AI / LLM Providers
Several features are AI-assisted and optional: newsfeed article analysis and report generation, email analysis, AI prompt templates, and photo geolocation. None of the rest of the app depends on an LLM being configured — these features simply won’t produce AI output without one.
Supported providers
Section titled “Supported providers”- Anthropic, Google, OpenAI — add the corresponding API key under Settings → API Keys to make that provider’s models available.
- Local Ollama — set
LLM_OLLAMA_BASE_URLto a running Ollama server’s OpenAI-compatible endpoint (e.g.http://host.docker.internal:11434/v1to reach a host-run Ollama from inside the container). No API key needed. Any model already pulled into that Ollama server is auto-discovered and shows up alongside the cloud providers. If the server is unreachable, it’s silently skipped rather than failing startup.
Choosing models
Section titled “Choosing models”Once at least one provider is configured, pick models under Settings → AI Settings: a default model used as a fallback, plus optional per-feature overrides for newsfeed analysis, newsfeed reports, email analysis, AI templates, and photo geolocation.
Multimodal use
Section titled “Multimodal use”Photo geolocation (Image Tools) sends the uploaded image directly to the configured model — pick a model with vision support for that feature to produce useful output.