AI Agent Hub
Back to plugins
dsh-provider-veark preview

dsh-provider-veark

Web Tools Updated 2026.09.01

Run the following command in DeepSeek Harness:

dsh plugin install IcedWatermelonJuice/dsh-provider-veark

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

Run dsh plugin install IcedWatermelonJuice/dsh-provider-veark (source: https://github.com/IcedWatermelonJuice/dsh-provider-veark) in your terminal, then restart DSH to configure the API key in the plugin card.

About this plugin

Volcano Ark Coding Plan is already great value, but the built-in DeepSeek Harness routing has a well-known weakness when wired to Ark: every single conversation turn must re-send the full image as a base64 blob, dragging multi-round chats down in both speed and token cost, and PDF document understanding is simply out of the picture.

This plugin cuts straight to the official Responses API and Coding gateway to sidestep those bottlenecks. Images are uploaded to the Ark Files API on first use; every subsequent turn references a lightweight file_id instead of re-transmitting pixels, with an automatic base64 fallback if the upload path is unavailable. PDFs travel as native input_file payloads through a plugin-local sidecar, leaving the Harness core untouched and requiring no TOS dependency. Text, image, and PDF inputs all work the moment you paste an API key.

If you are already paying for Volcano Ark Coding Plan and regularly paste images or PDFs into DSH for multi-turn analysis, this plugin dramatically cuts the repetitive cost of multimodal dialogue and removes the need to juggle providers or hand-roll upload scripts.

Screenshots

Use Cases

  • Re-analyzing the same screenshot or chart across turns without re-sending pixels
  • Uploading PDF technical docs for Q&A, summarization, and cross-section lookup
  • Running daily coding assistance and knowledge retrieval on a low-cost plan

Best For

  • Developers already subscribed to Volcano Ark Coding Plan
  • Researchers needing multi-turn image and document deep-dive analysis
  • Solo developers seeking low-cost multimodal inference