ai-agent-smart-assist

LangChain AI Agent — Your Smart Assistant 🚀

Hey there! This project is AI agent powered by LangChain. It's built to help with:

• 🧠 Classifying text and routing it to the right logic
• 📚 Adding files to a searchable knowledge base
• 🤖 Asking questions and getting answers from what I’ve uploaded

No rocket science — just upload stuff, click buttons, and the agent does the rest.

---

🔍 What can it do?

1. Classify

If I upload a doc or paste some text, it figures out what it's about — like:

• “Is this support-related?”
• “Should I extract questions from this?”

Then it calls the right tool automatically. Simple.

---

2. Ingest Knowledge

If I want to teach the system something (like feeding it docs), I just:

• Switch to Ingest mode
• Upload my files
• It breaks them into chunks and stores them smartly (FAISS)

Now I can query those chunks later. Boom.

---

3. Ask Questions (Query)

Once I’ve ingested stuff, I can:

• Ask it a question in normal language
• It finds the most relevant chunks
• Then gives me a smart answer (RAG-style)

---

🖥️ What’s in the UI?

• Upload files (multi-file supported)
• Toggle between Classify and Ingest mode
• Run pipelines
• Ask questions from my ingested data
• View full results and even export them

---

🛠️ How I run it

Backend (FastAPI)

cd langchain_ai_agent
uvicorn langchain_ai_agent.api.main:app --reload

Frontend (Next.js)

cd frontend
npm install
npm run dev

---

🧠 Tech I'm using

• LangChain + tools
• Sentence Transformers (MiniLM)
• FAISS vector store
• Unstructured for parsing docs
• FastAPI for serving
• Next.js + Tailwind for the frontend
• Gemini model from Vertex AI

---

🪜 What I’ll do next

I will next:

• Add RAG-style chat flow (so it answers naturally, not just chunks)
• Add support for namespaces (multi-project KBs)
• Maybe connect to Google Drive or Slack

---