dsh-tongflow

TongFlow as a DeepSeek Harness (dsh) plugin — a film-crew studio inside your agent.

Three layers, never mixed up:

Layer	Owns
dsh	the harness: sessions, model routing, tools, jobs, web UI
the agent	creativity: script, characters, shot lists, prompts, review notes — plain files
TongFlow	deterministic generation: every image / voice / music / video is produced by running a saved workflow file (workflows/*.tongflow.json) through the TongFlow engine and its plugins

There is deliberately no "generate an image" tool. The agent writes a workflow, binds its inputs to project assets, runs it, reviews the take, circles the good one. Users open the same .tongflow.json on the embedded canvas, tweak it, and re-run.

Install

npx @deepseek-ai/dsh@next plugin --profile web add dsh-tongflow      # from npm
# or from a tarball:  pnpm --filter dsh-tongflow pack  →  dsh plugin --profile web add ./dsh-tongflow-x.y.z.tgz
npx @deepseek-ai/dsh@next web

Requirements: dsh ≥ 0.1.0-rc.7 (Node ≥ 22.19), Python ≥ 3.10 on PATH (or pythonPath in the plugin config), git, and ffmpeg for video contact sheets. On first use the plugin creates ~/.dsh/tongflow/venv with the tongflow SDK; TongFlow plugins are cloned into ~/.dsh/tongflow/plugins on demand.

Start a session whose first message begins with @tongflow — that session becomes a studio session: the conversation view turns into the Studio (chat column · project tree · preview / canvas · drawers for takes, runs and details, all in the UI language of your browser), and the agent gets the tongflow_* tools and skills. Any other session is untouched dsh. In the Studio: create a project (template manga-drama ships, with English / Chinese starter files), install TongFlow plugins and paste API keys under , then talk to the agent — or click a workflow and use the canvas directly.