docs: add API guide for AI agents

Complete reference for interacting with OpusAgent: auth, async flow,
multi-turn sessions, model resolution, summarization, and error handling.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

CLAUDE.md

@@ -0,0 +1,125 @@
+# OpusAgent
+
+Self-hosted REST API proxy to Claude models via Claude Code SDK, leveraging a Claude Pro/Max subscription.
+
+## Architecture
+
+Two Docker containers behind Traefik reverse proxy:
+
+```
+Client → Traefik (HTTPS) → FastAPI (Python, :8000) ↔ Node.js Worker (:3001, internal)
+                                    ↓                         ↓
+                              SQLite (WAL)           Claude Code SDK
+```
+
+- **fastapi-service**: API REST, auth, rate limiting, async queue processing
+- **claude-worker**: Express server wrapping `@anthropic-ai/claude-code` SDK
+
+## Key Paths
+
+| Area | Path |
+|------|------|
+| FastAPI entrypoint | `src/backend/main.py` |
+| Config (Pydantic Settings) | `src/backend/config.py` |
+| Database layer (aiosqlite) | `src/backend/db/database.py` |
+| API models + envelope | `src/backend/models/api.py` |
+| Routes | `src/backend/api/routes/{health,topics,requests,sessions}.py` |
+| Auth + rate limiter | `src/backend/api/dependencies.py` |
+| Worker HTTP client | `src/backend/services/worker_client.py` |
+| Queue processor | `src/backend/services/queue.py` |
+| Node.js worker | `worker/index.js` |
+| Tests | `tests/` |
+| Design spec (Italian) | `docs/superpowers/specs/` |
+| Implementation plan | `docs/superpowers/plans/` |
+
+## Tech Stack
+
+- **Python 3.11+**: FastAPI, uvicorn, aiosqlite, httpx, slowapi, pydantic-settings
+- **Node.js 20**: Express, @anthropic-ai/claude-code SDK (v1.0.128+, async iterable API)
+- **Package manager**: uv (Python), npm (Node.js)
+- **Database**: SQLite with WAL mode, foreign keys enabled
+- **Infra**: Docker Compose, Traefik v3 (TLS via Let's Encrypt)
+
+## Database Schema
+
+Three tables in SQLite:
+
+- **topics**: id (UUID), name (unique), system_prompt, summary, model (nullable), timestamps
+- **requests**: id (UUID), session_id, sdk_conversation_id, topic_id (FK), prompt, result, error, model, summarize (bool), status (pending/processing/completed/failed), timestamps
+- **closed_sessions**: session_id (PK), closed_at
+
+## API Endpoints
+
+All require `X-Api-Key` header. All responses use envelope: `{success, data, error}`.
+
+- `GET /api/health` — health check (DB + worker)
+- `GET /api/models` — list available Claude models (aliases + default)
+- `POST|GET /api/topics`, `GET|PUT|DELETE /api/topics/{id}` — topic CRUD
+- `POST /api/topics/{id}/summarize` — regenerate complete topic summary
+- `POST /api/requests` (202) — submit prompt (async queue), `GET|DELETE /api/requests[/{id}]`
+- `GET /api/sessions`, `DELETE /api/sessions/{id}` — session management
+
+## Processing Flow
+
+1. `POST /api/requests` → 202 Accepted (queued)
+2. Background loop picks oldest pending, calls worker `POST /process`
+3. Worker runs Claude Code SDK (`query()` async iterable) with optional session resume
+4. Result stored in DB; SDK conversation ID saved in `sdk_conversation_id`; optional incremental summary appended to topic
+5. Client polls `GET /api/requests/{id}` for result
+
+Model resolution: request-level > topic-level > `CLAUDE_MODEL` env var.
+
+### Multi-turn sessions
+
+- `session_id: "new"` → stored as NULL; after completion, set to SDK's conversation ID
+- Custom `session_id` → preserved as-is; SDK conversation ID stored separately in `sdk_conversation_id`
+- Resume logic: queue looks up `sdk_conversation_id` from previous completed requests in the same session; only passes it to worker if found (avoids resume on non-existent SDK sessions)
+
+## Commands
+
+```bash
+# Development
+uv sync --all-groups
+uv run dev                          # uvicorn with --reload
+
+# Production
+docker compose up -d                # both services
+docker compose exec claude-worker npx claude login  # first-time auth (must run on active container, NOT run --rm)
+
+# Tests
+uv run pytest -v
+
+# Logs
+docker compose logs -f fastapi-service
+docker compose logs -f claude-worker
+```
+
+## Environment Variables (.env)
+
+| Variable | Default | Purpose |
+|----------|---------|---------|
+| `API_KEY` | (required) | Auth key for all endpoints |
+| `API_HOST` | 0.0.0.0 | Server bind address |
+| `API_PORT` | 8000 | Server port |
+| `WORKER_URL` | http://claude-worker:3001 | Internal worker address |
+| `DB_PATH` | /data/opus-agent.db | SQLite database path |
+| `CLAUDE_MODEL` | claude-sonnet-4-6 | Default Claude model |
+| `RATE_LIMIT_PER_MINUTE` | 10 | Rate limit per API key |
+| `RATE_LIMIT_PER_HOUR` | 100 | Rate limit per API key |
+
+## Design Decisions
+
+- **Sequential processing**: one request at a time (FIFO), no concurrency — simplifies Claude Code SDK session handling
+- **Async Python**: FastAPI + aiosqlite for non-blocking I/O while queue runs in background
+- **SQLite + WAL**: sufficient for single-writer pattern, no external DB needed
+- **Envelope pattern**: consistent `{success, data, error}` on all responses
+- **Session = conversation**: session_id maps to Claude Code SDK `resume` option
+- **Worker isolation**: not exposed to host, only reachable from fastapi-service via Docker internal network
+- **Claude auth volume**: credentials stored at `/root/.claude` (not `/home/node/.claude`) because the worker process runs as root inside the container
+- **SDK async iteration**: `query()` returns an async iterable (not a promise); iterate with `for await...of` and extract the message with `type === "result"`
+
+## Deployment
+
+Domain: `opus-agent.tielogic.xyz` → Traefik → fastapi-service:8000
+
+Traefik config: `/opt/docker/traefik/dynamic.yml` (file provider, auto-reload).

