XNews

Hugging Face Releases ML Intern as Installable Slack Bot

Hugging Face released its open-source ML Intern as a Slack bot that users can install in their own workspaces. The bot evaluates models, builds datasets and fine-tunes models on the fly, running the same agent as HuggingChat under each user's own Hugging Face account.

Event version 1.

Key insights

Installation and channel setup

Users install ML Intern through huggingface.co/chat/slack/install and click Allow. In channels, ML Intern is added via the channel details under Agents & apps.

Per-user account connection

Everyone who uses the bot connects their own Hugging Face account. On first mention, ML Intern replies with a private Connect Hugging Face link that expires after ten minutes.

Billing and publishing

Inference and compute use the user's account and follow HuggingChat billing settings, allowing billing to an organization. ML Intern publishes repositories and Spaces to the user's username unless told otherwise.

Session and file limits

Each thread is its own session with its own conversation, files and compute budget. ML Intern reads up to 200 messages in a thread and accepts up to 10 files per request at 10 MB each and 25 MB total.

Source posts