gateway: add public invoke/jobs facade with redis queue worker and SSE

gateway-bot/daarion_facade/invoke_api.py

@@ -0,0 +1,118 @@
+import asyncio
+from datetime import datetime, timezone
+import json
+import os
+import uuid
+from typing import Any, Dict, List
+
+from fastapi import APIRouter, HTTPException, Request, status
+from fastapi.responses import StreamingResponse
+from pydantic import BaseModel, Field
+
+from .redis_jobs import create_job, enqueue_job, get_job
+from .registry_api import _load_registry
+
+router = APIRouter(prefix="/v1", tags=["daarion-facade"])
+
+EVENT_TERMINAL_STATUSES = {"done", "failed"}
+EVENT_KNOWN_STATUSES = {"queued", "running", "done", "failed"}
+EVENT_POLL_SECONDS = float(os.getenv("DAARION_JOB_EVENTS_POLL_SECONDS", "0.5"))
+
+
+class InvokeInput(BaseModel):
+    prompt: str = Field(min_length=1)
+    images: List[str] = Field(default_factory=list)
+
+
+class InvokeRequest(BaseModel):
+    agent_id: str
+    input: InvokeInput
+    metadata: Dict[str, Any] = Field(default_factory=dict)
+
+
+class InvokeResponse(BaseModel):
+    job_id: str
+    status: str
+    status_url: str
+
+
+def _sse_message(event: str, payload: Dict[str, Any]) -> str:
+    return f"event: {event}\ndata: {json.dumps(payload, ensure_ascii=False)}\n\n"
+
+
+@router.post("/invoke", status_code=status.HTTP_202_ACCEPTED, response_model=InvokeResponse)
+async def invoke(payload: InvokeRequest) -> InvokeResponse:
+    registry = _load_registry().get("agents", {})
+    if payload.agent_id not in registry:
+        raise HTTPException(status_code=404, detail=f"Unknown agent_id: {payload.agent_id}")
+
+    job_id = f"job_{uuid.uuid4().hex}"
+    now = datetime.now(timezone.utc).isoformat()
+    job_doc = {
+        "job_id": job_id,
+        "status": "queued",
+        "agent_id": payload.agent_id,
+        "input": payload.input.model_dump(),
+        "metadata": payload.metadata,
+        "result": None,
+        "error": None,
+        "created_at": now,
+        "updated_at": now,
+        "started_at": None,
+        "finished_at": None,
+    }
+    await create_job(job_id, job_doc)
+    await enqueue_job(job_id)
+    return InvokeResponse(job_id=job_id, status="queued", status_url=f"/v1/jobs/{job_id}")
+
+
+@router.get("/jobs/{job_id}")
+async def job_status(job_id: str) -> Dict[str, Any]:
+    job = await get_job(job_id)
+    if not job:
+        raise HTTPException(status_code=404, detail="Job not found")
+    return job
+
+
+@router.get("/jobs/{job_id}/events")
+async def job_events(job_id: str, request: Request) -> StreamingResponse:
+    existing = await get_job(job_id)
+    if not existing:
+        raise HTTPException(status_code=404, detail="Job not found")
+
+    async def event_stream():
+        last_state = None
+        yield "retry: 1000\n\n"
+
+        while True:
+            if await request.is_disconnected():
+                break
+
+            job = await get_job(job_id)
+            if not job:
+                yield _sse_message("failed", {"job_id": job_id, "status": "failed", "error": {"message": "Job not found"}})
+                break
+
+            status_value = str(job.get("status", "unknown"))
+            updated_at = str(job.get("updated_at", ""))
+            state = (status_value, updated_at)
+
+            if state != last_state:
+                event_name = status_value if status_value in EVENT_KNOWN_STATUSES else "status"
+                yield _sse_message(event_name, job)
+                last_state = state
+
+                if status_value in EVENT_TERMINAL_STATUSES:
+                    break
+
+            await asyncio.sleep(EVENT_POLL_SECONDS)
+
+    return StreamingResponse(
+        event_stream(),
+        media_type="text/event-stream",
+        headers={
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no",
+        },
+    )