🤖 Multi-Domain AI Personal Assistant

An intelligent, modular, and production-grade AI Assistant built using LangChain, OpenAI, Anthropic, SymPy, Python, and Streamlit.

This project enables users to interact with an AI agent that can:

• 📚 Summarize and QA uploaded documents (PDFs, DOCX)
• 🧠 Solve complex mathematical problems (algebra, calculus, symbolic math)
• 📰 Fetch the latest news from real-world sources
• 🌍 Perform web and Wikipedia searches with detailed responses
• 🐍 Debug, explain, optimize, and generate Python code
• 🖥️ Execute Python snippets live
• 💱 Convert currencies with real-time rates
• 💬 Engage in open-ended, context-aware conversations

---

🚀 Features

🔥 AI Functionalities

• Smart Document QA: Upload any PDF/DOCX, ask questions, get intelligent answers.
• News Summarization: Fetches real-world news and summarizes it like a professional report.
• Web + Wikipedia Search: Queries DuckDuckGo and Wikipedia, compiles coherent answers.
• Advanced Math Solver: Handles symbolic differentiation, integration, equation solving, and standard math.
• Code Engineering Suite:
  • Debug Python code (find and fix errors).
  • Explain Python code (step-by-step breakdown).
  • Optimize Python code (performance, best practices).
  • Generate Python code from instructions.
  • Execute Python snippets live.

---

🛠️ Technologies Used

• Python 3.12+
• Streamlit (UI & Deployment)
• LangChain
• OpenAI API (gpt-4o / gpt-3.5-turbo)
• Anthropic Claude 3 Sonnet (backup LLM for reasoning)
• SymPy (Math engine)
• FAISS (Vector database for document retrieval)
• DuckDuckGo Search API (smart web search)
• NewsAPI / GNews API (real-time news)
• ExchangeRate API (currency conversion)

---

🧩 LangChain Concepts Applied

• Tools and Agents (Zero-Shot-ReAct Framework)
• Memory (Context retention across conversation turns)
• Smart retrievals using Contextual Compression Retriever + LLMChain Extractor
• Custom Runnable chains (RunnablePassthrough, RunnableParallel)
• Output parsing via StrOutputParser
• Error handling and fallback strategies

---

📥 Installation

# 1. Clone the repository
git clone https://github.com/your-username/ai-personal-assistant.git
cd ai-personal-assistant

2. (Optional) Create a virtual environment

python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install all required packages

pip install -r requirements.txt

---

Required Environment Variables

Create a .env file in the root directory with your API keys:

OPENAI_API_KEY=your-openai-api-key
ANTHROPIC_API_KEY=your-anthropic-api-key
GNEWS_API_KEY=your-newsapi-api-key
EXCHANGE_RATE_API_KEY=your-exchangerate-api-key

---

🧠 How It Works

1. Document Upload: Upload multiple PDFs or DOCX files to the sidebar.
2. Question Asking: Ask questions related to the uploaded documents.
3. Tool Routing: LangChain agent intelligently selects the best tool:
4. SmartMathSolver for math
5. PythonDebugger for code
6. SmartWebSearch for real-world info
7. Fallback Handling: If the document lacks an answer, the assistant gracefully transitions to general knowledge.
8. Multi-turn Chat: Context is retained using conversation memory.

---

✨ Future Enhancements

• Add support for image-based document QA (OCR).
• Add user authentication (Streamlit login).
• Deploy a production backend using FastAPI + LangServe.
• Expand toolset to include SQL Query Generation, Visualization Generator, and more.

---

🙌 Acknowledgements

• OpenAI
• LangChain
• Streamlit
• Anthropic
• FAISS
• DuckDuckGo Search API
• NewsAPI.org

---

🛡️ License

This project is licensed under the Apache 2.0 License.

---

👨‍💻 Author

Name: Sayam Kumar
Email: sayamk565@gmail.com
LinkedIn: https://www.linkedin.com/in/sayam-kumar/

Feel free to contact me through email or LinkedIn in case you have any queries about this project.

Built with ❤️, Intelligence, and Curiosity.