deepseek-channel-octo
by quanming1
把 DeepSeek Harness(dsh)agent 桥接进 Octo IM
Bridge DeepSeek Harness (dsh) agents into Octo IM
安装
dsh plugin --profile web add github:quanming1/deepseek-channel-octoGitHub 源码安装:首次需按提示配置 allowBuilds 构建授权后重试
安装与环境配置指引、插件开发教程见 DSH 中文社区文档 ↗
安装即在你的机器上以你的权限运行第三方代码——它可读写文件、使用凭据、访问网络,DSH 的工具审批不会为插件代码加沙箱。「检测到 manifest」仅代表发现 dsh.bundle / dsh.plugin 清单,不构成兼容性或安全审查;安装前请审阅源码,不熟悉的插件先在不含密钥的环境试用。
README
目录
Bridge DeepSeek Harness (dsh) agents into Octo IM — a dsh agent becomes an Octo bot/teammate: send/receive messages, execute tasks, stream cards, and handle approvals.
Status: prototype. The Octo WebSocket channel MVP is live-verified (C2); config-driven multi-bot support is done (C3). Implementation follows the staged plan in
docs/TODO.yaml.
Quick start
Connect an Octo bot to a dsh agent in three steps.
1. Install
npm install -g @deepseek-ai/dsh@0.1.0-rc.6 # dsh CLI
npm install -g deepseek-channel-octo # this package
2. Create a bot and configure it
Create a bot in the Octo admin console — you get a Bot ID and a Bot Token (bf_ prefix).
Copy the config template and fill in your credentials:
cp octo.config.example.yaml octo.config.yaml
# octo.config.yaml
apiUrl: https://your-octo-server.example/api
bots:
- name: my-bot
botUid: xxxxxxxxxxxxx_bot
botToken: bf_your_bot_token_here
Add one entry per bot under bots to connect multiple bots — each gets its own
WebSocket connection and shares the same dsh harness. octo.config.yaml is
git-ignored (it contains secrets); the path can be overridden with OCTO_CONFIG.
No config file? Legacy env vars are used as fallback:
OCTO_API_URL / OCTO_BOT_TOKEN / OCTO_BOT_UID.
3. Run
dsh-octo-bot octo run
Add the bot to a group and @mention it — the agent replies with the dsh answer.
Each group maps to its own dsh session (octo:<accountId>:<chatId>), so the bot
remembers the conversation per group (restart-safe).
Workflow (Rondo method)
This repository runs a PRD-driven, agent-constrained workflow. Read these files before touching anything:
| File | Purpose |
|---|---|
AGENTS.md |
Behavior contract for all AI agents and humans (read first) |
docs/TODO.yaml |
Staged, structured task list — the single source of execution truth |
docs/PROCESS.md |
How to move a stage forward (six-step closed loop) |
docs/prd/PRD-<stage>-*.md |
Stage PRD — the single source of truth for development |
.githooks/ |
Local enforcement: commit-msg (commit rules) + pre-push (main protection) |
Enable hooks
git config core.hooksPath .githooks
Branch model
Single-main, all-PR flow: feature branches are cut from main and merged only via
GitHub PRs. See AGENTS.md §4.
Development
pnpm install # install deps (pnpm 11, Node >= 22.19)
pnpm typecheck # type check (tsc --noEmit)
pnpm lint # ESLint
pnpm test # vitest unit tests
pnpm build # tsup build -> dist/
License
Apache-2.0.
原始 README: https://github.com/quanming1/deepseek-channel-octo/blob/main/README.md ↗
同类插件
查看全部 →
deepseek-harness
从仓库或系统描述生成经过校验的自包含交互式架构图、流程图、时序图、数据流图和生命周期图。

dsh-plugin
通过 DSH MCP 客户端挂载 Ouroboros 的纯配置包,在 DSH 中提供 36 个涵盖需求访谈、Seed、执行、评估与演化流程的工具。

dsh-tongflow
基于 TongFlow 的“片场”插件,用于图片、配音、音乐与视频制作:agent 为每个资产生成 TongFlow 工作流文件(.tongflow.json)并通过 TongFlow 插件执行,内嵌工作流画布,按镜头/角色/take 组织项目,附漫剧模板;以 @tongflow 开头的会话进入 Studio 界面。

helloagents
AI 编码 CLI 的工作流层:技能、项目知识、交付检查、更安全的配置写入与可恢复执行

dsh-ai-novel-writer
安装专用 AI 小说创作预设与工作台:提供带修订号的本地项目资产、紧凑侧边工作台,以及需要原生审批的逐文件变更。

rea
用 agent 逆向任何东西:从应用行为到原生二进制