Dockerfile

@@ -11,4 +11,4 @@ COPY src/ src/
 
 EXPOSE 8000
 
-CMD ["uv", "run", "start"]
+CMD ["uv", "run", "uvicorn", "src.backend.main:app", "--host", "0.0.0.0", "--port", "8000"]

README.md

@@ -24,16 +24,18 @@ Due container Docker:
    # Edita .env: imposta API_KEY e le altre variabili
    ```
 
-2. Autentica Claude Code nel worker:
-   ```bash
-   docker compose run --rm claude-worker npx claude login
-   ```
-
-3. Avvia il sistema:
+2. Avvia il sistema:
    ```bash
    docker compose up -d
    ```
 
+3. Autentica Claude Code nel worker (nel container attivo, non con `run --rm`):
+   ```bash
+   docker compose exec claude-worker npx claude login
+   ```
+   Le credenziali vengono salvate nel volume Docker `claude-auth` (mount: `/root/.claude`)
+   e persistono tra riavvii e rebuild. Si perdono solo con `docker compose down -v`.
+
 4. Verifica:
    ```bash
    curl -H "X-Api-Key: TUA_KEY" http://localhost:8000/api/health
@@ -56,7 +58,18 @@ uv run pytest -v
 
 ## API
 
-Documentazione completa: `http://localhost:8000/doc`
+Documentazione interattiva:
+- Swagger UI: `https://opus-agent.tielogic.xyz/docs`
+- ReDoc: `https://opus-agent.tielogic.xyz/redoc`
+
+### Modelli disponibili
+
+```bash
+curl http://localhost:8000/api/models -H "X-Api-Key: TUA_KEY"
+```
+
+Ritorna i modelli disponibili dall'SDK (`default`, `opus`, `sonnet`, `opusplan`) e il modello
+di default configurato. Gli alias possono essere usati nel campo `model` di topic e richieste.
 
