Skip to content

v0.1.0 - Local Coding Agent Stack

Choose a tag to compare

View all tags
@timfromhcs timfromhcs released this 20 Sep 18:03
· 17 commits to main since this release
v0.1.0
1b2980f

Local Coding Agent Stack v0.1.0

Autonomous, local-first coding assistant stack powered by native llama.cpp with Vulkan GPU acceleration, an Anthropic-to-OpenAI translation proxy, patched Claude Code CLI, and an automated Recursive Self-Improvement (RSI) continuous learning loop.

Key Features & Milestones

  • Native Inference Engine: llama.cpp with Vulkan compute shader acceleration, Flash Attention (-fa on), 8-bit quantized KV-cache (q8_0), and speculative decoding (ngram-mod).
  • Anthropic API Proxy: Port 4000 bridge translating Anthropic Messages API requests, tool use, and SSE streaming to OpenAI completions.
  • Claude Code CLI Integration: Full TypeScript source build running headless on local models with configurable endpoints and custom small model routing.
  • RSI Feedback Loop: 5-stage daemon (Collect, Curate, Train PEFT LoRA, GGUF Export, 20-task Benchmark, Promotion Gate).
  • Test Suite: 40 unit and integration tests passing with 80% coverage, strictly zero mocks.
  • CI/CD & Containers: Automated multi-platform build workflows, Dockerfiles, and docker-compose.
  • One-Line Installers: POSIX install.sh and Windows native install.ps1.

Full Changelog: https://github.com/timfromhcs/local-coding-agent-stack/commits/v0.1.0

Assets 4
Loading