Toks.love releases ModelHarbor, a self‑hosted model library
ModelHarbor is a self‑hosted software package that lets users serve models from the Hugging Face Hub using only the files they choose. It runs as a Docker container and requires two host directories—one for the model weights and another for the application state. After pulling the provided docker‑compose.yml and setting the MODELSTORAGEPATH environment variable, the service starts on port 7860,…
Key points
- ModelHarbor provides a self‑hosted container to serve Hugging Face models from local storage
- It runs via Docker Compose, needing two host directories and exposing port 7860
- Developed by toks.love with AI‑generated code, released under the MIT License, unaffiliated with Hugging Face
The project is released under the MIT License and is not affiliated with Hugging Face. It builds on the open‑source HuggingHack project, preserving attribution in its documentation. Development was led by the creator of toks.love, who used an AI‑driven coding agent for most of the code, while retaining final review and ownership of the repository.
ModelHarbor offers a straightforward way for teams to keep model weights locally, avoid external API calls, and maintain control over data privacy, especially in environments where internet access is limited or regulated.
Introducing ModelHarbor, a self-hosted model library
github.com · 2 October 2026Loading the full article…
This text was published by github.com. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Reddit discussionreddit.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- Researchers propose Environment Steering to block unsafe AI agent actions · 2 src
- OpenAI adds Ultrafast API tier, Decisions API and security scans at DevDay · 3 src
- Earendil releases Pi 1.0 agent harness with default MCP support · 1 src
- MIT, Google and Northeastern unveil InstructMesh for editing AI‑generated 3D models · 1 src
- AWS Bedrock AgentCore Runtime Instances enables multi-agent music production pipeline · 2 src
Comments
via GitHub Discussions