A gemini based chatbot built using TF-IDF + Cosine Similarity to answer queries from uploaded .txt or .pdf files — ideal for answering university or document-specific questions without relying on cloud storage or external databases.
This chatbot can:
- Read and process text or PDF files uploaded by the user
- Accept natural language queries
- Retrieve the most relevant sentence(s) from the document using TF-IDF vectorization
- Match based on cosine similarity between query and document vectors
- Maintain basic context for multi-turn conversations (semi-stateless)
| Tool/Library | Purpose |
|---|---|
| Python | Core programming language |
| Streamlit | Web app framework for UI |
| scikit-learn | TF-IDF Vectorizer and cosine similarity |
| PyPDF2 | PDF text extraction |
| python-dotenv | Environment variable management (local dev) |
| Gemini API | Can be used for LLM integration in future |
- Upload File (.txt or .pdf)
- Preprocessing: Clean and chunk the text
- Vectorization: Convert document + query into TF-IDF vectors
- Similarity Search: Use cosine similarity to rank and return the most relevant text chunk
- Semi-Context Awareness: Previous queries can be optionally combined to improve relevance
git clone https://github.com/your-username/context-aware-bot.git
cd context-aware-bot
pip install -r requirements.txt
GEMINI_API_KEY=your-api-key-here
streamlit run app.py