kyrielrving11/LoopForge

kyrielrving11★ 1TypeScript最后同步: 2026-08-21

在 GitHub 打开

LoopForge is a governance runtime for long-horizon AI coding tasks. It keeps the objective, hard constraints, near-term plan, evidence, approvals, and recovery state outside the coding Agent's conversation.

README 摘要

LoopForge A context window is not memory. Memory needs a runtime. v2.11.0 — npm install -g loopforge . Node.js ≥ 18. Zero runtime dependencies. 中文文档 The real problems with long-horizon tasks Any AI coding agent, given a long enough task, hits three walls. Not because the model isn't smart enough. Not because the context window isn't large enough. These are architectural problems. Problem 1: The summary cascade Context fills up → compress. Fills up again → compress again. The third compression is a summary of a summary of a summary. Every cycle evaporates information — critical constraints get silently deleted because the summarizer judged them "no longer relevant." Researchers found that after 3–4 summary cycles, constraint violation rates spike from 0% to 59%. The agent doesn't know what it forgot, because the forgetting itself was compressed away. This is what happens when you treat cognitive state as context. The bucket always leaks. Problem 2: Self-correction doesn't work Google DeepMind's conclusion is clear: models cannot reliably improve themselves through introspection. They talk themselves out of correct answers. Effective recovery requires an external verifier — an indepe…

在 GitHub 查看完整 README →
Agent/智能体dsh-pluginagentsearch

分类