v1.0.0

feat: init 支持 Codex CLI — AGENTS.md + RTK.md 规则文件

- Agent::Codex 枚举 + init_codex() 实现
- 全局模式: $CODEX_HOME → ~/.codex/，绝对路径 @ 引用
- 项目模式: 当前目录 AGENTS.md + RTK.md
- 幂等写入，保留已有 AGENTS.md 内容
- 6 个单元测试覆盖（解析、项目/全局、幂等、路径、保留）
- .gitignore 排除 .agents/ .codex/ AGENTS.md 运行时产物
- specs/index.md 补充 0019 Codex CLI Support 条目

ONNX Model Files v1

all-MiniLM-L6-v2 ONNX model and tokenizer for tokenless semantic-aware compression (Level 2).

Files:

• all-MiniLM-L6-v2.onnx (~86MB) — FP32 ONNX model
• tokenizer.json (~455KB) — HuggingFace tokenizer configuration

Usage: Place both files in ~/.tokenless/models/ or run make models-install.

v0.3.1

chore: bump workspace version to 0.3.1

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

v0.3.0

docs: update README and indexes for new features

- README: add format router, differential response, predictive cache,
  MCP server to strategy list and token savings table
- README: update architecture tree with new modules (cache, mcp,
  encoding/, shape_analyzer, format_router)
- README: add CLI usage for compress-auto, mcp start, hook diff
- README.zh.md: sync all changes to Chinese version
- docs/index.md: add specs 0011-0014
- specs/index.md: add specs 0011-0014
- specs/0010: update innovation roadmap with status markers

v0.2.0

chore: v0.2.0

v0.1.1

fix: release workflow - don't block GitHub Release on crates-io