Darlley / houston

Rocketseat AI platform.

Houston

Rocketseat platform for semantic search and AI chatbot.

Running locally

Setup containers

docker-compose up -d

Setup environment

cp ./packages/api/.env.example ./packages/api/.env
cp ./packages/consumer/.env.example ./packages/consumer/.env
cp ./packages/langchain/.env.example ./packages/langchain/.env

Open each .env file and fill with the desired variables.

Seed sample data

cd ./packages/langchain && pnpm tsx src/sample/seed.ts

The seeder includes data from 252 lessons of React Ignite trail.

Use the API (HTTPie recommended)

http --stream POST localhost:3000/api/messages text="O que é Redux?" title="Sample chat" --auth-type bearer --auth "VALID_SKYLAB_JWT"

V2

• Lesson context
• Semantic Cache
• Better memory

Packages

• packages/contracts: Typings and Zod schemas for all Houston requests/responses;
• packages/api: Houston Web API, which offers HTTP endpoints;
• packages/consumer: Kafka consumer that ingest platform data;
• packages/langchain: Functions to interact with OpenAI;