fix(opencode): route GPT models and discover OSS fallbacks

Configure the Databricks OpenAI Responses provider alongside the validated GLM/Kimi provider, and fall back to foundation-model serving endpoints when UC model services are unavailable.

Author: dgokeeffe

src/ucode/agents/opencode.py

@@ -42,6 +42,7 @@
 PROVIDER_KEYS: list[list[str]] = [
     ["provider", "databricks-anthropic"],
     ["provider", "databricks-google"],
+    ["provider", "databricks-openai"],
     ["provider", "databricks-oss"],
 ]
 
@@ -52,7 +53,14 @@ def is_update_available() -> tuple[str, str] | None:
 
 def _resolve_model_selector(model: str, opencode_models: dict[str, list[str]]) -> str:
     """Return an OpenCode model selector in provider/model form when possible."""
-    if model.startswith(("databricks-anthropic/", "databricks-google/", "databricks-oss/")):
+    if model.startswith(
+        (
+            "databricks-anthropic/",
+            "databricks-google/",
+            "databricks-openai/",
+            "databricks-oss/",
+        )
+    ):
         return model
 
     anthropic_models = opencode_models.get("anthropic") or []
@@ -63,6 +71,10 @@ def _resolve_model_selector(model: str, opencode_models: dict[str, list[str]]) -
     if model in gemini_models:
         return f"databricks-google/{model}"
 
+    openai_models = opencode_models.get("openai") or []
+    if model in openai_models:
+        return f"databricks-openai/{model}"
+
     oss_models = opencode_models.get("oss") or []
     if model in oss_models:
         return f"databricks-oss/{model}"
@@ -102,6 +114,7 @@ def render_overlay(
 
     anthropic_models = opencode_models.get("anthropic") or []
     gemini_models = opencode_models.get("gemini") or []
+    openai_models = opencode_models.get("openai") or []
     oss_models = opencode_models.get("oss") or []
 
     providers: dict = {}
@@ -137,6 +150,28 @@ def render_overlay(
             "models": {m: {"headers": ua_header} for m in gemini_models},
         }
         keys.append(["provider", "databricks-google"])
+    if openai_models:
+        # @ai-sdk/openai supports both the Responses API and the legacy
+        # chat-completions API. Databricks GPT-5 / GPT-5.6 / Codex models are
+        # Responses-only on /ai-gateway/codex/v1, so the per-model flag
+        # `useResponsesApi: true` lives in models.<m>.options where opencode
+        # reads it (provider-level options is read by the SDK only).
+        providers["databricks-openai"] = {
+            "npm": "@ai-sdk/openai",
+            "options": {
+                "baseURL": opencode_base_urls["openai"],
+                "apiKey": token,
+                "headers": auth_headers,
+            },
+            "models": {
+                m: {
+                    "headers": ua_header,
+                    "options": {"useResponsesApi": True},
+                }
+                for m in openai_models
+            },
+        }
+        keys.append(["provider", "databricks-openai"])
     if oss_models:
         providers["databricks-oss"] = {
             "npm": "@ai-sdk/openai",
@@ -233,6 +268,9 @@ def default_model(state: dict) -> str | None:
     anthropic = opencode_models.get("anthropic") or []
     if anthropic:
         return anthropic[0]
+    openai = opencode_models.get("openai") or []
+    if openai:
+        return openai[0]
     gemini = opencode_models.get("gemini") or []
     if gemini:
         return gemini[0]

src/ucode/cli.py

@@ -37,6 +37,7 @@
     discover_codex_models,
     discover_gemini_models,
     discover_model_services,
+    discover_oss_models,
     ensure_ai_gateway_v2,
     ensure_databricks_auth,
     ensure_pat_bearer,
@@ -92,7 +93,7 @@
 
