release: v0.1.45 — code graph index, intent-level API, N-gram optimization

Major additions since v0.1.44:
- Code graph persistent index (graph build/update/refs/impact/map)
- Intent-level commands: find (auto-routing search), edit (exact+fuzzy)
- N-gram index 3x optimization: bincode (12x smaller), memchr, regex trigrams
- 10 ADRs with invariant enforcement
- project-context.md maximized across all pipeline stages
- Memory type classification + auto pitfall capture
- Quick Commands enforcement + repo-map unlimited default
- Tool priority aligned: native first, flowctl for unique capabilities

370 tests pass. Binary: flowctl 0.1.45.

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

Author: z23cc

.claude-plugin/marketplace.json

@@ -6,13 +6,13 @@
   },
   "metadata": {
     "description": "Structured plan-first development plugin for Claude Code with .flow/ task tracking, Agent Teams, and Ralph autonomous mode",
-    "version": "0.1.44"
+    "version": "0.1.45"
   },
   "plugins": [
     {
       "name": "flow-code",
       "description": "Zero-dependency planning + execution with .flow/ task tracking. Three-layer quality (guard + RP plan-review + Codex adversarial). Full-auto, zero questions. Teams auto-parallel, DAG mutation, Codex-driven decisions, auto draft-PR, session summary. 5 prompt templates. 24 subagents, 22 commands, 74 skills.",
-      "version": "0.1.44",
+      "version": "0.1.45",
       "source": "./",
       "category": "workflow",
       "tags": [
@@ -26,5 +26,5 @@
       "strict": true
     }
   ],
-  "version": "0.1.44"
+  "version": "0.1.45"
 }

.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "flow-code",
-  "version": "0.1.44",
+  "version": "0.1.45",
   "description": "Zero-dependency planning + execution with .flow/ task tracking and Ralph autonomous mode (multi-model review gates). Worker subagent per task with git worktree isolation for parallel execution. Three-layer quality system (guard + RP plan-review + Codex adversarial). Full-auto by default — AI decides from context, zero questions. Teams-default with file locking, DAG mutation, Codex-driven conflict resolution, auto draft-PR. Auto-detected stack profiles with one-command guard (test/lint/typecheck). Enhanced agent definitions with permissionMode/maxTurns/effort. Lifecycle hooks with state preservation (PreCompact injects .flow state into compaction, TaskCompleted auto-unlocks files, SubagentStart context injection). Memory v2 with atomic entries, dedup, and progressive disclosure. TDD enforcement mode. Multi-epic queue with dependency visualization. Includes 24 subagents, 22 commands, 74 skills.",
   "author": {
     "name": "z23cc",

.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "flow-code",
-  "version": "0.1.44",
+  "version": "0.1.45",
   "description": "Zero-dependency planning + execution with .flow/ task tracking and Ralph autonomous mode (multi-model review gates). Worker subagent per task with git worktree isolation for parallel execution. Three-layer quality system (guard + RP plan-review + Codex adversarial). Full-auto by default. Teams-default with file locking, DAG mutation, Codex-driven conflict resolution, auto draft-PR.",
   "author": {
     "name": "z23cc",

CHANGELOG.md

@@ -4,6 +4,31 @@ All notable changes to Flow-Code are documented in this file.
 
 Format follows [Keep a Changelog](https://keepachangelog.com/). Versions use [Semantic Versioning](https://semver.org/).
 
