ModelHub
Docker app from allornothing's Repository
Self-hosted, open 3D print file library manager: STL/3MF/OBJ/STEP/FBX organization, thumbnails, duplicate detection, tagging, semantic search, smart collections, filament…
Requirements
Community Applications first) if you want local AI instead of an API key.
Overview
Readme
View on GitHubModel Hub
A self-hosted manager for your 3D-print file library, built for Unraid. Point it at a folder of STL, 3MF, OBJ, STEP or FBX files and it gives you thumbnails, duplicate detection, tags and a print queue.
Why use it:
- Find things fast. Thumbnails, collections, and search that understands what a model is, not just its filename.
- Optional AI tagging. Use a local Ollama, an OpenAI-compatible API, or turn AI off. Nothing leaves your network unless you pick an API.
- Run your print shelf. Track filament, queue prints across printers, and keep a print log.
Image: allornothing/model-hub · Source: aon082910/model-hub
Install
Unraid: Docker tab → Add Container, then paste this into the Template field:
https://raw.githubusercontent.com/aon082910/model-hub/master/unraid/model-hub.xml
Docker Compose: run docker compose up --build in the repo and open http://localhost:8420.
| Setting | Default | What it's for |
|---|---|---|
| Port | 8420 |
Web UI |
/data |
/mnt/user/models/model-hub-library |
Your model files |
/config |
/mnt/user/appdata/model-hub |
Database and settings |
PUID / PGID |
99 / 100 |
Owner of files it creates |
First run
- Open the UI and create your admin account (or set
AUTH_USERNAMEandAUTH_PASSWORDin the template). - Click Rescan Library.
- Optional: in Settings, pick an AI mode. For local AI, install Ollama from Community Applications and run
ollama pull llavaandollama pull nomic-embed-text, then setOLLAMA_HOSTto it.
Help
Questions or bugs: open an issue.
Categories
Download Statistics
Related apps
Explore more like this
Explore allLinks
Details
allornothing/model-hub:latestRuntime arguments
- Web UI
http://[IP]:[PORT:8420]/- Network
bridge- Shell
bash- Privileged
- false
Template configuration
Port for the Model Hub web interface.
- Target
- 8420
- Default
- 8420
- Value
- 8420
Folder containing your STL/3MF/OBJ/STEP/FBX files. Read-only usage: files are hashed and thumbnailed, never modified.
- Target
- /data
- Default
- /mnt/user/models/model-hub-library
- Value
- /mnt/user/models/model-hub-library
Where the SQLite database, thumbnails, and settings are stored.
- Target
- /config
- Default
- /mnt/user/appdata/model-hub
- Value
- /mnt/user/appdata/model-hub
Unraid's nobody user -- keeps /config editable from the host
- Default
- 99
- Value
- 99
Unraid's users group
- Default
- 100
- Value
- 100
Seconds between automatic background library scans.
- Target
- SCAN_INTERVAL_SECONDS
- Default
- 300
- Value
- 300
Memory ceiling for the separate process that parses model files during scans. Leave blank for automatic (half of the container limit or host RAM, 1-8 GB). Files needing more are still indexed, from a preview (a .3mf's embedded thumbnail, or a streamed render of an STL) instead of a full parse. Raise it only if you have RAM to spare and want full stats on very large files.
- Target
- MESH_WORKER_MEMORY_MB
Address of your Ollama container/instance for local AI tagging. Only used when AI mode is set to 'local' in the app's Settings tab.
- Target
- OLLAMA_HOST
- Default
- http://ollama:11434
- Value
- http://ollama:11434
Vision-capable Ollama model used for auto-tagging thumbnails.
- Target
- OLLAMA_VISION_MODEL
- Default
- llava
- Value
- llava
Ollama embedding model used for semantic search.
- Target
- OLLAMA_EMBED_MODEL
- Default
- nomic-embed-text
- Value
- nomic-embed-text
Creates the admin login automatically on first start. Leave blank to set it up from the WebUI's first-run screen instead.
- Target
- AUTH_USERNAME
Paired with Admin Username above. Leave both blank to use the in-app setup screen instead (recommended -- avoids the password sitting in plain text in the container's env).
- Target
- AUTH_PASSWORD
Optional: absolute path *inside the container* to a headless slicer CLI binary (e.g. an extracted PrusaSlicer/OrcaSlicer AppImage) for exact print time/filament estimates instead of the built-in volumetric approximation. Bind-mount the binary in via 'Add another Path' first, then point this at it.
- Target
- SLICER_CLI_PATH