 _DISCOVERY_CONSUMERS: dict[str, tuple[str, ...]] = {
     "claude": ("claude", "opencode", "copilot", "pi"),
-    "codex": ("codex", "copilot", "pi"),
+    "codex": ("codex", "copilot", "opencode", "pi"),
     "gemini": ("gemini", "opencode", "pi"),
     "oss": ("opencode", "pi"),
 }
@@ -371,7 +372,9 @@ def configure_shared_state(
         fetch_all or "claude" in tools or "opencode" in tools or "copilot" in tools or "pi" in tools
     )
     want_gemini = fetch_all or "gemini" in tools or "opencode" in tools or "pi" in tools
-    want_codex = fetch_all or "codex" in tools or "copilot" in tools or "pi" in tools
+    want_codex = (
+        fetch_all or "codex" in tools or "copilot" in tools or "opencode" in tools or "pi" in tools
+    )
     want_oss = fetch_all or "opencode" in tools or "pi" in tools
 
     claude_reason: str | None = None
@@ -424,10 +427,14 @@ def configure_shared_state(
                     codex_models, codex_reason = discover_codex_models(workspace, token)
             if want_oss:
                 oss_models, oss_reason = ms_oss, ms_reason
+                if not oss_models:
+                    oss_models, oss_reason = discover_oss_models(workspace, token)
         if claude_models:
             opencode_models["anthropic"] = list(claude_models.values())
         if gemini_models:
             opencode_models["gemini"] = gemini_models
+        if codex_models:
+            opencode_models["openai"] = codex_models
         if oss_models:
             opencode_models["oss"] = oss_models
 

src/ucode/databricks.py

@@ -2142,6 +2142,34 @@ def discover_codex_models(workspace: str, token: str) -> tuple[list[str], str |
     return discover_endpoints_with_api_type(workspace, token, "openai/v1/responses")
 
 
+def discover_oss_models(workspace: str, token: str) -> tuple[list[str], str | None]:
+    """Discover OSS chat models served as AI Gateway foundation-model endpoints.
+
+    Fallback for workspaces that don't register OSS foundation models as
+    `system.ai.*` UC model-services (see `discover_model_services`): those
+    workspaces expose the same models as regular `databricks-*` serving
+    endpoints instead. Lists every endpoint advertising the
+    `mlflow/v1/chat/completions` dialect, then keeps only the OSS chat families
+    (`_is_oss_chat_model`) — on some workspaces the Claude/Gemini endpoints also
+    advertise that dialect, so the family filter is what separates the OSS
+    cohort from them. Mirrors the AI-Gateway fallback the other families use
+    when the UC model-services listing is empty.
+    """
+    endpoints, reason = discover_endpoints_with_api_type(
+        workspace, token, "mlflow/v1/chat/completions"
+    )
+    if not endpoints:
+        return [], reason
+    oss = [e for e in endpoints if _is_oss_chat_model(e)]
+    if oss:
+        return oss, None
+    sample = ", ".join(endpoints[:5])
+    return [], (
+        "foundation-models exposing `mlflow/v1/chat/completions` matched no OSS "
+        f"chat family (got: {sample})"
+    )
+
+
 def fetch_gemini_models(workspace: str, token: str) -> list[str]:
     models, _ = discover_gemini_models(workspace, token)
     return models
@@ -2339,6 +2367,9 @@ def build_opencode_base_urls(workspace: str) -> dict[str, str]:
     return {
         "anthropic": build_tool_base_url("claude", workspace) + "/v1",
         "gemini": build_tool_base_url("gemini", workspace) + "/v1beta",
+        # @ai-sdk/openai appends "/responses" (or "/chat/completions") to baseURL,
+        # so stop just before that — matches the Pi adapter's build_pi_base_urls.
+        "openai": build_tool_base_url("codex", workspace),
         "oss": f"{workspace}/ai-gateway/mlflow/v1",
     }
 

tests/test_agent_opencode.py

@@ -14,6 +14,7 @@ def _base_urls() -> dict[str, str]:
     return {
         "anthropic": f"{WS}/ai-gateway/anthropic/v1",
         "gemini": f"{WS}/ai-gateway/gemini/v1beta",
+        "openai": f"{WS}/ai-gateway/codex/v1",
         "oss": f"{WS}/ai-gateway/mlflow/v1",
     }
 
@@ -211,6 +212,76 @@ def test_prefixes_oss_model_with_provider_id(self):
         assert overlay["model"] == "databricks-oss/system.ai.kimi-k2-7-code"
 
