DocChat AI — Multi-Tenant RAG Knowledge Base Chatbot for SaaS

DocChat AI is a multi-tenant SaaS application where each workspace uploads its knowledge base — PDFs, Word documents, HTML pages, and URLs — and instantly gets an AI support assistant that answers questions using only that content. Documents are processed through an asynchronous ingestion pipeline (extraction, semantic chunking, embedding generation) and stored in a Postgres vector index with strict tenant isolation enforced at the database level. At query time, the system combines vector similarity search with keyword matching (hybrid retrieval) to find the most relevant passages, then streams a token-by-token answer with inline source citations so users can verify every claim. Built for reliability and cost efficiency with queued background processing, embedding caching, incremental re-indexing, and a provider-agnostic LLM layer supporting OpenAI and Anthropic. Tech stack Backend: Laravel 12, PHP 8.4, Laravel Horizon (queued ingestion pipeline), Prism PHP (multi-provider LLM layer) AI/LLM: OpenAI GPT-4o + text-embedding-3, Anthropic Claude (switchable), streaming via SSE Vector search: PostgreSQL + pgvector (HNSW index), hybrid vector + keyword retrieval Frontend: Next.js 15, React, Tailwind CSS, Vercel AI SDK (streaming chat UI) Infrastructure: Redis (cache + queues), Docker, MySQL/Postgres, S3-compatible storage