Skip to main content

v0.1.0

SLMCortex v0.1.0 is the first public product release.

Summary​

This release packages specialized LoRA coding SLMs as self-describing artifacts, composes them into deterministic runtime bundles, and runs local inference, serving, and bounded agent workflows against the same runtime core.

Highlights​

  • canonical public identity: SLMCortex
  • product-first README with one install and demo story
  • architecture docs for SLM Factory, SLM Composer, Runtime Core, and Agent Runtime
  • no-model end-to-end demo using checked-in fixtures and dry-run execution
  • OSS release hygiene including license, changelog, contribution guide, issue templates, pull request template, and CI workflows

Included workflow​

SLM Factory
  -> SLM Package
  -> SLM Composer
  -> Runtime Bundle
  -> Runtime Core
  -> Agent Runtime

Try first​

python3 -m venv .venv
. .venv/bin/activate
pip install -e '.[test]'
python -m slmcortex --help
python scripts/run_slmcortex_demo.py

Limitations​

  • the documented demo flow does not retrain models or download model weights
  • demo inference and agent execution rely on dry-run validation of routing and control flow
  • compose-slms supports only the routed strategy in v0.1
  • the compatibility server is intentionally minimal and non-streaming
  • Agent Runtime is bounded and local, not a general IDE agent