Skip to content

KKenny0/Lode

Repository files navigation

Lode

Lode

Agentic coding's persistent memory: capture the why, then compound it into reports, reviews, and decision roadmaps.

Documentation · 中文

Why

AI makes software exploration cheap. It also makes continuity expensive.

In one coding session, you can compare multiple designs, reject plausible paths, discover hidden constraints, update prompts, change schemas, and make decisions that only make sense because of everything you tried along the way. Git records the final diff. Issue trackers record planned work. Neither keeps the reasoning that made the work possible.

When the session ends, that context usually disappears.

Lode is a local habit toolbox for agentic coding's persistent memory. It captures decisions, risks, abandoned paths, open questions, and architecture artifacts, then compounds them into weekly outlines, monthly reviews, and decision roadmaps. The differentiator is not another memory layer; it is structured decision replay that tells the story of what happened and why.

Habit Loop

Lode is organized around the loop you want to repeat until it becomes automatic:

开工 -> 实现探索 -> 收工 -> 周期复盘

Weekly, monthly, and roadmap outputs are compounding layers on top of that loop. Adaptive-depth recap, lightweight sync suggestions, hard-stuff signals, and candidate rules are absorbed into the surviving skills instead of living as separate triggers.

Skills

Each skill maps to a habit you already have. In plugin form, activate it with a namespaced command.

Skill When What it does
/lode:cold-start-interview First run Creates ~/.lode/config.yaml with vault path, project identity, language, and report preferences
/lode:capture Every session wrap-up Classifies the session archetype, captures decision/build/repair depth, and indexes durable artifacts when relevant
/lode:recall Session start Recalls recent decisions, risks, open questions, abandoned alternatives, relevant docs, and possible stale intent artifacts
/lode:daily Daily, on demand Updates Obsidian daily notes from raw entries and git history
/lode:weekly Weekly, on demand Builds a weekly outline from raw entries, with a conditional hard-stuff section when evidence exists
/lode:monthly Monthly, on demand Generates a monthly review and candidate rules from repeated evidence
/lode:roadmap On demand Generates a narrative decision roadmap, including accumulating risks and recurring open questions

Skills are independent. Lode is not a strict pipeline — each skill works on its own, but they share one local storage convention so downstream reports can reuse earlier context.

Install

# Codex Git-backed marketplace
codex plugin marketplace add KKenny0/Lode

# Or local development marketplace
codex plugin marketplace add ./path/to/Lode

# CLI verification remains available
npx @lode/cli doctor

Then run /lode:cold-start-interview once. Work in any git repo as usual: say 开工 or /lode:recall at the start of a session, 收工 or /lode:capture at the end, and /lode:roadmap when you want to see how your project's decisions evolved.

No vault? No problem — 收工 outputs structured Markdown directly in the conversation.

For configuration details see docs/configuration.md. For the data model see docs/data-model.md. Artifact ownership and roadmap storage rules are documented in docs/artifact-governance.md. Synthetic examples live in examples/.

Background

The name comes from lode: a vein of ore where valuable mineral is concentrated. Commits, sessions, diffs, code changes — that is the ore. Lode refines it into work knowledge worth keeping. The word shares its root with load and with the Chinese character 载 (zài, to carry, to record). The original tagline, "the knowledge vein in your codebase", still describes the metaphor; the product promise is persistent memory for agentic coding that humans can read, review, and share.

Lode writes local Markdown and JSON only. Code-adjacent artifacts such as architecture docs stay in the project repo by default; shared memory, indexes, and reports live in your knowledge vault. No remote services, accounts, sync backends, or hosted databases. If your knowledge vault is a git repo, you control where it gets pushed.

Development
npm --prefix cli run build
npm --prefix cli run copy-skills
npm --prefix cli run check-skills

Design principles:

  • Self-contained skills: each skill carries its own references so it can be installed individually.
  • Raw-first reporting: weekly reports use raw entries as the primary semantic source; git is fallback and coverage evidence.
  • Adaptive-depth recap: session wrap-up entries carry archetype-specific fields so reports can explain decisions, repairs, investigations, and builds without a second write skill.
  • Artifact governance: full repo-local docs stay near the code, while recap-owned vault indexes make them discoverable for recall and reports.
  • Graceful side effects: when a raw write is only a side effect, failures do not block the primary deliverable.
  • Deterministic helpers: scripts handle path resolution, date calculation, parsing, and aggregation where consistency matters.
  • Local evals, public protocols: local fixtures stay ignored; public benchmark guidance lives under benchmarks/.
Benchmarks

Public benchmark protocols document the quality bar without publishing local fixtures:

License

MIT

About

⛏️ Capture the context git leaves behind. — 代码库中的知识矿脉

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors