SearchMind

Inspiration

The inspiration for SearchMind came from the need to unify productivity, analytics, and personal knowledge management in a single, privacy-respecting platform. We wanted to empower users to not only track their digital life (expenses, notes, tasks, browsing, etc.) but also to gain actionable insights and automate routine tasks using AI.

What it does

SearchMind is a full-stack productivity and analytics platform. It:

• Tracks expenses, meals, notes, tasks, events, transactions, and more.
• Stores and auto-categorizes web and search history using LLMs.
• Provides analytics and pattern discovery on user data.
• Features a chatbot that can answer questions using both user data (via vector search) and generative AI (Gemini), and can take actions like creating tasks from chat.
• Offers a modern, extension-friendly frontend for seamless daily use.

How we built it

• Backend: FastAPI, async Motor (MongoDB), Pydantic models, modular routes for each resource, LangChain+FAISS for vector search, Gemini API for LLM-powered features.
• Frontend: React + TypeScript, Tailwind CSS, modular components for each productivity/analytics feature, browser extension integration.
• AI Integration: Gemini API for RAG, classification, and chat; LangChain for vector search; robust error handling and context management.
• DevOps: MongoDB Atlas, environment-based config, CORS, .env for secrets.

Challenges we ran into

• Ensuring robust, scalable async CRUD for many resource types.
• Handling LLM API quirks (formatting, error handling, quota).
• Mapping between frontend camelCase and backend snake_case data.
• Efficiently embedding and searching large volumes of user data.
• Making the chatbot reliably take actions (like creating tasks) from natural language.

Accomplishments that we're proud of

• Seamless integration of LLM-powered features with user data.
• Modular, scalable backend and frontend architecture.
• Real-time analytics and actionable insights from personal data.
• A chatbot that can both answer and act, not just chat.

What we learned

• The importance of robust data modeling and API design for extensibility.
• How to combine vector search, RAG, and generative AI for practical productivity.
• Best practices for error handling and user feedback in AI-driven apps.
• The value of clear communication between frontend and backend (naming, types).

What's next for SearchMind

• Smarter, more proactive AI suggestions and automations.
• Deeper integrations with third-party services (calendar, email, finance).
• More advanced analytics and visualization features.
• Enhanced privacy controls and on-device AI options.
• Open-sourcing and building a community around extensible productivity AI.

Built With

• apis
• browser
• built-with-**languages:**-python
• classification
• cloud-hosted)-**orm/odm:**-motor-(async-mongodb-driver-for-python)
• cors-middleware
• custom-rest-apis-for-all-resources-**devops/infra:**-.env-for-secrets
• environment-based-config-**other:**-pdfplumber-(pdf-parsing)
• extension
• faiss)-**frontend-tooling:**-vite
• for
• google-cloud-(gemini-api)-**apis:**-gemini-generative-language-api
• javascript-**frameworks:**-fastapi-(backend)
• langchain-(vector-search
• lucide-icons-**cloud-services:**-mongodb-atlas
• pydantic-(data-validation)-**ai/ml:**-google-gemini-api-(llm-for-chat
• pytesseract-(ocr)
• rag)
• react-(frontend)-**databases:**-mongodb-atlas-(nosql
• tailwind-css
• typescript
• web/search