v1.0.0 — First Stable

The first public release of Inovus AMA delivers a compact, end-to-end RAG loop: ingest common files, chunk them, create embeddings with Gemini, index in Pinecone, retrieve top-k matches, and synthesize grounded answers with Gemini 2.5 Flash.

A minimal HTTP surface handles asking questions and managing conversation state, while simple ingestion scripts and environment-based configuration make setup straightforward. This cut is aimed at pilots/internal use — optimized for correctness and traceability — so built-in auth/rate-limits and streaming are intentionally out of scope for now.

Features

• End-to-end RAG: ingest → chunk → embed → retrieve → generate
• Gemini 2.5 Flash responses + Gemini Embeddings for search
• Pinecone vector index with configurable retrieval (e.g., top-k)
• Minimal HTTP endpoints for Q&A and health checks
• Conversation state with reset/clear
• Ingestion scripts, environment validation, and basic logging