 ### Esempio d'uso
 
@@ -77,3 +90,23 @@ curl -X POST http://localhost:8000/api/requests \
 curl http://localhost:8000/api/requests/ID_RICHIESTA \
   -H "X-Api-Key: TUA_KEY"
 ```
+
+### Sessioni multi-turn
+
+```bash
+# Primo messaggio: session_id "new" o un ID personalizzato
+curl -X POST http://localhost:8000/api/requests \
+  -H "X-Api-Key: TUA_KEY" \
+  -H "Content-Type: application/json" \
+  -d '{"prompt": "Mi chiamo Marco.", "topic_id": "ID_TOPIC", "session_id": "new"}'
+
+# Il risultato contiene il session_id (generato dall'SDK)
+# Usalo per i messaggi successivi nella stessa conversazione
+curl -X POST http://localhost:8000/api/requests \
+  -H "X-Api-Key: TUA_KEY" \
+  -H "Content-Type: application/json" \
+  -d '{"prompt": "Come mi chiamo?", "topic_id": "ID_TOPIC", "session_id": "SESSION_ID_DAL_RISULTATO"}'
+```
+
+Il campo `sdk_conversation_id` nella risposta contiene l'ID interno della conversazione
+Claude Code SDK, usato automaticamente per il resume nei turni successivi.

docker-compose.yml

@@ -33,7 +33,7 @@ services:
     env_file: .env
     volumes:
       - opus-data:/data
-      - claude-auth:/home/node/.claude
+      - claude-auth:/root/.claude
     restart: unless-stopped
     healthcheck:
       test: ["CMD", "wget", "--spider", "-q", "http://localhost:3001/health"]

docs/agent-guide.md

@@ -0,0 +1,264 @@
+# OpusAgent - Guida per Agenti AI
+
+Questo documento spiega come interagire con l'API REST di OpusAgent.
+OpusAgent e' un proxy che inoltra richieste ai modelli Claude tramite Claude Code SDK.
+
+## Informazioni base
+
+- **Base URL**: `https://opus-agent.tielogic.xyz`
+- **Autenticazione**: header `X-Api-Key` obbligatorio su ogni richiesta
+- **Formato risposte**: JSON con envelope `{"success": bool, "data": any, "error": {"code": str, "message": str} | null}`
+- **Elaborazione asincrona**: le richieste vengono accodate e processate sequenzialmente (FIFO)
+
+## Flusso operativo
+
+### 1. Crea un topic
+
+Un topic definisce il contesto della conversazione (system prompt) e opzionalmente un modello di default.
+
+```
+POST /api/topics
+Content-Type: application/json
+X-Api-Key: TUA_KEY
+
+{
+  "name": "analisi-dati",
+  "system_prompt": "Sei un esperto analista dati. Rispondi in italiano.",
+  "model": "sonnet"  // opzionale, default dal server
+}
+```
+
+Risposta (201):
+```json
+{
+  "success": true,
+  "data": {
+    "id": "uuid-del-topic",
+    "name": "analisi-dati",
+    "system_prompt": "...",
+    "summary": "",
+    "model": "sonnet",
+    "created_at": "...",
+    "updated_at": "..."
+  }
+}
+```
+
+### 2. Invia una richiesta
+
+Le richieste vengono accodate e processate in background. La risposta immediata e' un 202 Accepted con l'ID della richiesta.
+
+```
+POST /api/requests
+Content-Type: application/json
+X-Api-Key: TUA_KEY
+
+{
+  "topic_id": "uuid-del-topic",
+  "prompt": "Analizza i trend di vendita Q1 2026",
+  "model": "opus",       // opzionale, sovrascrive il modello del topic
+  "session_id": "new",   // opzionale, per conversazioni multi-turn
+  "summarize": true       // opzionale, genera riassunto incrementale nel topic
+}
+```
+
+Risposta (202):
+```json
+{
+  "success": true,
+  "data": {
+    "id": "uuid-della-richiesta",
+    "session_id": null,
+    "status": "pending",
+    "created_at": "..."
+  }
+}
+```
+
+### 3. Attendi il risultato (polling)
+
+Interroga periodicamente lo stato della richiesta fino a `status: "completed"` o `status: "failed"`.
