This repository is a minimal template for implementing a custom Celery task that runs inside the toxindex.com workflow engine. Use it to adapt your own logic, tools, or model calls and submit a pull request to contribute your task.

• A starter Celery task in task_template_v1.py demonstrating:
  • Receiving a payload with file_id and user_query field
  • Emitting progress updates visible in the toxindex UI
  • Sending a chat-style message back to the task
  • Creating and uploading a Markdown result file to GCS and emitting it to the frontend

• task_template_v1.py: Replace the placeholder logic with your implementation. The Celery task is defined as toolname and is queued on queue='toolname'. Rename this and the queue to match your tool name.

1. Rename the task and queue

   • Change @celery.task(bind=True, queue='toolname') and the function name toolname to your tool name.

2. Wire your logic

   • Replace the placeholder call yourtool_function(user_query) with your function or integration.
   • The input text comes from payload.get("payload"). If your task needs files, see how the sample fetches a File record and downloads it from GCS.

3. Emit output to the user

   • Use the provided helpers already imported in the template:
     • Status updates: emit_status(task_id, "your message")
     • Chat message: build a MessageSchema and call emit_task_message(...)
     • File output: write a temporary file, upload with GCSFileStorage().upload_file(...), then emit_task_file(...)

4. Return and finish

   • Mark completion via Task.mark_finished(task_id) as shown. Keep the try/except so failures correctly surface to the platform.

While the full toxindex stack provides the Celery app, models, and storage, you can still validate your core logic locally:

• Extract your core function (the part you would put into yourtool_function) into a plain Python function and test it with inputs.
• Keep the platform-specific calls (emit, storage, models) inside the task wrapper so your logic remains testable without the platform.

1. Fork or clone your repository containing your implementation.
2. Copy your task code into this template structure, ensuring task_template_v1.py contains your adapted task.
3. Open a Pull Request against our template repository with a clear description of your task and any runtime needs.

• Minimal required payload fields: task_id, user_id. Your input text is expected in payload.
• If your task produces a file, prefer Markdown for easy in-UI preview; set content_type='text/markdown' when uploading.
• Keep logs and status messages concise and user-friendly.

Currently available built-in integrations on the platform include OpenAI, Google Gemini, and Google Search. If you need additional API keys/integrations, reach out to us; we’re expanding support and will add self-service API key registration soon.