DeepSeek Launches Comprehensive API for Online Search and Voice Interaction

DeepSeek has officially launched its comprehensive application API, integrating online search, deep thinking search, document parsing, web reading, and voice capabilities for third-party developers. Previously, developers could only access the DeepSeek API for model inference capabilities, and to achieve an experience similar to the official app, they had to independently integrate search engines, parse PDFs, and scrape web pages. Now, developers simply need to enable the corresponding parameters in the API, and DeepSeek will automatically handle tool calls and context stitching, allowing the model to think while searching the web, reading links, and answering questions based on uploaded PDFs, Office documents, and images. Additionally, DeepSeek has opened independent interfaces for search, web scraping, OCR, ASR, and TTS, enabling developers to either use the entire set of dialogue functions or call specific features individually. The official team has also open-sourced a chat application demo that supports multi-turn conversations, voice input, read-aloud responses, and importing historical chat records, allowing developers to build their own DeepSeek chat products based on this.