+
+```
+GET /api/requests/{id}
+X-Api-Key: TUA_KEY
+```
+
+Risposta quando completata:
+```json
+{
+  "success": true,
+  "data": {
+    "id": "uuid-della-richiesta",
+    "session_id": "sdk-conversation-id",
+    "topic_id": "uuid-del-topic",
+    "prompt": "Analizza i trend di vendita Q1 2026",
+    "model": "opus",
+    "summarize": true,
+    "status": "completed",
+    "result": "Risposta di Claude...",
+    "error": null,
+    "created_at": "...",
+    "completed_at": "...",
+    "sdk_conversation_id": "sdk-uuid"
+  }
+}
+```
+
+**Intervallo di polling consigliato**: 2-3 secondi. Il tempo medio di risposta e' 3-5 secondi per richieste semplici.
+
+**Stati possibili**: `pending` -> `processing` -> `completed` | `failed`
+
+## Conversazioni multi-turn
+
+Per mantenere il contesto tra messaggi successivi, usa le sessioni.
+
+### Prima richiesta
+
+Invia con `"session_id": "new"` oppure un ID personalizzato a tua scelta.
+
+```json
+{
+  "topic_id": "uuid",
+  "prompt": "Mi chiamo Marco. Lavorero' su un progetto Python.",
+  "session_id": "new"
+}
+```
+
+Il `session_id` nella risposta 202 sara' `null`. Dopo il completamento, il campo `session_id` conterra' l'ID della conversazione generato dall'SDK.
+
+### Richieste successive
+
+Usa il `session_id` ottenuto dal risultato della prima richiesta completata.
+
+```json
+{
+  "topic_id": "uuid",
+  "prompt": "Come mi chiamo? Su cosa lavoriamo?",
+  "session_id": "session-id-dal-risultato-precedente"
+}
+```
+
+Claude ricordera' tutto il contesto della conversazione.
+
+### Session ID personalizzato
+
+Puoi passare un tuo ID (es. `"progetto-alpha-chat-1"`) invece di `"new"`. Il sistema lo mantiene come identificativo della sessione e gestisce internamente la mappatura con la conversazione SDK.
+
+### Chiudere una sessione
+
+```
+DELETE /api/sessions/{session_id}
+X-Api-Key: TUA_KEY
+```
+
+Dopo la chiusura, nuove richieste con lo stesso `session_id` verranno rifiutate con status 410 Gone.
+
+## Risoluzione del modello
+
+Il modello usato segue questa priorita' (dal piu' alto al piu' basso):
+
+1. **Campo `model` nella richiesta** - sovrascrive tutto
+2. **Campo `model` nel topic** - default per il topic
+3. **Variabile d'ambiente `CLAUDE_MODEL`** - default globale del server
+
+### Valori accettati per il campo model
+
+Alias corti (consigliati):
+- `default` - Opus o Sonnet in base ai limiti d'uso
+- `opus` - Opus 4.1, per task complessi
+- `sonnet` - Sonnet 4, per uso quotidiano
+- `opusplan` - Opus in plan mode, Sonnet altrimenti
+
+Model ID completi:
+- `claude-opus-4-7`
+- `claude-sonnet-4-6`
+- `claude-haiku-4-5-20251001`
+
+Per ottenere la lista aggiornata dei modelli disponibili:
+
+```
+GET /api/models
+X-Api-Key: TUA_KEY
+```
+
+## Riassunti automatici
+
+### Riassunto incrementale
+
+Aggiungi `"summarize": true` alla richiesta. Dopo il completamento, una sintesi di 1-2 frasi viene appesa al campo `summary` del topic con la data.
+
+### Riassunto completo
+
+Rigenera il riassunto del topic analizzando tutte le richieste completate:
+
+```
+POST /api/topics/{id}/summarize
+X-Api-Key: TUA_KEY
+```
+
+Il summary viene sovrascritto con un'overview completa di tutte le interazioni.
+
+## Riferimento completo degli endpoint
+
+| Metodo | Endpoint | Descrizione |
+|--------|----------|-------------|
+| GET | `/api/health` | Stato del sistema (DB, worker) |
