AI Shield Bot Project Overview
1. Core Purpose
- Intercept and analyze `!shield` requests/commands in the #team-passport slack channel
- Retrieve information from multiple sources to either automatically resolve requests or provide actionable steps to on-call engineers
- Provide instant, accurate responses to technical questions using AI-powered knowledge retrieval
AI Shield Bot Project Overview
2. Data Sources
- GitHub repositories
- OpenAPI specification from a specific S3 bucket
- Documentation from docs.immutable.com
- Confluence spaces
AI Shield Bot Project Overview
3. Technical Stack
- Runtime: Node.js
- Language: TypeScript
- Vector Database: ChromaDB for storing and retrieving document embeddings
- AI/LLM: OpenAI (using o3-mini model)
- Embeddings: OpenAI's text-embedding-3-large model
- Web Framework: Express.js for the API server
AI Shield Bot Project Overview
4. Data Flow Diagram
AI Shield Bot Project Overview
5. Security Considerations
- All contextual information is stored locally in ChromaDB
- OpenAI only receives the specific query and relevant context
- No sensitive data is stored in OpenAI
- Enterprise-grade OpenAI account with enhanced security features
AI Shield Bot Project Overview
6. Future State
- Automated Confluence document updates
- Incremental updates to reduce processing costs
- Performance optimization through selective indexing
- Smart caching of frequently accessed information
AI Shield Bot Project Overview
7. Closing Thoughts
- AI-assisted development throughout the project lifecycle
- Requirements gathering and technical approach defined with AI collaboration
- System diagrams and presentations created with AI assistance
- Continuous improvement through AI-powered insights
- Focus on delivering value while maintaining security and efficiency