DeepSeek opens up full App capabilities: web search, file reading, voice interaction, all handled with one set of APIs.

Beating AI News Flash: DeepSeek has officially opened the "full suite" application API, providing third-party developers with web search, deep-thinking search, file parsing, web reading, and voice capabilities that were originally integrated into the official App.

In the past, developers calling the DeepSeek API mainly obtained the model's own reasoning capabilities. If they wanted to create an experience similar to the official App, they had to integrate search engines, parse PDFs, scrape web pages, and then feed the results to the model for processing.

Now, simply by enabling the corresponding parameters in the original API, DeepSeek will automatically complete tool calls and context stitching on the server side. The model can search web pages and read links while thinking, and can also answer questions by combining uploaded PDFs, Office documents, and images.

DeepSeek has also opened independent interfaces for search, web scraping, OCR (image text recognition), ASR (speech-to-text), and TTS (text-to-speech). Developers can either use the entire chat function directly or call only one of them.

The official team has also open-sourced a chat application Demo that supports multi-turn conversations, voice input, read-aloud replies, and importing DeepSeek chat history. Developers can directly modify it to build their own DeepSeek chat products.