alcheme-labs/dsh-experience-map
Experience Map for DeepSeek Harness turns completed work into evidence-governed, reusable Experiences that can be checked, approved, executed, verified, revised and forgotten: it stores typed, versioned components in SQLite, runs a current Preflight before proposing reuse, and asks the user to approve the exact reuse plan before it affects a task.
Experience Map for DeepSeek Harness turns completed work into evidence-governed, reusable Experiences that can be checked, approved, executed, verified, revised and forgotten. Ordinary chat history does not reliably say which steps worked, under which conditions, who approved them, or whether an old result is still valid; Experience Map stores the successful route as typed, versioned components (Procedure, Diagnostic, Strategy, Preference Policy, Fact, Causal), binds each claim and step to exact source references and evidence grades, and runs a current Preflight before proposing reuse. An LLM may propose content, but the Host derives eligibility and evidence limits and persists a Candidate that the user accepts, edits, rejects or publishes field by field. Reuse composes an exact Plan that discloses selected and discarded contributions and only an authorized decision creates a one-time binding, so automatic tool execution stays disabled. SQLite owns the durable records; Markdown, learning views and the relation map are readable projections. The README reports one paired task with 65.9% lower provider token volume, 45.5% fewer tool calls and 46.2% fewer model steps, explicitly a single task family and not a guarantee.
Install
dsh plugin --profile web add dsh-experience-map@0.1.0-beta.6Copied verbatim from the README's install block for a DSH 0.2.0-rc.2 Web profile; the README notes DSH 0.1.5-rc.2 users must explicitly install dsh-experience-map@0.1.0-beta.3, and headless profiles use --profile headless with the same version.
Compatibility
Requires DeepSeek Harness 0.2.0-rc.2 for plugin 0.1.0-beta.6 (use 0.1.0-beta.3 with DSH 0.1.5-rc.2), Node.js ^22.19.0 or >=24.0.0, and a configured Harness LLM provider only when you want a model-proposed Candidate. The npm package does not bundle Transformers.js or model weights; the optional local dense-retrieval adapter is an explicit opt-in peer. Pre-release schema v8 rejects older local databases. English README with a Chinese translation.
Details
- Repo: alcheme-labs/dsh-experience-map
- Category: Memory
- Stars: 1
- Version: GitHub repo alcheme-labs/dsh-experience-map (1★, MIT); npm package dsh-experience-map at 0.1.0-beta.6 on both the latest and beta dist-tags (registry check 2026-10-13; the registry repository field resolves back to this repo).
Alternatives
863683348/dsh-plugin-focus · akslcw/dsh-negative-ledger · aLIZELI/belief-merge