Python ViewComfy API Integration Rules: Best Practices for FastAPI ComfyUI Workflows

Follow these Do/Don't guidelines to integrate ViewComfy API—a serverless FastAPI service for ComfyUI workflows—into Python applications effectively.

**Do deploy your ComfyUI workflow first**
Upload workflow_api.json to the ViewComfy dashboard to get your unique endpoint.
Example:
```python
# Use dashboard to deploy, then set:
view_comfy_api_url = "https://your-endpoint.viewcomfy.com/api/infer"
```

**Don't call the API without API keys**
Always obtain and use client_id and client_secret from the dashboard.
Example:
```python
client_id = "your_client_id"
client_secret = "your_client_secret"
headers = {"client_id": client_id, "client_secret": client_secret}
```

**Do extract and flatten workflow parameters**
Run workflow_parameters_maker.py on workflow_api.json to identify editable inputs like prompts or images.
Example:
```bash
python workflow_parameters_maker.py --workflow_api_path "workflow_api.json"
```
Result: `{"6-inputs-text": "new prompt", "52-inputs-image": "path/to/image"}`

**Don't send empty params**
Include at least a seed change to avoid errors; update relevant keys for custom inputs.
Example:
```python
params = {"seed": 123456}  # Minimum
params["6-inputs-text"] = "Custom prompt here"
params["52-inputs-image"] = open("image.png", "rb")
```

**Do account for cold starts and variable times**
First calls may delay due to startup (up to minutes); use timeouts like 2400s.
Example:
```python
import httpx
response = await client.post(url, timeout=httpx.Timeout(2400.0), ...)
```

**Do choose infer or infer_with_logs**
Use `infer` for simple POST; `infer_with_logs` for real-time ComfyUI progress via callback.
Example:
```python
def log_callback(message): print(message)
result = await client.infer_with_logs(params=params, log_callback=log_callback)
```

**Do parse responses into structured objects**
Handle JSON with prompt_id, status, outputs (base64 files); save files locally.
Example:
```python
if result.completed:
    for output in result.outputs:
        with open(output.filename, "wb") as f:
            f.write(base64.b64decode(output.data))
```

**Don't override workflows without node compatibility**
Stick to deployed nodes or ensure new workflow_api.json matches installed packs to avoid startup issues.