Feb 1

Our Kamiwaza community edition 0.1.0 is available by clicking contact above and filling out the request form. Currently only available for Mac's with Metal processors and Linux with Nvidia RTX cards.

‍

‍

March 26

Our new release of the Kamiwaza community edition 0.2.0 is now in broad availability through a simple download!

Support provided through our Discord server - https://discord.gg/cVGBS5rD2U

This full feature release is only limited by a single node non-clustered now including:

• Model repository with hf hub integration
• Model deployment from repository, supporting llama.cpp and vLLM
• Helper middleware (optional) for embeddings, vector databases, retrieval
• Retrieval middleware that stitches together catalog+secret management+retrieval+pipelines on Ray
• Pre-integrated components with an installer for both OSX / Linux
• Support for CPU and GPU on both OSX and Linux, both arm64 and amd64/x86_64 platforms
• Notebook environment pre-integrated with the Kamiwaza tools
• REST API for building applications
• Response React UI (using the REST API) for viewing and managing models, model deployments, catalog items, and more
• Pre-integrated and installer auto-deploys Acryl Datahub (catalog), Milvus as a vectordb, and CockroachDB for managing Kamiwaza items
• Built with Ray (Ray Serve, Ray Data on top of Ray) to allow large parallel pipelines

Kamiwaza is also announcing the open-source release of AgentZero, an internal ChatBot tool built by Kamiwaza as the first leg of a chatbot with RAG and Agent framework built in. This first release supports a fully functional responsive chatbot web interface as a FastAPI app, with:

https://github.com/kamiwaza-ai/agentzero

1. Familiar sidebar history+conversations
2. Support for OpenAI and openai-compatible model endpoints (including Kamiwaza model deployments)
3. Code highlighting
4. Token streaming
5. Model selector (including the ability to change models mid-conversation)
6. Automatic retrieval functions; if you have Kamiwaza community edition installed (not required for AgentZero), your model deployments will auto-populate, as well your data sources for RAG retrieval

‍