+| GET | `/api/models` | Modelli Claude disponibili |
+| POST | `/api/topics` | Crea topic |
+| GET | `/api/topics` | Lista topic |
+| GET | `/api/topics/{id}` | Dettaglio topic |
+| PUT | `/api/topics/{id}` | Aggiorna topic |
+| DELETE | `/api/topics/{id}` | Elimina topic (fallisce se ha richieste pendenti) |
+| POST | `/api/topics/{id}/summarize` | Rigenera riassunto completo |
+| POST | `/api/requests` | Invia richiesta (202 Accepted) |
+| GET | `/api/requests` | Lista richieste (filtri: `?status=`, `?topic_id=`, `?session_id=`) |
+| GET | `/api/requests/{id}` | Stato e risultato richiesta |
+| DELETE | `/api/requests/{id}` | Elimina richiesta pendente |
+| GET | `/api/sessions` | Lista sessioni attive |
+| DELETE | `/api/sessions/{id}` | Chiudi sessione |
+
+## Gestione errori
+
+Le risposte di errore seguono lo stesso formato envelope:
+
+```json
+{
+  "success": false,
+  "data": null,
+  "error": {
+    "code": "NOT_FOUND",
+    "message": "Topic not found"
+  }
+}
+```
+
+Codici HTTP comuni:
+- **202**: richiesta accettata e accodata
+- **400**: parametri mancanti o invalidi
+- **403**: API key mancante o errata
+- **404**: risorsa non trovata
+- **409**: conflitto (es. nome topic duplicato)
+- **410**: sessione chiusa
+- **429**: rate limit superato
+
+## Rate limiting
+
+Le richieste `POST /api/requests` sono limitate per API key:
+- **Per minuto**: configurabile (default 10)
+- **Per ora**: configurabile (default 100)
+
+Le altre operazioni (lettura, topic CRUD) non hanno limiti.
+
+## Esempio completo di workflow
+
+```
+1. GET  /api/models                     -> scegli il modello
+2. POST /api/topics                     -> crea il contesto
+3. POST /api/requests (session_id=new)  -> primo messaggio
+4. GET  /api/requests/{id}              -> polling fino a completed
+5. POST /api/requests (session_id=...)  -> messaggi successivi
+6. GET  /api/requests/{id}              -> polling risultato
+7. POST /api/topics/{id}/summarize      -> riassunto finale
+8. DELETE /api/sessions/{id}            -> chiudi sessione
+```

src/backend/api/routes/health.py

@@ -1,8 +1,9 @@
 from fastapi import APIRouter, Depends
 
 from src.backend.api.dependencies import verify_api_key
+from src.backend.config import settings
 from src.backend.db.database import get_db
-from src.backend.models.api import ok
+from src.backend.models.api import ok, fail
 from src.backend.services.worker_client import WorkerClient
 
 router = APIRouter(dependencies=[Depends(verify_api_key)])
@@ -26,3 +27,17 @@ async def health():
         "database": db_ok,
         "worker": worker_ok,
     })
+
+
+@router.get("/models", summary="List available models", tags=["system"])
+async def list_models():
+    worker = WorkerClient()
+    result = await worker.models()
+
+    if not result.get("success"):
+        return fail("WORKER_ERROR", result.get("error", "Failed to fetch models"))
+
+    return ok({
+        "default_model": settings.claude_model,
+        "models": result["models"],
+    })

src/backend/db/database.py

@@ -19,6 +19,7 @@ CREATE TABLE IF NOT EXISTS topics (
 CREATE TABLE IF NOT EXISTS requests (
     id TEXT PRIMARY KEY,
     session_id TEXT,
+    sdk_conversation_id TEXT,
     topic_id TEXT NOT NULL REFERENCES topics(id),
     prompt TEXT NOT NULL,
     model TEXT,
@@ -156,7 +157,7 @@ async def create_request(
     now = _now()
 
     if session_id == "new":
-        session_id = _new_id()
