synapticai_backend

SynapticAI is a backend service designed to power a personal productivity and learning assistant. It enables users to interact with web content, manage notes and tasks, and leverage AI capabilities for summarization, Q&A, and knowledge recall. The system features a robust API built with Hono, integrates with various AI models and services, and utilizes a PostgreSQL database with Drizzle ORM.

Core Features

• AI-Powered Content Interaction: Summarize websites and YouTube videos, ask questions about online content using voice or text.
• Knowledge Management: Create, manage, and retrieve notes. Notes are vectorized for semantic search, enabling users to "chat with their knowledge base."
• Task Management: Create and manage a TO-DO list with AI assistance.
• Voice Interaction: Supports real-time Speech-to-Text (STT) and Text-to-Speech (TTS) for hands-free operation via WebSocket.
• User Authentication: Secure user management using Clerk.
• Data Persistence: Notes, tasks, and user data are stored in a PostgreSQL database, managed with Drizzle ORM.
• Semantic Search: Find relevant notes based on natural language queries using vector embeddings.
• Background Processing: Uses Redis for asynchronous job queuing (e.g., generating note embeddings).
• Agentic AI: Employs Langchain/Langgraph to create AI agents capable of using various tools (e.g., web search, note retrieval, task creation).

Tech Stack

• Runtime: Bun
• Web Framework: Hono
• Database: PostgreSQL (specifically Neon serverless)
• ORM: Drizzle ORM
• Authentication: Clerk
• AI/LLM Orchestration: Langchain, Langgraph
• AI Models: OpenAI (GPT-4o-mini, GPT-4.1-nano, text-embedding-3-small)
• RAG/Vector Search: LlamaIndex, pgvector extension for PostgreSQL
• Speech-to-Text (STT): Deepgram
• Text-to-Speech (TTS): ElevenLabs
• Caching & Job Queue: Redis (ioredis)
• Web Scraping: Puppeteer, Cheerio
• Language: TypeScript

Project Structure

The src/ directory is organized as follows:

• config/: Environment variable loading.
• controllers/: HTTP request handlers and WebSocket controllers.
• db/: Database schema (schema.ts, vectordb_schema.ts), migrations, and Drizzle ORM setup.
• middlewares/: Custom middleware for authentication, session management, and usage metrics.
• repositories/: Data access layer for interacting with the database.
• routes/: API endpoint definitions using Hono.
• services/: Core business logic, AI agent implementation (agentGraph.ts, agentTools.ts), integrations with external services (STT, TTS, AI Models), and Redis utilities.
• utils/: Utility functions, custom error classes, API request/response models, and constants.