot_does_not_fall_back_from_responses_error():
+    """On /responses failure, github_copilot must re-raise instead of hitting /chat/completions."""
+    from nanobot.providers.github_copilot_provider import GitHubCopilotProvider
+
+    mock_client = MagicMock()
+    mock_client.api_key = "no-key"
+
+    class _CompatError(Exception):
+        """Looks like a fallback-eligible error on other providers."""
+        status_code = 400
+        body = "Unsupported parameter responses api"
+
+    mock_client.responses.create = AsyncMock(side_effect=_CompatError("boom"))
+    mock_client.chat.completions.create = AsyncMock()
+
+    with patch("nanobot.providers.openai_compat_provider.AsyncOpenAI", return_value=mock_client):
+        provider = GitHubCopilotProvider(default_model="github_copilot/gpt-5.4-mini")
+    provider._get_copilot_access_token = AsyncMock(return_value="copilot-access-token")
+
+    response = await provider.chat(
+        messages=[{"role": "user", "content": "hi"}],
+        model="github_copilot/gpt-5.4-mini",
+        max_tokens=16,
+        temperature=0.1,
+    )
+
+    assert response.finish_reason == "error"
+    mock_client.responses.create.assert_awaited_once()
+    mock_client.chat.completions.create.assert_not_awaited()