AI Assistant

/ask — an AI assistant that answers questions about the bot and server. Uses RAG (Retrieval-Augmented Generation) to search documentation. Prompt injection protection: 20+ blocking patterns.

Command: /ask question:<your question> Max question length: 500 characters. Response comes as an ephemeral message (only visible to the author).

The bot indexes server documents into a vector database: • Documents split into chunks (800 chars, 100 overlap) • Embeddings created via Ollama (nomic-embed-text) • On question, similar chunks are found (cosine similarity, threshold 0.7) • Found context is added to the prompt Indexing: Dashboard → AI → Index Documents.

Models depend on subscription tier: • Free — Ollama llama3.1:8b (local, 500 tokens) • Pro — Ollama llama3.1:70b (local, 1000 tokens, fallback to 8b) • Business — OpenAI gpt-4o-mini (cloud, 2000 tokens) Anthropic API is also supported as a provider.

Daily request limits: • Free — 10 requests/day • Pro — 50 requests/day • Business — unlimited Cooldown between requests: • Free — 60 seconds • Pro — 10 seconds • Business — 5 seconds

Dashboard → AI: • Enable/disable the module • Index documents • Query and response log • Usage statistics