AI Agent Hub
Back to plugins
🤖

dsh-seedance2

Model Inference Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install synmindai/dsh-seedance2

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install synmindai/dsh-seedance2 in DeepSeek Harness to install this plugin; the source repository is at https://github.com/synmindai/dsh-seedance2

About this plugin

DeepSeek Harness covers text inference and code generation well, but image and video generation have been a gap. dsh-seedance2 bridges that gap through the Seedance 2 public API: a single plugin lets you generate images with Nano Banana Pro, Nano Banana 2 Lite, or GPT Image 2, and produce videos with Seedance 2 or Seedance 2.5, plus task-polling and account-assistance tools so your multimodal pipeline stays inside one Harness session.\n\nBilling and safety are handled thoughtfully. Paid requests require explicit approval by default, every call carries an idempotency key so automatic retries never double-charge, transient 429 and 5xx responses trigger bounded backoff, failed tasks surface a credits_refunded value from the API, and a foreground timeout guides you to resume the original task rather than resubmit and pay again. These guardrails matter a lot when you are running batch generation and do not want a runaway invoice.\n\nIf you already have a DeepSeek Harness inference pipeline and want controllable image and video generation without spinning up extra services, this plugin is a lightweight, build-free, production-ready option. Its defaults are tuned for real workloads; override individual settings in the profile patch only when you need to.

Use Cases

  • Generate promotional or concept images inline within a Harness workflow using multiple engines
  • Batch-submit short video generation tasks and track status via seedance2_get_task
  • Plug multimodal output into automation pipelines without extra SDKs or services

Best For

  • Teams running inference pipelines on DeepSeek Harness that need a multimodal layer
  • Generative-AI developers who care about billing safety, idempotent retries, and approval flows
  • Users who want image and video workflows contained in a single Harness session