π Live Demo: https://memorypalai.onrender.com
π¦ Tech Stack: Streamlit β’ LangGraph β’ Gemini 2.5 Flash β’ Pinecone β’ SentenceTransformers β’ PyMuPDF β’ Python
MemoryPalAI is an Agentic AI Learning System that autonomously:
- Processes documents, audio, PDFs, URLs
- Extracts text + style + tags + topics
- Generates embeddings and stores them in Pinecone
- Performs RAG (retrieval augmented generation) for grounded answering
- Generates quizzes based on retrieved content
- Evaluates user answers
- Creates revision notes for weak topics
- Maintains long-term user performance memory
It acts as a complete AI learning companion.
- Supports PDF, TXT, audio (mp3/wav/m4a), URLs
- Automatic text extraction
- Detects subject, style, tone, tags, topic
- Splits text using RecursiveCharacterTextSplitter
- Embeds with Gemini text-embedding-004
- Retrieves top-k relevant chunks from Pinecone
- Answers ONLY using retrieved context
- If unavailable β βI don't know based on the provided documents.β
Generates a 3-phase learning plan from your query and context.
- Creates quizzes from retrieved content
- Fully interactive MCQ interface
- Explains each answer
- Computes final score
- Highlights weak concepts
Triggered when score < 70%:
- Generates concise revision notes
- Suggests YouTube search links
- Provides self-check questions
- Stores revision history per topic
- Builds relationships between topics and tags
- Visualized using custom graph UI
βββββββββββββββββββββββββββββββββββββββ
β User Uploads β
β (PDF / Text / Audio / URL Input) β
βββββββββββββββββββββββββββββββββββββββ
β
βΌ
βββββββββββββββββββββββ
β Ingestion Agent β
β Text Extraction + β
β Chunking + Tagging β
βββββββββββββββββββββββ
β
βΌ
βββββββββββββββββββββββ
β Embedding Service β
β (Gemini Embeddings) β
βββββββββββββββββββββββ
β
βΌ
βββββββββββββββββββββββ
β Pinecone DB β
β Vector + Metadata β
βββββββββββββββββββββββ
β
ββββββββββββ΄βββββββββββ
βΌ βΌ
βββββββββββββββββββββ βββββββββββββββββββββββ
β Retrieve & Answer β β Knowledge Graph β
βββββββββββββββββββββ βββββββββββββββββββββββ
β
βΌ
βββββββββββββββββββββββ
β Quiz Generator β
βββββββββββββββββββββββ
β
βΌ
βββββββββββββββββββββββ
β Quiz Evaluation β
βββββββββββββββββββββββ
β
Score < 70% ?
β
ββββββββββββ΄ββββββββββββ
βΌ βΌ
ββββββββββββββββββββββ ββββββββββββββββββββββ
β Revision Agent β β End of Cycle β
β Weak Topic Notes β β Good Score! β
ββββββββββββββββββββββ ββββββββββββββββββββββ
Frontend
- Streamlit
- Custom UI components
Backend
- Python
- LangGraph Agents
- Gemini 2.5 Flash
- Sentence Transformers
- PyMuPDF
Database
- Pinecone Vector DB
- Local JSON memory
git clone https://github.com/your-username/MemoryPalAI.git
cd MemoryPalAIconda create -n memorypal python=3.11 -y
conda activate memorypalpip install -r requirements.txtGOOGLE_API_KEY=your_key
GOOGLE_API_KEY_2=optional_secondary_key
PINECONE_API_KEY=your_pinecone_key
PINECONE_INDEX_NAME=memorypal-ai
PINECONE_NAMESPACE=memorypal
streamlit run frontend/streamlit_app.pystreamlit run frontend/streamlit_app.py --server.port=$PORT --server.address=0.0.0.0Deployed at:
π https://memorypalai.onrender.com
| Characteristic | Implementation |
|---|---|
| Autonomous | Full ingestion β retrieval β quiz β revision loop |
| Goal Oriented | Learning goalβdriven reasoning |
| Planning | Auto 3-phase learning plan |
| Reasoning | Retrieval-grounded responses |
| Adaptability | Adaptive revision notes |
| Content Awareness | Style, tone, topic, tag extraction |
| Component | Description |
|---|---|
| Brain | Gemini Flash 2.5 |
| Orchestrator | LangGraph |
| Tools | Embedding, Style Detector, Pinecone, Graph |
| Memory | Pinecone + user_profile.json |
| HITL | User uploads, quiz answering |
- Abhishek (AD23B1012) β Architected retrieval pipeline, UI, embedding flow.
- Izhaar Ahmed (23BDS053) β Implemented ingestion, style/tone/tag extraction, Pinecone indexing.
- Jashwanth (AD23B1020) β Built Quiz Agent & Evaluation Agent workflows.
- Swarup G L (AD23B1020) β Developed knowledge graph, session manager, and agent orchestration.
MemoryPalAI is a complete autonomous learning system powered by Agentic AI.
It processes knowledge, quizzes users, evaluates performance, and generates personalized revision notes.
MemoryPalAI is fully open to contributions!
We welcome improvements, bug fixes, new ideas, and feature enhancements.
If you find this project useful, please β star the repository β
your support motivates us and helps the project grow! π
Feel free to submit issues or pull requests anytime.
Weβre excited to see your contributions! β€οΈ