dsh-honcho-sync: A Honcho Sync Plugin for Persistent Memory in DeepSeek Harness
nanpaidashi/dsh-honcho-sync
DSH plugin that auto-syncs conversation turns to Honcho memory, injecting persistent context and enabling semantic recall.
Let the DSH plugin for DeepSeek Harness automatically sync conversations to the Honcho memory service and inject memory context at session start, giving the AI semantic search and dialectic reasoning capabilities.
dsh plugin --profile web add github:nanpaidashi/dsh-honcho-sync- Category
- Memory & Context
- Platform
- DSH-Plugin
- Author
- nanpaidashi
- Distribution
- Plugin
dsh-honcho-sync Key Features
dsh-honcho-sync Repository
dsh-honcho-sync Repo Summary
dsh-honcho-sync is a Honcho memory bridge DSH plugin for DeepSeek Harness, maintained by nanpaidashi under the MIT license. It solves the problem of DSH sessions lacking persistent memory: after installation, DSH automatically syncs every conversation turn incrementally to a self-hosted Honcho memory service (NAS, server, or cloud), and at session start, it auto-injects Layer-1 memory context (peer cards + representation) into the system prompt, giving the AI semantic search and dialectic reasoning tools. Core capabilities include a visual settings panel, incremental auto-sync, per-day session splitting, and the honcho_recall / honcho_ask / honcho_status tools.
What core features does the dsh-honcho-sync DSH plugin support?
This DSH plugin offers: a visual settings panel (DSH Settings → Honcho Memory, no manual YAML editing); incremental auto-sync (event-count based, default 3s debounce, no lost or duplicate messages); per-day session splitting (Session ID with date suffix, e.g., dsh-home-gsq-DeepSeek-2026-08-19, keeping sessions small for faster dialectic search); Layer-1 memory injection (on first user message, fetches peer cards + representation + session summary and injects into systemPrompt); honcho_recall semantic search (2-5s, returns relevant message snippets from last 7 days); honcho_ask dialectic reasoning (2-5min, deep Q&A across 7-day sessions); honcho_status health check; timezone conversion (UTC to Beijing time UTC+8); and sync state persistence (written to ~/.dsh/honcho-sync-state.json, survives restarts).
How to install dsh-honcho-sync?
Prerequisite: a deployed Honcho memory service (one-line Docker command). Recommended installation from GitHub:
dsh plugin --profile web add github:nanpaidashi/dsh-honcho-sync
Alternatively, install from npm:
dsh plugin --profile web add @nanpaidashi/dsh-honcho-sync@latest
How to configure dsh-honcho-sync?
There are three configuration methods: visual panel (recommended), environment variables, and cordis.patch.yml. In the visual panel, go to DSH Settings → Honcho Memory and fill in Honcho URL, Workspace, User/Agent Peer, Debounce, Recall Budget, Max Chars, etc., then click Save to apply without restart. For environment variables, set HONCHO_URL and HONCHO_WORKSPACE (required), with HONCHO_USER_PEER and HONCHO_AGENT_PEER optional. For cordis.patch.yml, append a config block to ~/.dsh/profiles/web/cordis.patch.yml, supporting parameters like sessionStrategy (per-day or per-directory), debounceMs, autoRecall, recallBudget, etc.
What is the architecture of dsh-honcho-sync?
Architecturally, DSH Session triggers incremental sync (with debounce and event-count persistence) to the Honcho API via session/event; on the first user message, Layer-1 injection fetches Peer Card, Representation, and Session Summary and injects them into systemPrompt's scoped dynamic context; honcho_recall calls POST /search (7-day session scope), and honcho_ask calls POST /peers/{peer}/chat (dialectic reasoning).
Which DeepSeek Harness versions are compatible with dsh-honcho-sync?
This DSH plugin requires DSH ≥ 0.1.0-rc.6, is maintained by nanpaidashi, last updated on 2026-08-19, and is licensed under MIT.