+        session_id = None
 
     await db.execute(
         "INSERT INTO requests (id, session_id, topic_id, prompt, model, summarize, status, created_at) "
@@ -246,19 +247,14 @@ async def complete_request(
     db: aiosqlite.Connection,
     request_id: str,
     result: str,
-    session_id: str | None = None,
+    sdk_conversation_id: str | None = None,
 ) -> None:
     now = _now()
-    if session_id:
-        await db.execute(
-            "UPDATE requests SET status = 'completed', result = ?, completed_at = ?, session_id = ? WHERE id = ?",
-            (result, now, session_id, request_id),
-        )
-    else:
-        await db.execute(
-            "UPDATE requests SET status = 'completed', result = ?, completed_at = ? WHERE id = ?",
-            (result, now, request_id),
-        )
+    await db.execute(
+        "UPDATE requests SET status = 'completed', result = ?, completed_at = ?, "
+        "sdk_conversation_id = ?, session_id = COALESCE(session_id, ?) WHERE id = ?",
+        (result, now, sdk_conversation_id, sdk_conversation_id, request_id),
+    )
     await db.commit()
 
 
@@ -297,6 +293,17 @@ async def list_sessions(db: aiosqlite.Connection) -> list[dict]:
     return [dict(row) for row in rows]
 
 
+async def get_session_sdk_id(db: aiosqlite.Connection, session_id: str) -> str | None:
+    cursor = await db.execute(
+        "SELECT sdk_conversation_id FROM requests "
+        "WHERE session_id = ? AND status = 'completed' AND sdk_conversation_id IS NOT NULL "
+        "ORDER BY completed_at DESC LIMIT 1",
+        (session_id,),
+    )
+    row = await cursor.fetchone()
+    return row[0] if row else None
+
+
 async def session_exists(db: aiosqlite.Connection, session_id: str) -> bool:
     cursor = await db.execute(
         "SELECT 1 FROM requests WHERE session_id = ? LIMIT 1", (session_id,)

src/backend/main.py

@@ -4,10 +4,14 @@ from contextlib import asynccontextmanager
 
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import JSONResponse
 from slowapi import _rate_limit_exceeded_handler
 from slowapi.errors import RateLimitExceeded
+from starlette.middleware.base import BaseHTTPMiddleware
+from starlette.requests import Request
 
 from src.backend.api.dependencies import limiter
+from src.backend.config import settings
 from src.backend.api.routes.health import router as health_router
 from src.backend.api.routes.requests import router as requests_router
 from src.backend.api.routes.sessions import router as sessions_router
@@ -35,12 +39,25 @@ async def lifespan(app: FastAPI):
 app = FastAPI(
     title="OpusAgent",
     version="0.1.0",
-    docs_url="/doc",
+    docs_url="/docs",
     redoc_url="/redoc",
     openapi_url="/openapi.json",
     lifespan=lifespan,
 )
 
+PROTECTED_DOCS_PATHS = {"/docs", "/redoc"}
+
+
+class DocsAuthMiddleware(BaseHTTPMiddleware):
+    async def dispatch(self, request: Request, call_next):
+        if request.url.path in PROTECTED_DOCS_PATHS:
+            key = request.headers.get("X-Api-Key") or request.query_params.get("api_key")
+            if key != settings.api_key:
+                return JSONResponse(status_code=403, content={"detail": "API key required"})
+        return await call_next(request)
+
+
+app.add_middleware(DocsAuthMiddleware)
 app.state.limiter = limiter
 app.add_exception_handler(RateLimitExceeded, _rate_limit_exceeded_handler)
 

src/backend/services/queue.py

