Configure · LLM Settings

LLM Settings

Chat and embeddings providers are configured via environment variables (primary + fallback chain). This view is read-only — editing that chain in the dashboard is planned for the full configuration window (T3.8).

Provider health
Pings each configured provider with a minimal request to confirm it's reachable.
Chat providers (primary → fallback order)
primary — qwen-tailscale / qwen2.5:7b-instruct
https://montu.tail327002.ts.net/v1
configured — run a health check
Embeddings provider
embeddings — gemini / gemini-embedding-001
https://generativelanguage.googleapis.com/v1beta/openai
configured — run a health check

Local LLM override

Local LLM override

Point this firm's live chat replies at a self-hosted model (e.g. Ollama + Qwen via a Tailscale Funnel — see docs/LOCAL-LLM-SETUP.md) instead of the deployment's configured provider. Takes effect on the next chat message, no redeploy needed. Test Cases, lead extraction, and conversation summaries are unaffected — they keep using the primary configured provider, not this override.

Must be a public https:// address (not localhost/LAN) — include the trailing /v1.

Leave blank for a local Ollama endpoint (defaults to ollama). Never pre-filled with — or shared with — your primary provider's key.