Inspiration

As developers, we spend hours writing code, switching between tools, and explaining our work in reports no one reads. We wanted to build something that turns daily developer activity into a searchable, intelligent memory — so your past work can help your future self.

What it does

Synthex automatically captures developer tasks (JIRA, Git commits, IDE activity), collects justifications via a lightweight UI, generates weekly reports, and transforms this data into AI-powered memory. This memory becomes searchable directly inside the IDE, giving developers instant, context-rich answers based on their own work.

How we built it

• Frontend: Lightweight UI for submitting justifications and generating reports.
• Backend:
• CompleteReportExtractor.py – parses reports
• SupabaseManager.py – stores structured data
• KnowledgeChunkCreator.py – creates searchable chunks
• EmbeddingService.py – embeds content using SentenceTransformers
• ChromaDBManager.py – vector DB for search
• HonchoIntegration.py – builds AI memory
• MCP WebSocket Server: Handles IDE queries, runs parallel search across Supabase, ChromaDB, and Honcho.
• VS Code Extension: Connects to MCP server and brings insights to the developer.

Challenges we ran into

• Designing a schema that works across tasks, commits, and reports
• Handling daily report submissions without adding friction
• Syncing embeddings with structured data
• Making vector search results explainable and useful in the IDE

Accomplishments that we're proud of

• End-to-end pipeline from developer activity to IDE-based assistance
• Seamless report generation + approval flow
• Contextual memory that understands each developer’s workflow
• Built everything using open-source tools + local embeddings

What we learned

• How to combine structured and unstructured developer data
• How to build multi-source search pipelines
• How AI can assist not just with code — but with context

What's next for Synthex

• Natural language prompts for weekly report generation
• Admin dashboards for team intelligence
• Multi-developer memory linking for onboarding and collaboration
• Support for more IDEs and teams

Built With

  • api
  • apis
  • axios
  • cloud-services
  • databases
  • eslint
  • flowglad
  • framer-motion
  • frameworks
  • gemini-api
  • github
  • material-ui
  • pexels
  • platforms
  • postcss
  • react
  • react-hook-form
  • react-speech-recognition
  • recharts
  • tailwind-css
  • typescript
  • vite
  • yup
Share this project:

Updates