OpenRouter API key Get a key

Model

Retrieval Augmented Generation

Response mode

Changing the response mode clears the current conversation so the baseline and RAG histories stay comparable.

Supabase project URL

Supabase publishable key

Retrieved chunks 4

Similarity threshold 0.55

Philosophy Frames

Temperature 0.8

Top-p 1.00

This model locks temperature and top-p to the default (1.0). Pick an older model to experiment with sampling.

Show token probabilities

Token probabilities are unavailable for this model/provider. The reply will render as normal markdown.

Token probabilities are only available in Standard mode. RAG and Compare use full-answer responses so the retrieved citations stay coherent.

Max tokens 1024

Xenophon

Reflective RAG assistant with an inspectable agent pipeline

Run Standard, RAG, Agent, or Compare mode to inspect retrieval quality, cited sources, token usage, and grounding checks.