AI features

OpenFishing has optional AI features powered by a LiteLLM proxy that you run yourself. Your API keys stay server-side — they’re never exposed to the browser.

Enabling the chatbot

Set these environment variables (see Environment variables):

environment:
  - CHATBOT=true
  - LITELLM_URL=http://litellm:4000
  - LITELLM_MODEL=your-model-name

CHATBOT — any truthy value turns on the floating chat widget.
LITELLM_URL — the base URL of your LiteLLM proxy.
LITELLM_MODEL — a model name matching an entry in your litellm.config.yaml.

The chatbot can answer questions about your own data (your lures, spots, catches and tackle).

Running LiteLLM as a sidecar

Run LiteLLM alongside OpenFishing in the same compose project:

services:
  openfishing:
    image: ghcr.io/m1ndgames/openfishing:latest
    ports:
      - "3000:3000"
    environment:
      - CHATBOT=true
      - LITELLM_URL=http://litellm:4000
      - LITELLM_MODEL=your-model-name
      - LITELLM_VISION_MODEL=your-vision-model   # optional
    volumes:
      - openfishing-db:/app/data
      - openfishing-uploads:/app/uploads

  litellm:
    image: ghcr.io/berriai/litellm:main-latest
    command: ["--config", "/app/config.yaml"]
    volumes:
      - ./litellm.config.yaml:/app/config.yaml
    # Provide your provider API keys here, e.g. via an env_file

volumes:
  openfishing-db:
  openfishing-uploads:

Your litellm.config.yaml defines the models (LITELLM_MODEL / LITELLM_VISION_MODEL must match names it declares). Provider API keys are configured on the LiteLLM side and never reach OpenFishing’s frontend.

Photo identification

If you set LITELLM_VISION_MODEL (a vision-capable model), OpenFishing adds:

Fish identification — suggest a species from a photo on the catch form.
Lure identification — recognise a lure from a photo on the add-lure form.

If LITELLM_VISION_MODEL is unset, identification falls back to LITELLM_MODEL.