A modern fullstack app for chatting with LLMs, documents, or a web links.

• Node.js — version 20 or higher
• Ollama — Download here
• LLM — gemma3:4b, or any other model supported by your system specifications
• Embedding Model — dengcao/Qwen3-Embedding-0.6B:Q8_0
• Minimum Free RAM — 8 GB

- Nextjs
- Langchain
- Perplexity API
- Ollama
- Pinecone
- Upstash (Redis)

1. Chatbot — Talk directly with your chosen LLM.
2. Chat with Document — Perform RAG-based queries using your uploaded files.
3. Chat with URL (coming soon) — Interact with web content directly.

• Chat with Perplexity API for up-to-date responses.
• Chat with local LLMs via Ollama (offline).
• Use RAG (Retrieval-Augmented Generation) to chat with your documents.
• Simple UI to switch between:
  • Chat with LLM
  • Chat with Document
  • Chat with URL (coming soon)

• Namespace-based vector separation for files.
• Contextual memory for persistent conversations.
• Middleware to limit API requests and prevent overuse.

PERPLEXITY_API_KEY=""
NEXT_PUBLIC_HOST="http://localhost:3000"
EMBEDDING_MODEL="dengcao/Qwen3-Embedding-0.6B:Q8_0" # Ollama
OLLAMA_BASE_URL="http://localhost:11434"
OLLAMA_MODEL="gemma3:4b"
PINECONE_KEY=""
UPSTASH_REDIS_URL="" # for rate-limiter
UPSTASH_REDIS_TOKEN="" # for rate-limiter

We appriciate contributions! Please follow CONTRIBUTING.md for more infromation.

1. Next.js — for an amazing framework
2. LangChain.js — for simplifying complex LLM integrations and RAG workflows.
3. Ollama — for letting me run LLMs on my half-alive laptop!
4. Perplexity API — for sparking the idea of building an RAG application.