+## [0.1.45] - 2026-04-09
+
+### Added
+- **Code graph persistent index** (`flowctl graph build/update/refs/impact/map`) — 1407 symbols, 107K edges, bincode persistence to `.flow/graph.bin`, incremental update via git diff
+- **`flowctl find`** — intent-level search that auto-routes: regex → index regex, symbol → graph refs, literal → trigram, fallback → nucleo fuzzy
+- **`flowctl edit`** — intent-level edit: exact str::replacen first, fuzzy fudiff fallback
+- **`flowctl graph refs`** — find all references to a symbol (<16ms from cache)
+- **`flowctl graph impact`** — transitive impact analysis: what files break if you change this file (BFS depth 3)
+- **N-gram index optimized**: bincode serialization (6.2MB → 502KB, 12x smaller), memchr verification (2-5x faster), regex→trigram extraction for indexed regex search
+- **`flowctl index regex`** — indexed regex search via trigram pre-filtering (12ms)
+- **10 ADRs** with YAML frontmatter (verify + scope) and 5 invariant checks
+- **project-context.md maximized** — now read by all pipeline stages (brainstorm/plan/review/worker/close)
+- **Memory type classification** — `--type pitfall|convention|decision|general`, auto-capture pitfalls on NEEDS_WORK
+- **Quick Commands enforced** — Worker Phase 6 and close phase run epic smoke tests
+- **repo-map default unlimited** — outputs all ranked symbols by default
+- Deep comparison docs: flow-code vs flow-next, fff.nvim lessons, ADR strategy, Rust crate optimization research
+
+### Changed
+- Skills/agents updated to use intent-level API (`find`/`graph refs`/`graph impact`/`edit`) instead of raw tool commands
+- Tool priority: native Claude Code tools (Grep/Glob/Read/Edit) first, flowctl for unique capabilities only
+- Pipeline alignment fixes: checklist wired into Worker Phase 10 + close, memory docs corrected, frecency docs corrected
+
+### Dependencies
+- Added: `bincode` 2, `memchr` 2, `regex-syntax` 0.8
+
 ## [0.1.44] - 2026-04-09
 
 ### Added

README.md

@@ -6,7 +6,7 @@
 
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](../../LICENSE)
 [![Claude Code](https://img.shields.io/badge/Claude_Code-Plugin-blueviolet)](https://claude.ai/code)
-[![Version](https://img.shields.io/badge/Version-0.1.44-green)](https://github.com/z23cc/flow-code/releases)
+[![Version](https://img.shields.io/badge/Version-0.1.45-green)](https://github.com/z23cc/flow-code/releases)
 
 **A production-grade harness for Claude Code. Full-auto development from idea to PR.**
 

README_CN.md

@@ -6,7 +6,7 @@
 
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](../../LICENSE)
 [![Claude Code](https://img.shields.io/badge/Claude_Code-Plugin-blueviolet)](https://claude.ai/code)
-[![Version](https://img.shields.io/badge/Version-0.1.44-green)](https://github.com/z23cc/flow-code/releases)
+[![Version](https://img.shields.io/badge/Version-0.1.45-green)](https://github.com/z23cc/flow-code/releases)
 
