v0.4.0

What's Changed

• Multimodal RAG for images using multi-modal embeddings for search
• Chat file upload in both backend and React app. Upload files or specify URLs to pass to model.
• Streaming of chat responses in React frontend
• RBAC for model access. See README for LLM Service for docs on how to manage access to models for users.
• Added user role management in React frontend, used for RBAC for models and query engines
• Added documentation on authentication in the platform in docs/AUTH.md
• Updated to recent releases of FastAPI (0.112.2) and associated libraries
• Use L4 GPUs with Truss models
• Chunk size and chunking class are now query engine build params
• Add chunk size to React Query Admin engine build form
• Switched to using llama_index.core.node_parser.SentenceSplitter for chunking by default
• Updated default query generation model to Gemini Flash 1.5 - was set to Palm2
• Added Microsoft login in React frontend
• Add webscraping component written in Go for performance. Used for web data query engines with depth > 0.
• Add timestamps to logging
• New deploy script for React frontend

Fixes

• Fixed download of PDFs from scraped sites
• Fixed deletion of query engines in React frontend; made hard delete the default

Full Changelog: v0.3.2...v0.4.0

v0.3.2

See CHANGELOG for new features and fixes.

Full Changelog: v0.3.1...v0.3.2

v0.3.1

See CHANGELOG for new features and fixes.

Full Changelog: v0.3.0...v0.3.1

v0.3.0

See CHANGELOG for new features and fixes.

Full Changelog: v0.2.0...v0.3.0

GENIE v0.2.0

v0.2.0

Features:

GENIE ("GenAI for Enterprise" platform on GCP)

• Chat, text generation and embeddings using Vertex models (Feb 2024) and Langchain supported LLMs
• Basic RAG pipeline using AlloyDB / PGVector (recommended) or Vertex Matching Engine as vector store. Supports web-based datasources and GCS as a data source.
• Agents using Langchain
• Agent tool examples: send email using gmail, create spreadsheets, consult a rules engine
• DB Agents to retrieve data from SQL databases
• Basic UX in Streamlit and Flutterflow (see READMEs in those components for details)
• Support for Llama2 via Truss (experimental)

Core Solution Services

• Authentication
• User Management
• Rules Engine
• Job Service

Known Issues

• e2e testing not currently enabled
• Some known install issues (see INSTALL guide)
• Flutterflow UX not deploying successfully