Practical maintenance forks, interoperability fixes, and reproducible packages for self-hosted infrastructure and local AI.
Designed for agents, by agents.
- A2A - standards-compliant Agent2Agent provider framework built on the official A2A 1.0 SDK, with dedicated Codex and Antigravity plugins
- Bridgette - production-ready access to 185 tools across 13 Nextcloud application surfaces
- Ava - upstream-connected voice-agent fork with a ready-to-use Docker deployment kit
- Cara - local-first desktop agent workspace for self-hosted models, MCP tools, and private knowledge
- llama.cpp - narrow maintenance fork carrying qualified AMD gfx1151 corrections for LocalAI and LM Studio backends
- llm_proxy - compact Ollama and OpenAI-compatible proxy for self-hosted inference
- Bridgette container
- Ava Docker deployment kits
- Cara Windows installers
- llm_proxy container and binary releases
Each package page lists its immutable image tags, supported architectures, and pull command. Release pages provide checksums and installation assets where applicable. Source, release notes, and deployment guidance remain in each linked repository.



