OverDustD7/dsh-experience-recall

dsh-experience-recall watches what the model is reasoning, calling and failing at, and when the running work matches a lesson your memory store already learned in another project it injects that one conclusion as a short message right before the next step - entirely locally.

dsh-experience-recall hands your agent back what it learned in other projects at the right moment. It watches what the model is doing - its reasoning, its tool calls, its failures - and when the running work matches something your memory store already learned somewhere else, it injects that one lesson as a short message right before the next step. Mechanically, a session event runs a free keyword match against a vocabulary derived from your memories, then gates apply (once per term per turn, once per memory, cooldown, per-turn quota), and a local small model judges whether the lesson changes what the agent should do next. Each card carries a reusable conclusion plus a memory id so the model can pull the full text, at most a few per turn and roughly 1.2 KB each.

Memory ★ 0 updated 2026-10-14
View on GitHub ↗

Install

dsh plugin --profile web add dsh-experience-recall

Copied verbatim from the README's npm install block; the README also documents a GitHub form (github:OverDustD7/dsh-experience-recall) and a local-checkout form.

Compatibility

Everything runs locally: the keyword table, the retrieval and the relevance judge use your own machine by default. Injected messages still consume the host model's context and input tokens, which the README states plainly. English README with a Simplified Chinese translation.

Details

Alternatives

863683348/dsh-plugin-focus · akslcw/dsh-negative-ledger · alcheme-labs/dsh-experience-map

More plugins in Memory

Browse more in Memory

Guides for Memory plugins