Val-sss 5a6803fc30 release: v2.1.0 — ReAct多轮交互循环，Eval 22%→53% (26/50 PASS) ...

新增ReAct循环(kaiwu/agent/react_loop.py)：
- 工具集: read_file/write_file/run_test/grep/list_dir/submit
- 门控: retry_count≥1时触发，小模型5步/中8步/大10步
- 上下文管理: 保留system+初始user+最近6轮，早期压缩
- 安全: 写入前Python语法检查，禁写测试文件，失败恢复

Eval结果 (qwen2.5-coder:32b-instruct-q4_K_M, 50题):
- 总通过率: 26/50 (53%), 对比v2.0.0的8/36(22%)提升141%
- Go题: 6/9 PASS (t42/t43/t44/t46/t56/t57)
- 多文件API: 3/3 全过 (t51/t52/t53)
- 综合题: t10_comprehensive PASS (425s, 之前timeout)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-05-14 07:11:38 +08:00

..

agent

release: v2.1.0 — ReAct多轮交互循环，Eval 22%→53% (26/50 PASS)

2026-05-14 07:11:38 +08:00

ast_engine

release: v1.5.1 — update docs, fix remaining audit issues

2026-05-06 21:11:45 +08:00

audit

feat: v1.7.0 KAIJU架构借鉴 — bounded context + sub-task decomposition + DetailedLogger

2026-05-07 18:24:56 +08:00

builtin_experts

Revert "feat: Blueprint两步生成流程（LLM施工图→模板填充→LLM fallback）"

2026-05-07 09:12:47 +08:00

cli

feat: v1.7.0 KAIJU架构借鉴 — bounded context + sub-task decomposition + DetailedLogger

2026-05-07 18:24:56 +08:00

core

fix: eval-driven bugfixes — 20% → 47% PASS (same-task comparison)

2026-05-12 04:55:06 +08:00

experts

release: v2.1.0 — ReAct多轮交互循环，Eval 22%→53% (26/50 PASS)

2026-05-14 07:11:38 +08:00

flywheel

feat: add detailed attempts list to trajectory records

2026-05-08 09:22:47 +08:00

knowledge

…

llm

fix: eval-driven bugfixes — 20% → 47% PASS (same-task comparison)

2026-05-12 04:55:06 +08:00

mcp

…

memory

…

notification

…

registry

…

scripts

…

search

…

server

release: v1.6.2 版本号统一 + README/CHANGELOG/STATUS更新

2026-05-07 14:42:21 +08:00

stats

…

telemetry

release: v1.6.2 版本号统一 + README/CHANGELOG/STATUS更新

2026-05-07 14:42:21 +08:00

tests

feat: v1.9.0 — 9项架构优化，工程消化反馈替代raw输出注入

2026-05-09 18:38:03 +08:00

tools

feat: 21-fix eval optimization — 0/3 → 3/15 PASS (+83 tests)

2026-05-09 14:03:56 +08:00

tui

…

validation

feat: mechanism_audit.py — 7个最小验证任务走查v1.6.2全部机制

2026-05-07 13:23:09 +08:00

__init__.py

release: v2.0.0 — Eval驱动架构升级，数据量化验证每个改动

2026-05-10 15:59:59 +08:00

__main__.py

…

PRIVACY.md

feat: three flywheels + anonymous telemetry + prompt quantification

2026-05-06 19:39:17 +08:00