dsh-honcho-sync — Honcho Memory Plugin for DeepSeek Harness GitHub License: MIT DSH ≥ 0.1.0-rc.6 > Give your DeepSeek Harness AI persistent memory. > Auto-sync every conversation turn to a self-hosted Honcho service, and give the AI built-in tools to recall, search, and remember across sessions. Configure everything from a visual settings panel — no YAML editing required. --- ## 中文 ### 这是什么? DeepSeek Harness (DSH) ↔ Honcho 记忆桥接插件。 安装后 DSH 自动将每轮对话同步到用户自建的 Honcho 记忆服务(NAS、服务器或云端均可),AI 获得内置记忆工具,可直接检索、保存跨会话信息。 ### 功能 | 功能 | 说明 | |------|------| | 可视化设置面板 | DSH 设置 → "Honcho Memory",所有配置项图形化编辑,无需手动改 YAML | | 自动同步 | 每轮对话自动推送到 Honcho(默认 debounce 3s),无需 AI 主动操作 | | honcho_recall | 从长期记忆中语义搜索相关上下文,会话开始或重要任务前调用 | | honcho_search | 跨所有 session 精确搜索,适合查找具体信息 | | honcho_remember | 保存重要事实、决策、偏好到持久记忆 | | honcho_context | 读取当前 session 的完整对话上下文 | | honcho_status | 检查 Honcho 服务器健康状态 | | 上下文注入 | 会话启动时自动从 Honcho 加载近期上下文 | | 去重 | 同一 session 不重复推送相同内容 | ### 安装 前提: 已部署 Honcho 记忆服务(Docker 一行命令即可) bash # 方式 1:从 GitHub 直接安装(推荐) dsh plugin --profile web add link:https://github.com/nanpaidashi/dsh-honcho-sync dsh web # 方式 2:从 npm 安装 dsh plugin --profile web add @nanpaidashi/dsh-honcho-sync@latest ### 配置 方式 1:可视化面板(推荐) 打开 DSH → 左下角 设置Honcho Memory,直接填写: - Honcho URL(API 地址) - Workspace(工作区名称) - User / Agent Peer(标识符) - Debounce、Recall Budget、Max Chars(调优参数) - Auto Recall / Auto Sync(开关) 点击 Save 即生效,无需重启。 方式 2:环境变量 bash export HONCHO_URL="http://your-honcho-server:8000" # 必填 export HONCHO_WORKSPACE="your-workspace" # 必填 方式 3:cordis.patch.yml~/.dsh/profiles/web/cordis.patch.yml 中追加: yaml - id: honcho-sync config: honchoUrl: "http://your-honcho-server:8000" # Honcho API 地址(必填) workspace: "your-workspace" # Workspace 名称 userPeer: "user" # 用户标识 agentPeer: "agent" # Agent 标识 debounceMs: 3000 # 同步防抖(毫秒) autoRecall: true # 会话启动时自动 recall recallBudget: 2000 # recall token 上限 autoSync: true # 自动同步开关 messageMaxChars: 25000 # 单条消息最大字符数 环境变量速查: | 变量 | 默认值 | 必填 | 说明 | |------|--------|------|------| | HONCHO_URL | — | | Honcho API 地址 | | HONCHO_WORKSPACE | — | | Workspace 名称 | | HONCHO_USER_PEER | user | 否 | 用户标识 | | HONCHO_AGENT_PEER | agent | 否 | Agent 标识 | | HONCHO_DEBOUNCE_MS | 3000 | 否 | 同步防抖延迟(毫秒) | ### 架构 DSH Session │ ├─ session/event ──→ 自动同步(debounce)──→ Honcho API (POST messages) │ ├─ honcho_recall / search / remember / context / status │ └────────→ AI 可调用的记忆工具 │ ├─ session/start ──→ 上下文自动注入(autoRecall) │ └─ 设置面板 ──→ /_dsh/dsh-honcho-sync/status (loopback-only) ### 依赖 - Honcho — 需自行部署(Docker 一行命令) - DeepSeek Harness ≥ 0.1.0-rc.6 ### 许可证 MIT --- ## English ### What is this? A bridge plugin connecting DeepSeek Harness to Honcho memory. After installation, DSH auto-syncs every conversation turn to a self-hosted Honcho service (NAS, server, or cloud), and the AI gains built-in tools to recall, search, and remember across sessions. All settings are editable from a visual panel — no YAML editing required. ### Features | Feature | Description | |---------|-------------| | Visual Settings Panel | DSH Settings → "Honcho Memory" — all configuration in a graphical UI, no YAML editing | | Auto-sync | Every conversation turn synced to Honcho (debounced 3s default), no AI action needed | | honcho_recall | Semantic search across long-term memory. Call at session start or before important tasks. | | honcho_search | Search across ALL sessions for precise lookups. | | honcho_remember | Save facts, decisions, preferences to persistent memory. | | honcho_context | Get full conversation context for the current session. | | honcho_status | Check Honcho server health. | | Context injection | Auto-load recent context from Honcho on session start. | | Deduplication | Same content in same session won't be pushed twice. | ### Installation Prerequisite: Deploy Honcho first (one Docker command) bash # Option 1: Install from GitHub (recommended) dsh plugin --profile web add link:https://github.com/nanpaidashi/dsh-honcho-sync dsh web # Option 2: Install from npm dsh plugin --profile web add @nanpaidashi/dsh-honcho-sync@latest ### Configuration Option 1: Visual panel (recommended) Open DSH → bottom-left SettingsHoncho Memory, then fill in: - Honcho URL (API address) - Workspace (name) - User / Agent Peer (identifiers) - Debounce, Recall Budget, Max Chars (tuning) - Auto Recall / Auto Sync (toggles) Click Save — takes effect immediately, no restart needed. Option 2: Environment variables bash export HONCHO_URL="http://your-honcho-server:8000" # required export HONCHO_WORKSPACE="your-workspace" # required Option 3: cordis.patch.yml Append to ~/.dsh/profiles/web/cordis.patch.yml: yaml - id: honcho-sync config: honchoUrl: "http://your-honcho-server:8000" # Honcho API URL (required) workspace: "your-workspace" # Workspace name userPeer: "user" # User identifier agentPeer: "agent" # Agent identifier debounceMs: 3000 # Sync debounce (ms) autoRecall: true # Auto-recall on session start recallBudget: 2000 # Token budget for recall autoSync: true # Auto-sync toggle messageMaxChars: 25000 # Max chars per message Environment variables: | Variable | Default | Required | Description | |----------|---------|----------|-------------| | HONCHO_URL | — | Yes | Honcho API base URL | | HONCHO_WORKSPACE | — | Yes | Workspace name | | HONCHO_USER_PEER | user | No | User peer_id | | HONCHO_AGENT_PEER | agent | No | Agent peer_id | | HONCHO_DEBOUNCE_MS | 3000 | No | Sync debounce delay (ms) | ### Architecture DSH Session │ ├─ session/event ──→ Auto-sync (debounced) ──→ Honcho API (POST messages) │ ├─ honcho_recall / search / remember / context / status │ └────────→ Memory tools callable by the AI │ ├─ session/start ──→ Context auto-injection (autoRecall) │ └─ Settings panel ──→ /_dsh/dsh-honcho-sync/status (loopback-only) ### Dependencies - Honcho — Self-hosted memory service (one Docker command) - DeepSeek Harness ≥ 0.1.0-rc.6 ### License MIT