The Pulse
DeepSeek makes app search, files and voice available via API
Developers can use DeepSeek's app features through existing APIs or call standalone search, OCR and voice services. The extended capabilities carry tool fees in addition to token charges.
Developers can now add the search, file-reading, safety and voice features from the DeepSeek app to their own applications through its API. DeepSeek's API documentation also lists standalone services for search, page fetching, optical character recognition, speech recognition and text-to-speech.
The release was recorded in the company's change log on November 13. Extended capabilities carry tool usage fees in addition to standard token charges.
Search and files in conversations
The built-in features work with the V4.1-Flash model, called deepseek-flash, across the Messages, Responses and Chat Completions APIs. Tool calls and the assembly of context are handled by the service.
With thinking disabled and the deepseek_web_search tool enabled, the model can search the web or images and answer with citations. With thinking enabled, it can search and read links over multiple reasoning steps.
File input accepts PDFs, Office files, images and other formats. Extracted text or parsed content is added to the conversation context. A separate URL-reading tool, deepseek_read_url, detects links in user messages and fetches their contents for answers.
The service runs searches and URL reading; the client displays returned blocks and passes them back unchanged in the next round, according to the chatbot guide.
Safety and standalone tools
The optional safety check reviews model inputs and outputs and returns a refusal marker when it detects a risk. DeepSeek tells developers providing services to the public to enable the check and remain responsible for obligations including content safety review, model registration and AI-generated content labeling.
Search and Fetch are also available as standalone APIs, alongside optical character recognition, automatic speech recognition and text-to-speech. The chatbot guide describes voice input as transcription into a message and read-aloud as conversion of a model answer into speech.
Demo client
DeepSeek has also published an open-source chat client demonstrating the APIs. It supports importing exported DeepSeek Chat history, as well as search, document and image uploads, voice input and reading answers aloud.
The demo supports both importing and exporting chat history, but an import replaces the existing history. The guide says API keys, conversations and uploaded images are stored only in the browser, and describes the client as a demonstration.