Aether

Elevator Pitch: An automated competitive intelligence platform that researches your market and competitors, turns it into a growth strategy, and keeps updating itself.

The Story

What it does

Aether runs a pipeline that researches a business's market and competitors, turns that research into a growth strategy, and reruns itself every 24 hours to catch new competitor moves, pricing changes, or market shifts. The whole process is visible on screen — each step shows up live as it finishes, instead of a loading spinner.

How we built it

The pipeline is split into single-purpose steps — market research, competitor research, strategy generation, change detection — each with a fixed input and output so they hand off to each other cleanly. The frontend (React, Vite, ReactFlow) shows the pipeline as a live node graph, fed by a NestJS backend over Server-Sent Events. A Python worker layer runs the actual research and generation steps using Gemini and Tavily, and everything is stored in Postgres with pgvector for embeddings. The database was built multi-tenant from the start, with a persona table already in place for the next round of features.

Challenges we faced

Getting four separate steps to hand off reliably to each other, with the UI staying in sync in real time, took a fair amount of debugging — keeping the SSE stream stable, making sure every step returned clean structured output, and getting the scheduler and manual "run now" trigger to share state without stepping on each other.

What we're proud of

It works end to end — trigger a run and watch real research turn into a real strategy, live. The live pipeline view, the multi-tenant schema, and the consistent step-by-step design are the parts we'd point to first.

Built With

• docker
• fastapi
• gemini-api
• multi-agent-pipeline
• nestjs
• openai
• postgress
• prisma
• python
• react
• reactflow
• restapi
• shadcn
• tailwindcss
• tavily
• typescript
• vector-embeddings
• vite