StatsRaven

AI

Chat in StatsRaven and the names it gives log patterns come from an AI model you choose. You can also use no AI in StatsRaven at all and ask through Claude over MCP.

Choose a provider

In Settings, AI, pick one:

Test connection checks the key and model, and Load models lists what the provider offers. Settings apply without a restart. The key stays on your server; requests go from there straight to the provider.

The same can come from the environment, for example in /etc/statsraven/statsraven.env. A value saved in Settings wins.

VariableWhat it sets
ANTHROPIC_API_KEYThe Anthropic key
ANTHROPIC_BASE_URLA gateway in front of Anthropic
STATSRAVEN_CHAT_MODELThe chat model
STATSRAVEN_LLM_MODELThe pattern naming model
STATSRAVEN_CHAT_DAILY_TOKENSDaily token limit for chat, 1,000,000 unless set
STATSRAVEN_LLM_DAILY_TOKENSDaily token limit for pattern naming, 200,000 unless set

When a limit is reached, that feature waits until midnight UTC.

Pattern naming

StatsRaven groups log lines into patterns. Every few minutes it asks the naming model to name patterns seen at least three times, so you can ask about "failed checkouts" instead of a raw line. Turn it off in Settings, AI; you can still name patterns yourself or let Claude do it over MCP.

What the provider sees