tongflow-api-deepseek

Official TongFlow plugin. Routes text generation through the DeepSeek V4 API — an OpenAI-compatible, text-only model family.

Capabilities

Implements these ABI slots (runs locally as a Python process, no GPU):

• Generate / rewrite text (gen-text) — create or edit copy from a prompt.
• Combine text (combine-text) — merge multiple text inputs into one.
• Split long text (split-text) — break a long passage into chunks.
• Arrange & batch groups (arrange-group) — group and arrange batches.
• Filter or drop clips (drop-video) — drop unwanted clips by rule.

Models

Each text node exposes a model dropdown with four choices:

Selection	Model	Thinking
deepseek-v4-flash (default)	deepseek-v4-flash	off
deepseek-v4-pro	deepseek-v4-pro	off
deepseek-v4-flash-thinking	deepseek-v4-flash	on
deepseek-v4-pro-thinking	deepseek-v4-pro	on

With a *-thinking selection the completion is streamed and the model's reasoning appears live in an auto-scrolling bubble beside the node; the final answer is returned as the node output.

Credentials

Add in TongFlow Settings (gear icon, top-right):

Key	Required	Notes
DEEPSEEK_API_KEY	✅	Create one at platform.deepseek.com/api_keys.
DEEPSEEK_BASE_URL	optional	Override the endpoint (defaults to https://api.deepseek.com).

Values are stored locally and take effect without a restart.

Development

Requires tongflow>=0.2.17 (the streaming thinking bubble uses progress(..., thinking=True)). Install into a clean venv and type-check:

pip install "tongflow==0.2.17"
pyright entry.py