Live Chat Assistant

BellamyBook Assistant answers product questions using your docs-app markdown and OpenAI (default gpt-4o-mini). It replaces the old n8n webhook proxy.

How it works

1. Frontend or docs widget POST /api/live-chat/send with header X-Live-Chat-Token.
2. API validates the token and applies IP + session rate limits.
3. Docs are loaded from LiveChat:DocsPath, chunked, and ranked with a cheap keyword score (synonyms, light typo fix, FAQ pin — no embedding API cost). Public Q&A (founder, GitHub, contact, pricing) lives in FAQ.
4. Top chunks + recent Redis session history are sent to OpenAI Chat Completions.
5. Response shape: { "output": "..." } (widgets also accept message / text).

Configuration

Key	Description
LiveChat__Enabled	Turn the endpoint on/off
LiveChat__Token	Shared secret with frontend / docs widgets
LiveChat__OpenAiApiKey	OpenAI API key
LiveChat__Model	Default gpt-4o-mini
LiveChat__DocsPath	Folder of docs markdown (Docker: /app/docs)
LiveChat__SessionRateLimitPerMinute	Per-session Redis limit
LiveChat__DailyRequestLimitPerIp	Soft daily OpenAI request cap per IP

Also keep AspNetCoreRateLimit rule post:/api/live-chat/* (15/min, 60/hour).

Client env

• Frontend: VITE_SOCIAL_ENABLE_LIVE_CHAT, VITE_LIVE_CHAT_TOKEN
• Docs-app build: LIVE_CHAT_ENABLED, LIVE_CHAT_API_URL, LIVE_CHAT_TOKEN

See Self-host: Live Chat.