This project provides a Slack app that fine-tunes a language model (Llama 3.1) to mimic a specific user's communication style based on their Slack messages. It's designed for users who want to automate responses or create a digital persona within Slack, leveraging serverless infrastructure for training and inference.

How It Works

The core of DoppelBot involves scraping a target user's Slack messages, fine-tuning a Llama 3.1 model on this data, and then deploying an inference endpoint. All these operations are managed and executed serverlessly on the Modal platform, abstracting away infrastructure concerns and allowing for scalable, on-demand processing.

Quick Start & Requirements

• Install: modal deploy src.bot
• Prerequisites: Modal account and client (pip install modal), Slack app setup with specific scopes (app_mentions:read, channels:history, channels:join, channels:read, chat:write, chat:write:customize, users.profile:read, users:read), Modal secrets (slack-finetune-secret). Optional: Weights & Biases account, Neon Postgres database for multi-workspace support.
• Setup Time: Training can take "few minutes to an hour" depending on message volume.
• Docs: https://modal.com/docs

Highlighted Details

• Fine-tunes Llama 3.1 for personalized Slack responses.
• Serverless architecture powered entirely by Modal.
• Supports single-workspace and optional multi-workspace distribution.
• Includes optional Weights & Biases integration for tracking fine-tuning runs.

Maintenance & Community

• Developed by Modal Labs.
• Links to Modal documentation and platform.

Licensing & Compatibility

• The README does not explicitly state a license. The project is hosted on GitHub under the modal-labs organization, suggesting potential adherence to Modal's general terms or a permissive license, but this requires verification.

Limitations & Caveats

• Currently limits each workspace to one target user, which cannot be changed post-installation.
• The project's reliance on Modal means users are dependent on the Modal platform's availability and pricing.