@@ -9,6 +9,7 @@ from src.backend.db.database import (
     complete_request,
     fail_request,
     append_topic_summary,
+    get_session_sdk_id,
 )
 from src.backend.services.worker_client import WorkerClient
 
@@ -38,6 +39,11 @@ async def process_one(worker: WorkerClient) -> bool:
             return True
 
         model = resolve_model(req["model"], topic["model"])
+
+        # Only resume if there's a previous completed request with an SDK conversation ID
+        resume_id = None
+        if req["session_id"]:
+            resume_id = await get_session_sdk_id(db, req["session_id"])
     finally:
         await db.close()
 
@@ -46,7 +52,7 @@ async def process_one(worker: WorkerClient) -> bool:
             prompt=req["prompt"],
             system_prompt=topic["system_prompt"],
             model=model,
-            session_id=req["session_id"],
+            session_id=resume_id,
         )
     except Exception as e:
         db = await get_db()
@@ -63,7 +69,7 @@ async def process_one(worker: WorkerClient) -> bool:
                 db,
                 req["id"],
                 result["result"],
-                session_id=result.get("session_id"),
+                sdk_conversation_id=result.get("session_id"),
             )
 
             if req["summarize"]:

src/backend/services/worker_client.py

@@ -38,6 +38,11 @@ class WorkerClient:
             )
             return resp.json()
 
+    async def models(self) -> dict:
+        async with httpx.AsyncClient(timeout=30.0) as client:
+            resp = await client.get(f"{self.base_url}/models")
+            return resp.json()
+
     async def health(self) -> bool:
         try:
             async with httpx.AsyncClient(timeout=5.0) as client:

tests/test_api_requests.py

@@ -49,8 +49,8 @@ async def test_create_request_with_new_session(client, topic_id):
     )
     assert resp.status_code == 202
     data = resp.json()
-    assert data["data"]["session_id"] is not None
-    assert data["data"]["session_id"] != "new"
+    # session_id is None at creation; gets set from SDK after completion
+    assert data["data"]["session_id"] is None
 
 
 @pytest.mark.asyncio

tests/test_api_sessions.py

@@ -37,12 +37,12 @@ async def test_list_sessions_empty(client):
 
 @pytest.mark.asyncio
 async def test_list_sessions_with_data(client, topic_id):
-    resp1 = await client.post(
+    # Use a custom session_id (not "new") so it's set immediately
+    session_id = "test-session-abc"
+    await client.post(
         "/api/requests",
-        json={"prompt": "First", "topic_id": topic_id, "session_id": "new"},
+        json={"prompt": "First", "topic_id": topic_id, "session_id": session_id},
     )
-    session_id = resp1.json()["data"]["session_id"]
-
     await client.post(
         "/api/requests",
         json={"prompt": "Second", "topic_id": topic_id, "session_id": session_id},
@@ -58,11 +58,11 @@ async def test_list_sessions_with_data(client, topic_id):
 
 @pytest.mark.asyncio
 async def test_delete_session(client, topic_id):
-    resp1 = await client.post(
+    session_id = "test-session-delete"
+    await client.post(
         "/api/requests",
-        json={"prompt": "Hello", "topic_id": topic_id, "session_id": "new"},
+        json={"prompt": "Hello", "topic_id": topic_id, "session_id": session_id},
     )
-    session_id = resp1.json()["data"]["session_id"]
 
     resp = await client.delete(f"/api/sessions/{session_id}")