 
+class TestOpenAIProvider:
+    """OpenCode reaches Databricks GPT-5 / GPT-5.6 / Codex models through the
+    databricks-openai provider (@ai-sdk/openai against /ai-gateway/codex/v1).
+    Without this wiring the codex/openai family is unreachable from OpenCode."""
+
+    def test_openai_provider_added_when_codex_models_present(self):
+        models = {"openai": ["databricks-gpt-5-6-sol"]}
+        overlay, _ = opencode.render_overlay("databricks-gpt-5-6-sol", "tok", _base_urls(), models)
+        assert "databricks-openai" in overlay["provider"]
+
+    def test_openai_provider_uses_ai_sdk_openai_npm(self):
+        models = {"openai": ["databricks-gpt-5-6-sol"]}
+        overlay, _ = opencode.render_overlay("databricks-gpt-5-6-sol", "tok", _base_urls(), models)
+        assert overlay["provider"]["databricks-openai"]["npm"] == "@ai-sdk/openai"
+
+    def test_openai_base_url_points_at_codex_gateway(self):
+        models = {"openai": ["databricks-gpt-5-6-sol"]}
+        overlay, _ = opencode.render_overlay("databricks-gpt-5-6-sol", "tok", _base_urls(), models)
+        options = overlay["provider"]["databricks-openai"]["options"]
+        assert options["baseURL"] == f"{WS}/ai-gateway/codex/v1"
+
+    def test_use_responses_api_set_on_every_codex_model(self):
+        models = {"openai": ["databricks-gpt-5-6-sol", "databricks-gpt-codex"]}
+        overlay, _ = opencode.render_overlay("databricks-gpt-5-6-sol", "tok", _base_urls(), models)
+        provider_models = overlay["provider"]["databricks-openai"]["models"]
+        for m in ("databricks-gpt-5-6-sol", "databricks-gpt-codex"):
+            assert provider_models[m]["options"]["useResponsesApi"] is True
+
+    def test_openai_authorization_header(self):
+        models = {"openai": ["databricks-gpt-5-6-sol"]}
+        overlay, _ = opencode.render_overlay("databricks-gpt-5-6-sol", "tok", _base_urls(), models)
+        headers = overlay["provider"]["databricks-openai"]["options"]["headers"]
+        assert headers["Authorization"] == "Bearer tok"
+
+    def test_managed_keys_include_openai_provider(self):
+        models = {"openai": ["databricks-gpt-5-6-sol"]}
+        _, keys = opencode.render_overlay("databricks-gpt-5-6-sol", "tok", _base_urls(), models)
+        assert ["provider", "databricks-openai"] in keys
+
+    def test_prefixes_openai_model_with_provider_id(self):
+        models = {"openai": ["databricks-gpt-5-6-sol"]}
+        overlay, _ = opencode.render_overlay("databricks-gpt-5-6-sol", "tok", _base_urls(), models)
+        assert overlay["model"] == "databricks-openai/databricks-gpt-5-6-sol"
+
+    def test_already_prefixed_openai_model_is_preserved(self):
+        models = {"openai": ["databricks-gpt-5-6-sol"]}
+        overlay, _ = opencode.render_overlay(
+            "databricks-openai/databricks-gpt-5-6-sol", "tok", _base_urls(), models
+        )
+        assert overlay["model"] == "databricks-openai/databricks-gpt-5-6-sol"
+
+    def test_all_four_providers_when_all_present(self):
+        models = {
+            "anthropic": ["claude-sonnet"],
+            "gemini": ["gemini-2"],
+            "openai": ["databricks-gpt-5-6-sol"],
+            "oss": ["system.ai.kimi-k2-7-code"],
+        }
+        overlay, _ = opencode.render_overlay("claude-sonnet", "tok", _base_urls(), models)
+        assert set(overlay["provider"].keys()) == {
+            "databricks-anthropic",
+            "databricks-google",
+            "databricks-openai",
+            "databricks-oss",
+        }
+
+    def test_provider_keys_listed_in_module(self):
+        assert ["provider", "databricks-openai"] in opencode.PROVIDER_KEYS
+
+
 class TestMcpServerConfig:
     def test_builds_remote_server_entry_with_oauth_token_env_header(self):
         entry = opencode.build_mcp_server_entry(f"{WS}/api/2.0/mcp/external/github")
@@ -323,6 +394,16 @@ def test_prefers_anthropic(self):
         state = {"opencode_models": {"anthropic": ["claude-sonnet"], "gemini": ["gemini-2"]}}
         assert opencode.default_model(state) == "claude-sonnet"
 
