Advanced Configuration

Control how Raasie searches your knowledge base:

- Top-K Results — How many document chunks to retrieve per query (default: 5). Higher values provide more context but may include less relevant results. - Similarity Threshold — Minimum relevance score for a chunk to be included. Increase this to filter out marginally relevant results.

Raasie uses hybrid search combining two methods:

- Vector Search — Semantic similarity using 512-dimension embeddings. Finds conceptually related content even if the exact words differ. - Full-Text Search — Traditional keyword matching for exact terms. - Reciprocal Rank Fusion (RRF) — Combines both result sets for the best of both worlds.

This ensures the widget catches both semantically similar content and exact keyword matches.

Fine-tune how the LLM generates responses:

- Temperature — Controls response randomness. Lower values (0.1-0.3) produce more focused, deterministic answers. Higher values (0.7-1.0) allow more creative responses. - Max Tokens — Limits response length. Adjust based on whether your use case requires brief or detailed answers.

Each website gets unique API keys for widget authentication:

- Keys are hashed and stored securely — only the prefix is visible after creation - Generate multiple keys for different environments (staging, production) - Rotate keys by generating a new one and revoking the old one - Revoked keys stop working immediately

Rate limits are enforced per plan at the gateway level:

- Free — 500 messages/day/website - Starter — 1,000 messages/day/website - Professional — 5,000 messages/day/website - Enterprise — Unlimited

When limits are reached, the widget displays a friendly message asking visitors to try again later. Limits reset daily.