     assert resp.status_code == 200

uv.lock

@@ -226,6 +226,12 @@ dev = [
     { name = "pytest-asyncio" },
 ]
 
+[package.dev-dependencies]
+dev = [
+    { name = "pytest" },
+    { name = "pytest-asyncio" },
+]
+
 [package.metadata]
 requires-dist = [
     { name = "aiosqlite", specifier = ">=0.20" },
@@ -241,6 +247,12 @@ requires-dist = [
 ]
 provides-extras = ["dev"]
 
+[package.metadata.requires-dev]
+dev = [
+    { name = "pytest", specifier = ">=9.0.3" },
+    { name = "pytest-asyncio", specifier = ">=1.3.0" },
+]
+
 [[package]]
 name = "packaging"
 version = "26.2"

worker/index.js

@@ -11,6 +11,20 @@ app.get("/health", (_req, res) => {
   res.json({ status: "ok" });
 });
 
+app.get("/models", async (_req, res) => {
+  try {
+    const conv = query({
+      prompt: ".",
+      options: { model: "claude-sonnet-4-6", allowedTools: [] },
+    });
+    const models = await conv.supportedModels();
+    await conv.return();
+    res.json({ success: true, models });
+  } catch (error) {
+    res.json({ success: false, error: error.message });
+  }
+});
+
 app.post("/process", async (req, res) => {
   const { prompt, system_prompt, model, session_id } = req.body;
 
@@ -31,22 +45,28 @@ app.post("/process", async (req, res) => {
       options.resume = session_id;
     }
 
-    const messages = await query({
-      prompt,
-      options,
-    });
+    const conv = query({ prompt, options });
+    const messages = [];
+    try {
+      for await (const msg of conv) {
+        messages.push(msg);
+      }
+    } catch (_) {
+      // SDK throws after process exits; messages already collected
+    }
 
-    const resultText = messages
-      .filter((m) => m.type === "text")
-      .map((m) => m.text)
-      .join("\n");
-
-    const conversationId = messages.length > 0 ? messages[0].conversationId : null;
+    const resultMsg = messages.find((m) => m.type === "result");
+    if (!resultMsg) {
+      return res.json({ success: false, error: "No result received from Claude" });
+    }
+    if (resultMsg.is_error) {
+      return res.json({ success: false, error: resultMsg.result });
+    }
 
     res.json({
       success: true,
-      result: resultText,
-      session_id: conversationId || session_id,
+      result: resultMsg.result || "",
+      session_id: resultMsg.session_id || session_id,
     });
   } catch (error) {
     console.error("Process error:", error.message);
@@ -76,7 +96,7 @@ app.post("/summarize", async (req, res) => {
       : `Summarize this exchange in 1-2 sentences:\n\n${exchangeText}`;
 
   try {
-    const messages = await query({
+    const conv = query({
       prompt: summaryPrompt,
       options: {
         model: DEFAULT_MODEL,
@@ -85,15 +105,19 @@ app.post("/summarize", async (req, res) => {
         allowedTools: [],
       },
     });
+    const msgs = [];
+    try {
+      for await (const msg of conv) {
+        msgs.push(msg);
+      }
+    } catch (_) {
+      // SDK throws after process exits; messages already collected
+    }
 
-    const summaryText = messages
-      .filter((m) => m.type === "text")
-      .map((m) => m.text)
-      .join("\n");
-
+    const resultMsg = msgs.find((m) => m.type === "result");
     res.json({
       success: true,
-      summary: summaryText.trim(),
+      summary: (resultMsg?.result || "").trim(),
     });
   } catch (error) {
     console.error("Summarize error:", error.message);