+    def test_falls_back_to_openai_before_gemini(self):
+        state = {
+            "opencode_models": {
+                "anthropic": [],
+                "openai": ["databricks-gpt-5-6-sol"],
+                "gemini": ["gemini-2"],
+            }
+        }
+        assert opencode.default_model(state) == "databricks-gpt-5-6-sol"
+
     def test_falls_back_to_gemini(self):
         state = {"opencode_models": {"anthropic": [], "gemini": ["gemini-2"]}}
         assert opencode.default_model(state) == "gemini-2"

tests/test_databricks.py

@@ -102,6 +102,12 @@ def test_returns_anthropic_gemini_and_oss(self):
         assert urls["gemini"] == f"{WS}/ai-gateway/gemini/v1beta"
         assert urls["oss"] == f"{WS}/ai-gateway/mlflow/v1"
 
+    def test_returns_openai_codex_gateway(self):
+        # @ai-sdk/openai appends /responses (Responses API) or /chat/completions
+        # to baseURL, so stop just before that suffix. Mirrors build_pi_base_urls.
+        urls = build_opencode_base_urls(WS)
+        assert urls["openai"] == f"{WS}/ai-gateway/codex/v1"
+
 
 class TestBuildSharedBaseUrls:
     def test_contains_all_tools(self):
@@ -919,6 +925,77 @@ def test_codex_discovery_keeps_alphabetical_order(self, monkeypatch):
         assert models == ["databricks-gpt-4-1", "databricks-gpt-5-2-codex"]
 
 
+def _mlflow_chat_payload(names, *, api_type="mlflow/v1/chat/completions", v2=True):
+    return {
+        "endpoints": [
+            {
+                "name": name,
+                "config": {
+                    "served_entities": [
+                        {
+                            "foundation_model": {
+                                "ai_gateway_v2_supported": v2,
+                                "api_types": [api_type],
+                            }
+                        }
+                    ]
+                },
+            }
+            for name in names
+        ]
+    }
+
+
+class TestDiscoverOssModels:
+    def test_finds_oss_endpoints_via_foundation_models(self, monkeypatch):
+        # Mirrors a workspace with no system.ai UC model-services: OSS models are
+        # plain databricks-* serving endpoints under the mlflow chat dialect.
+        payload = _mlflow_chat_payload(
+            [
+                "databricks-glm-5-2",
+                "databricks-kimi-k2-7-code",
+                "databricks-inkling",
+                "databricks-qwen35-122b-a10b",
+                "databricks-gemma-3-12b",
+            ]
+        )
+        monkeypatch.setattr(db_mod, "_http_get_json", lambda url, token: (payload, None))
+
+        models, reason = db_mod.discover_oss_models(WS, "token")
+
+        assert reason is None
+        assert models == ["databricks-glm-5-2", "databricks-kimi-k2-7-code"]
+
+    def test_excludes_claude_and_gemini_sharing_the_mlflow_dialect(self, monkeypatch):
+        # On some workspaces every foundation model advertises the mlflow chat
+        # dialect, so the api_type filter alone is too broad — the OSS family
+        # filter must drop Claude/Gemini and keep only the OSS cohort.
+        payload = _mlflow_chat_payload(
+            [
+                "databricks-claude-opus-4-8",
+                "databricks-gemini-2-5-pro",
+                "databricks-glm-5-2",
+                "databricks-qwen3-embedding-0-6b",
+            ]
+        )
+        monkeypatch.setattr(db_mod, "_http_get_json", lambda url, token: (payload, None))
+
+        models, reason = db_mod.discover_oss_models(WS, "token")
+
+        assert reason is None
+        assert models == ["databricks-glm-5-2"]
+
+    def test_reports_reason_when_no_oss_family_matches(self, monkeypatch):
+        payload = _mlflow_chat_payload(["databricks-claude-opus-4-8"])
+        monkeypatch.setattr(db_mod, "_http_get_json", lambda url, token: (payload, None))
+
+        models, reason = db_mod.discover_oss_models(WS, "token")
+
+        assert models == []
+        assert reason is not None
+        assert "no OSS" in reason
+
+
 class TestResolvePatToken:
     def test_reads_pat_profile_token_from_cfg(self, monkeypatch, tmp_path):
         cfg = tmp_path / "databrickscfg"