Architecture

About AIRA

Explore the core systems powering AIRA, including document retrieval, web research, memory, citations, deployment, and LLM-based answer generation.

Online

Backend API

Connected

/backend

Active

LLM Provider

Groq

Primary generation model configured from backend environment.

Ready

Web Search

Tavily

Used when Include web search is enabled in chat.

On

Memory

Enabled

Stores recent conversation context for better follow-up answers.

System Capabilities

Core modules enabled in this AI research workflow.

Documents

PDF, TXT, DOCX, Markdown

Retrieval Mode

Document RAG + Optional Web Search

Vector Store

Local persistent vector index

Security

Environment-based API keys

Retrieval Info

Current RAG defaults.

Chunk Size1000 tokens
Chunk Overlap150 tokens
Retrieval K6 sources
Max Upload Size10 MB

Deployment Notes

Environment variables required for production deployment.

Frontend

NEXT_PUBLIC_API_URL

LLM

GROQ_API_KEY

Web Search

TAVILY_API_KEY