 **Claude Code 的生产级开发框架。从想法到 PR，全自动。**
 

bin/flowctl

@@ -0,0 +1,161 @@
+# 深度对比：flow-code vs flow-next (gmickel-claude-marketplace)
+
+> flow-code v0.1.44 vs flow-next v0.29.1 | 2026-04-09
+
+---
+
+## 关系
+
+flow-code 最初 fork 自 flow-next，但已经走上了完全不同的架构路线。两者共享核心理念（plan-first、re-anchoring、.flow/ 状态、零依赖），但在引擎、规模和方向上已大幅分化。
+
+---
+
+## 1. 核心引擎
+
+| 维度 | flow-code | flow-next |
+|------|-----------|-----------|
+| **引擎语言** | **Rust** (31,419 行) | **Python** (7,716 行 flowctl.py) |
+| **二进制** | 编译后单文件 (~3.5MB) | 解释执行 Python 脚本 |
+| **启动速度** | ~5ms | ~200-500ms (Python 启动) |
+| **CLI 命令数** | **68 个** | ~30 个 (估算，Python 子命令) |
+| **并发安全** | fs2 advisory lock + Rust 类型安全 | Python 文件锁 |
+| **DAG 引擎** | petgraph (拓扑排序 + 环检测) | 自建依赖排序 |
+| **状态机** | 8 种任务状态，形式化 | 隐式状态 (JSON 字段) |
+
+### 分析
+flow-code 用 Rust 重写了引擎，代码量是 flow-next 的 4x（31K vs 7.7K 行），但获得了：
+- 编译时类型检查（Python 只有运行时错误）
+- 毫秒级 CLI 响应（vs Python 数百毫秒）
+- 形式化 DAG 验证（petgraph vs 手写排序）
+- 68 个 CLI 命令（vs ~30 个）
+
+flow-next 的优势是**修改简单** — 改 Python 脚本不需要编译。
+
+---
+
+## 2. 规模对比
+
+| 维度 | flow-code | flow-next |
+|------|-----------|-----------|
+| **命令** | 22 个斜杠命令 | 11 个 |
+| **技能** | 73 个 | 16 个 |
+| **Agent** | 24 个 | 20 个 |
+| **flowctl 命令** | 68 个 | ~30 个 |
+| **领域技能** | 47 个（安全/认证/缓存/数据库/API/国际化/容器/微服务/实时/状态管理/错误处理/监控/文档等） | 0 个 |
+| **步骤文件** | 15 个（JIT 加载） | 0 个（整体加载） |
+| **prompt 模板** | 8 个（blind-hunter/edge-case/acceptance-auditor 等） | 0 个 |
+
+### 分析
+flow-code 在 fork 后扩展了 4.6x 技能（73 vs 16）和 2x 命令（22 vs 11）。最大的增量是 47 个**领域专用技能**（flow-next 完全没有）和 15 个**步骤文件**（JIT 加载节省 token）。
+
+---
+
+## 3. flow-code 独有功能
+
+| 功能 | 描述 | 对应 flow-next |
+|------|------|---------------|
+| **Rust 二进制** | 编译时类型安全，毫秒级响应 | Python 脚本 |
+| **68 个 CLI 命令** | 完整的 CLI 工具链 | ~30 个 |
+| **模糊搜索 (nucleo)** | `flowctl search` — 防错别字 + frecency + git 加权 | 无 |
+| **N-gram 索引** | `flowctl index` — trigram 倒排索引，<1ms 搜索 | 无 |
+| **代码结构** | `flowctl code-structure` — 9 语言符号提取 | 无 |
+| **Repo Map** | `flowctl repo-map` — PageRank 排序符号概览 | 无 |
+| **模糊 Patch** | `flowctl patch` — fudiff 3 层回退 | 无 |
+| **project-context.md** | 自动检测技术栈 + Guard Commands + File Conventions + 全管线读取 | 无 |
+| **结构化完成清单** | `flowctl checklist` — 8 项 DoD | 无 |
+| **步骤文件架构** | 15 个步骤文件 JIT 加载，省 60% token | 整体加载 |
+| **三层并行代码审查** | Blind Hunter + Edge Case Hunter + Acceptance Auditor | 单一审查 |
+| **零发现暂停规则** | 审查必须找到问题 | 无 |
+| **高级引导方法** | Pre-mortem/第一性原理/逆向思维 | 无 |
+| **Quick Dev** | `--quick` 标志跳过规划 | 无 |
+| **47 个领域技能** | 安全/认证/数据库/API/缓存/容器/微服务等 | 无 |
+| **8 个 prompt 模板** | 审查/对抗/PR 模板 | 无 |
+| **DAG 环检测** | petgraph 拓扑排序 | 无 |
+| **任务重启级联** | `flowctl restart` 重置下游 | 无 |
+| **架构不变量** | `flowctl invariant add/check` | 无 |
+| **Gap 管理** | `flowctl gap` 追踪缺失需求 | 无 |
+| **事件溯源** | `flowctl events` 全程记录 | 无 |
+| **文件锁 (Teams)** | `flowctl lock/unlock` | 无 (Python 锁不同层级) |
