Sign inSign up

bitdruid/lamahub

By bitdruid

•Updated 29 days ago

Image
0

4.6K

bitdruid/lamahub repository overview

⁠ Lamahub

Mini webapp for managing Ollama LLM server models

Docker Hub

Example

⁠about

Lamahub is a web UI for one or more Ollama⁠ servers.

  • view running and installed models
  • pull, update and delete models
  • pin models to keep them warm-loaded at a fixed context length (in the UI, or via FIXED_MODELS)
  • browse HuggingFace GGUF models and deploy any (sharded) quant straight into Ollama (Deploy tab)
  • chat / generate with any model
  • switch between multiple Ollama endpoints from a dropdown
  • view the application log in a dedicated tab (with level filtering)
  • optional single-user login

⁠run

⁠docker

⁠get image
docker pull bitdruid/lamahub:latest

docker buildx build -t bitdruid/lamahub:latest . --load
⁠compose
docker-compose up -d

⁠script

bash start.sh [options]

Options:

  • -h, --host HOST - Ollama server host (default: localhost)
  • -p, --port PORT - Ollama server port (default: 11434)
  • -b, --base-path PATH - Public base path when hosting under a subdirectory (example: /subdir1/app)
  • -u, --login-user USER - Enable login with this username
  • -w, --login-pw PW - Password for the login user
  • -t, --login-timeout MIN - Session idle timeout in minutes (default: 60)
  • --help - Show help message

⁠cli

Requires Python >= 3.13.

pipx install .
# or: uv tool install .
lamahub

Options:

  • --host HOST - Interface to bind the UI server to (default: 0.0.0.0)
  • --port PORT - Port for the UI server (default: 5000)
  • --ollama-url URL - Ollama server URL(s), same format as the OLLAMA_URL env variable
  • --base-path PATH - Public base path when hosting under a subdirectory (example: /subdir1/app)
  • --login-user USER - Enable login with this username (requires --login-pw)
  • --login-pw PW - Password for the login user (requires --login-user)
  • --login-timeout MIN - Session idle timeout in minutes (default: 60)
  • --reload - Auto-reload on code changes (development only)

⁠envs

Table for envs:

VariableDefaultDescription
OLLAMA_URLhttp://localhost:11434Ollama server URL, or a comma-separated list of Name=URL endpoints for a dropdown switcher (example: Dev=http://dev.ollama,Prod=http://prod.ollama). The first entry is the default and manages FIXED_MODELS.
BASE_PATH(empty)Public base path when hosting under a subdirectory (example: /subdir1/app)
FIXED_MODELS(empty)Comma-separated models pulled automatically when missing (example: llama3.2,mistral). Append @<num_ctx> to bake that context length into the model as its default — applied to every client that doesn't send its own num_ctx, overriding the server's OLLAMA_CONTEXT_LENGTH, and kept warm-loaded (example: glm4:9b@96000). Values above the model's native max are clamped; unpinning reverts it.
HF_TOKEN(empty)Optional HuggingFace token for the Deploy tab. Raises the anonymous API/download rate limits and unlocks gated repos.
HF_STAGING_MAX_GB100Hard cap (GB) on the local shard-download cache used by the Deploy tab. When a new deploy would exceed it, least-recently-used completed families are pruned to fit (0 = unlimited). Staged shards live under INSTANCE_DIR.
LOG_LEVELDEBUGDEBUG, INFO, WARNING, ERROR, or CRITICAL. DEBUG includes periodic fetches; INFO shows only actions.
INSTANCE_DIRinstance/ inside the packageDirectory for instance data like the log file
LOGINfalseSet to true to require login with LOGIN_USER / LOGIN_PW. Sessions are in-memory only.
LOGIN_USER(empty)Username for the login
LOGIN_PW(empty)Password for the login
LOGIN_TIMEOUT60Session idle timeout in minutes

⁠volumes

Table for volumes:

VolumeDescription
INSTANCE_DIR (e.g. /data)Instance data: Logs, pinned model stats, hf model downloads.

Disclaimer: fully agentic project — built entirely by AI against the druidforms⁠ design framework (see AGENTS.md⁠).

Tag summary

Content type

Image

Digest

sha256:0157f7d95…

Size

135.7 MB

Last updated

29 days ago

docker pull bitdruid/lamahub