joeseesun/qiaomu-deepseek-api

The skill includes a command-line interface that uses only the Python standard library, sending requests directly to the official DeepSeek API: standalone search and web scraping do not require a conversational model; answers with citations, deep search, link reading, and image-based search use the Flash conversational tool; document uploads are cleaned up after Q&A; additionally, it offers page-by-page OCR, audio-to-text transcription, and Mandarin WAV voiceover, prompting you to choose between two voices if no voice is specified.

The skill includes a command-line interface that uses only the Python standard library, sending requests directly to the official DeepSeek API: standalone search and web scraping do not require a conversational model; answers with citations, deep search, link reading, and image-based search use the Flash conversational tool; document uploads are cleaned up after Q&A; additionally, it offers page-by-page OCR, audio-to-text transcription, and Mandarin WAV voiceover, prompting you to choose between two voices if no voice is specified.

Why It Matters This skill consolidates DeepSeek's newly released search, link reading, file Q&A, OCR, speech recognition, and synthesis into a single tool, eliminating the need to connect three separate tools for scanned documents, recordings, and web pages.

Stars 21

Reasons

  • The repository 11 was created on 14 under the MIT license; the README has been reviewed, covering usage, common errors, and data handling
  • It corresponds to the application API released by DeepSeek on 11 13

Usage

  • Install via npx skills add joeseesun/qiaomu-deepseek-api, configure DEEPSEEK_API_KEY in the host environment (or run python3 scripts/deepseek_api.py setup in your own terminal), then state requirements such as "Use DeepSeek to review this scanned PDF"

Risks

  • Uploaded and input content is sent to DeepSeek, and tool calls may incur additional charges; audio transcription supports only WAV or Ogg Opus, with a maximum duration of 120 seconds per segment
  • The end of the README includes the author's WeChat Official Account and donation promotion

Open

GitHub

Explore the latest AI Radar updates