First commit

apps/ai_gateway/Dockerfile

@@ -0,0 +1,7 @@
+FROM python:3.11-slim
+WORKDIR /app
+COPY apps/ai_gateway/requirements.txt /app/requirements.txt
+RUN pip install --no-cache-dir -r /app/requirements.txt
+COPY apps/ai_gateway /app
+EXPOSE 9100
+CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "9100"]

apps/ai_gateway/README.md

@@ -0,0 +1,18 @@
+# AI Gateway
+
+Camada desacoplada para abstração, roteamento, controle e governança de chamadas LLM.
+
+Este componente não substitui o Agent Runtime. Ele centraliza políticas de modelo, seleção de provider, fallback, telemetria e controles corporativos.
+
+## Rotas
+
+- `GET /health`
+- `GET /models`
+- `POST /v1/chat/completions`
+
+## Execução local
+
+```bash
+cd apps/ai_gateway
+uvicorn app.main:app --host 0.0.0.0 --port 9100 --reload
+```

apps/ai_gateway/app/main.py

@@ -0,0 +1,81 @@
+from __future__ import annotations
+
+import os
+from typing import Any, Literal
+
+import httpx
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel, Field
+
+
+class Message(BaseModel):
+    role: Literal["system", "user", "assistant", "tool"]
+    content: str | list[Any] | None = None
+
+
+class ChatCompletionRequest(BaseModel):
+    model: str | None = None
+    messages: list[Message]
+    temperature: float | None = None
+    max_tokens: int | None = None
+    metadata: dict[str, Any] = Field(default_factory=dict)
+
+
+class ModelRoute(BaseModel):
+    provider: str
+    model: str
+    base_url: str | None = None
+
+
+app = FastAPI(title="Agent Framework OCI - AI Gateway", version="0.1.0")
+
+
+def resolve_route(requested_model: str | None, metadata: dict[str, Any]) -> ModelRoute:
+    provider = metadata.get("provider") or os.getenv("AI_GATEWAY_DEFAULT_PROVIDER", "oci_openai")
+    model = requested_model or metadata.get("profile_model") or os.getenv("AI_GATEWAY_DEFAULT_MODEL", "openai.gpt-4.1")
+    base_url = os.getenv("AI_GATEWAY_OPENAI_COMPAT_BASE_URL")
+    return ModelRoute(provider=provider, model=model, base_url=base_url)
+
+
+@app.get("/health")
+def health() -> dict[str, Any]:
+    return {"status": "ok", "component": "ai_gateway"}
+
+
+@app.get("/models")
+def models() -> dict[str, Any]:
+    return {
+        "default_provider": os.getenv("AI_GATEWAY_DEFAULT_PROVIDER", "oci_openai"),
+        "default_model": os.getenv("AI_GATEWAY_DEFAULT_MODEL", "openai.gpt-4.1"),
+        "purpose": "model routing, policy control and LLM abstraction",
+    }
+
+
+@app.post("/v1/chat/completions")
+async def chat_completions(payload: ChatCompletionRequest) -> dict[str, Any]:
+    route = resolve_route(payload.model, payload.metadata)
+    # Safe default: when no upstream is configured, return route decision only.
+    # Production deployments should configure AI_GATEWAY_OPENAI_COMPAT_BASE_URL and credentials.
+    if not route.base_url:
+        return {
+            "gateway": "ai_gateway",
+            "mode": "dry_run",
+            "route": route.model_dump(),
+            "message": "No upstream base URL configured. Set AI_GATEWAY_OPENAI_COMPAT_BASE_URL to proxy requests.",
+        }
+
+    api_key = os.getenv("AI_GATEWAY_API_KEY") or os.getenv("OCI_GENAI_API_KEY") or os.getenv("OPENAI_API_KEY")
+    headers = {"Authorization": f"Bearer {api_key}"} if api_key else {}
+    body = payload.model_dump(exclude_none=True)
+    body["model"] = route.model
+    try:
+        async with httpx.AsyncClient(timeout=float(os.getenv("AI_GATEWAY_TIMEOUT_SECONDS", "60"))) as client:
+            resp = await client.post(f"{route.base_url.rstrip('/')}/chat/completions", json=body, headers=headers)
+            resp.raise_for_status()
+            data = resp.json()
+            if isinstance(data, dict):
+                data.setdefault("gateway", "ai_gateway")
+                data.setdefault("route", route.model_dump())
+            return data
+    except httpx.HTTPError as exc:
+        raise HTTPException(status_code=502, detail=f"AI upstream request failed: {exc}") from exc

apps/ai_gateway/requirements.txt

@@ -0,0 +1,5 @@
+fastapi>=0.111
+uvicorn[standard]>=0.30
+pydantic>=2
+httpx>=0.27
+PyYAML>=6