+| **write-file** | 流水线零交互文件写入 | 无 |
+| **doctor 增强** | 9 类健康检查 | 基础检查 |
+| **Frecency 记忆** | 指数衰减文件排序 | 无 |
+| **配置弃用迁移** | 自动检测旧配置 | 无 |
+
+---
+
+## 4. flow-next 独有功能
+
+| 功能 | 描述 | flow-code 状态 |
+|------|------|---------------|
+| **Python 引擎** | 修改简单，无需编译 | Rust 需编译 |
+| **TUI 监控** | flow-next-tui (Bun/TypeScript) 实时 Ralph 监控 | 无 TUI |
+| **Beads 集成** | 原版 flow 插件有 Beads 上下文管理 | 无 |
+| **双插件架构** | flow (legacy) + flow-next (推荐) 共存 | 单插件 |
+| **watch mode** | Ralph `--watch` / `--watch verbose` | Ralph 无 watch |
+| **git add -A** | 总是 git add -A（不选择性 add） | 选择性 git add |
+
+### 分析
+flow-next 的独有优势主要是 **TUI 监控**（Ralph 运行可视化）和**修改无需编译**。其他差异是设计选择（git add -A vs 选择性 add）。
+
+---
+
+## 5. 共享核心理念
+
+两者保留的共同基因：
+
+| 理念 | 实现 |
+|------|------|
+| **Plan-first** | brainstorm → plan → work → review → close |
+| **Re-anchoring** | 每个 Worker 读取 spec + 状态 |
+| **.flow/ 状态** | JSON 文件，无数据库 |
+| **零外部依赖** | 单二进制/脚本 + git/jq/gh |
+| **Multi-model review** | RP + Codex 双后端 |
+| **Ralph 自主模式** | 无人值守循环执行 |
+| **Evidence recording** | 每个任务记录 commits/tests |
+| **Memory system** | 学习 pitfalls，跨会话 |
+| **研究侦察兵** | 并行 scout 收集上下文 |
+| **Prime 评估** | 8 维 48 项就绪评估 |
+
+---
+
+## 6. 架构差异
+
+| 维度 | flow-code | flow-next |
+|------|-----------|-----------|
+| **Epic 阶段** | 6 阶段（brainstorm → close），flowctl 强制顺序 | 隐式阶段，技能自行管理 |
+| **Worker 阶段** | 12 个内部阶段（flowctl worker-phase） | 技能内嵌 phases.md |
+| **技能加载** | 步骤文件 JIT 加载（省 60% token） | 整体加载 |
+| **审查** | 三层并行（Blind + Edge + Acceptance） | 单层（RP 或 Codex） |
+| **Guard** | project-context.md Guard Commands → stack config → 自动检测 | 手动配置 |
+| **Domain 分配** | File Conventions 自动匹配 | 手动 |
+| **搜索** | nucleo 模糊 + N-gram 索引 + frecency | Grep/Glob |
+| **代码理解** | code-structure + repo-map (PageRank) | 无 |
+| **Patch** | fudiff 模糊匹配（3 层回退） | 标准 Edit |
+| **健康检查** | 9 类（binary/git/state/tools/search/context） | 基础 |
+| **跨平台** | Claude Code + Codex sync | Claude Code + Codex + Factory Droid |
+
+---
+
+## 7. 总结
+
+### flow-code 的方向
+> **工程深度** — 用 Rust 重写引擎，加入搜索/索引/代码结构/模糊 patch 等底层能力，扩展 47 个领域技能，project-context.md 贯穿全管线。往"AI agent 的操作系统"方向发展。
+
+### flow-next 的方向
+> **稳定实用** — 保持 Python 引擎简洁（7.7K 行），加 TUI 监控，稳步迭代。往"可靠的 plan-first 工具"方向发展。
+
+### 数字对比
+
+| 指标 | flow-code | flow-next | 倍数 |
+|------|-----------|-----------|------|
+| 技能数 | 73 | 16 | **4.6x** |
+| 命令数 | 22 | 11 | **2x** |
+| CLI 命令 | 68 | ~30 | **2.3x** |
+| 引擎代码 | 31,419 行 Rust | 7,716 行 Python | **4x** |
+| Agent 数 | 24 | 20 | **1.2x** |
+| 领域技能 | 47 | 0 | **∞** |
+| 搜索能力 | 4 种（fuzzy/index/structure/map） | 0 | **∞** |
+
+---
+
+*Generated 2026-04-09 by flow-code*

docs/comparison-flow-code-vs-flow-next.md

@@ -1,6 +1,6 @@
 [package]
 name = "flowctl-cli"
-version = "0.1.44"
+version = "0.1.45"
 description = "CLI entry point for flowctl"
 edition.workspace = true
 rust-version.workspace = true