Online self-improvement layer for pi: unified in-trajectory harness-state store + manual /refine with evidence-backed structured CRUD deltas. Composes with pi-reflect/pi-mem.
-
Updated
Aug 10, 2026 - TypeScript
#
continual-harness
Here are 3 public repositories matching this topic...
DeepSeek Harness (dsh) plugin — /refine command + settings panel for the dsh-continual-harness self-refinement engine. npm: dsh-refine
plugin agent npm-package refinement self-improvement harness dsh cordis llm deepseek agent-harness deepseek-harness dsh-plugin continual-harness cordis-plugin
-
Updated
Aug 21, 2026 - JavaScript
Dedicated-model proposer companion for pi-continual-harness. Registers the 'model' /refine proposer: a one-shot LLM call (via the harness's injected complete closure) producing strict JSON deltas validated against current state, with audited spend.
-
Updated
Aug 9, 2026 - TypeScript
Add this topic to your repo
To associate your repository with the continual-harness topic, visit your repo's landing page and select "manage topics."