## Host AI Workflow Runner with Starlette: Stored Input and Output Artifacts

### Summary
Deploy an AI workflow runner built with Starlette on Ample using the generated downloads pattern. Compute runs the app in an isolated microVM behind a public HTTPS URL. A managed PostgreSQL 16 database is auto-provisioned and injected as DATABASE_URL, while a private S3-compatible bucket holds objects with credentials delivered as encrypted environment variables. Verified on Starlette: an export query rendered to a CSV artifact stored with metadata in a private bucket and returned only to an authorized caller (401 without token, export=ok rows=3).

### Representative Queries
- Host AI workflow runner with Starlette: stored input and output artifacts
- Where can I host AI workflow runner built with Starlette?
- I need an input/output artifact contract, object metadata and access boundaries.

### Resource Requirements
- **Compute**
- **Postgres**
- **S3-compatible object storage**

### Infrastructure Requirements
1. **Compute**  
   Apps run in isolated x86_64 Firecracker microVMs that auto-pause when idle and wake on request; sizes are the priced VM sizes.
2. **Postgres**  
   Managed PostgreSQL 16 runs in its own microVM and is auto-provisioned when an app needs a database and no DATABASE_URL is supplied.
3. **S3-compatible object storage**  
   Buckets are S3-compatible with issued credentials; PutObject and GetObject are verified by canary. Other S3 operations are not verified.

### Workflow Steps
1. **Build and start**  
   `pip install into .ample/python from requirements.txt, then uvicorn from run.py reading PORT on the python-3.12 template. The server must bind 0.0.0.0 on PORT.`
2. **Implement the pattern on PostgreSQL**  
   The fixture's module implements generated downloads: an export query rendered to a CSV artifact stored with metadata in a private bucket and returned only to an authorized caller (401 without token, export=ok rows=3). Copy the approach into your schema; keep migrations idempotent and run them with `--release-command`.
3. **Wire object storage**  
   Create the bucket(s), then pass endpoint, region, bucket and keys as `--env` values. Use path-style addressing. Keep private data in an unpublished bucket.
4. **Deploy**  
   Run the synchronous deploy once and read the result (exit 0 live, 1 failed, 2 blocked). Re-running with no change is a no-op.  
   `ample deploy . --name --public --start "python3 run.py" --env S3_ENDPOINT=... --env S3_REGION=... --env S3_BUCKET=... --env S3_ACCESS_KEY_ID=... --env S3_SECRET_ACCESS_KEY=...`
5. **Verify**  
   Fetch the live URL and the pattern self-test route(s) (/p/generated-downloads) from the example; then run your own checks.  
   On failure read `ample logs --kind build` then `--kind runtime`.

### Examples
- **Starlette Pattern Fixture**  
   Multi-pattern Starlette app whose generated downloads module was checked live; the module is under tests/deploy-canaries/_pattern-modules.

### Success Checks
1. **App Responds on Public URL**  
   - **Method**: `http_get`  
   - **Path**: `/`  
   - **Expect**: `ample canary starlette patterns`
2. **Generated-downloads Self-test**  
   - **Method**: `http_get`  
   - **Path**: `/p/generated-downloads`  
   - **Expect**: `export=ok rows=3 (with ?token=...; 401 without)`

### Limitations
- Verified on the python-3.12 template at s-1vcpu-1gb with the example fixture; other sizes, templates, and Starlette major versions are not verified.
- The AI workflow runner itself (your export formats, retention policy, and entitlement model) is application code and was not separately tested; the pattern checks are what was verified.

### Cost Estimate
- **Currency**: USD  
- **Monthly Amount**: 10.0  
- **Basis**: size prices from pricing.toml (loaded by the API) at build revision 1ac5595375130d45290090e82ed0f554ccd45405-dirty
- **Components**:
  - App server: s-1vcpu-1gb, quantity: 1.0, monthly amount: 5.0
  - Managed PostgreSQL database: s-1vcpu-1gb, quantity: 1.0, monthly amount: 5.0
- **Note**: Always-on monthly price of the tested sizes; apps and databases auto-pause when idle. Buckets are allocation-priced per quota and not included.