AI News

DeepSeek Opens Full-Suite App API, Allowing Developers to Replicate the DeepSeek App 1:1

ZhidxUpdated Nov 14 04:18

AI summary

DeepSeek has officially opened the full-suite app API, integrating App capabilities such as web search, deep-thinking search, file input, and link reading into the existing API; developers can call them by enabling the corresponding parameters. It has also opened independent tool interfaces including search, web scraping, OCR, speech recognition, and speech synthesis, and DeepSeek says developers can replicate the complete DeepSeek App 1:1. In addition to model Token fees, using the extended capabilities also requires paying tool invocation fees based on the tools called and usage; DeepSeek-V4.1-Flash (deepseek-flash) already supports multiple new capabilities across the three types of interfaces: Messages, Responses, and Chat Completions.

Why read this

AI score 76

The original text lists tool interfaces such as web search, file parsing, and voice, as well as the billing method, so readers can judge the cost and barrier to building their own app.

Topics

Tags

Body · AI translation

Zhidongxi November 13 news: Just now, DeepSeek officially opened the full-suite application API, integrating capabilities from the DeepSeek App such as web search, deep-thinking search, file input, and link reading into the original API. Developers only need to enable the corresponding parameters to call the relevant functions.

At the same time, DeepSeek also opened independent tool interfaces such as search, web scraping, OCR, speech recognition, and speech synthesis, supporting developers to call them individually or in combination.

DeepSeek said that with the capabilities opened up this time, developers can even replicate the complete DeepSeek App 1:1.

To demonstrate the actual effect of the new API, DeepSeek also built a demo of the DeepSeek App based on the above interfaces, and it can be seen that it is almost identical to the DeepSeek web version.

Demo experience address:

GitHub sample code address:

In terms of pricing, DeepSeek said that when developers use extended capabilities such as web search, file parsing, and speech through the API, in addition to model Token fees, they also need to pay the corresponding tool invocation fees. The relevant fees are calculated based on the tools actually invoked and the usage volume, with specific prices as follows:

▲Tool API price list (Source: DeepSeek)

Currently, the DeepSeek-V4.1-Flash model (deepseek-flash) already supports multiple new capabilities in the three types of interfaces: Messages, Responses, and Chat Completions.

Specifically, the functions opened up by this full-suite application API include:

Web search: In non-thinking mode, after enabling the deepseek_web_search tool, the model can automatically initiate web or image searches based on user questions and attach cited sources in the answer;

Deep-thinking search: In thinking mode, after enabling the deepseek_web_search tool, the model can autonomously search the web and read links during multi-step reasoning, and complete answers by combining retrieval results;

File input: Supports formats such as PDF, Office documents, and images, and can perform plain text extraction or mixed text-image parsing on files, and add the results to the conversation context;

Link reading: After enabling the deepseek_read_url tool, the model can automatically identify web links in user messages, scrape web page content, and answer questions in combination with the context;

Model security: Developers can enable security verification parameters to conduct basic risk control review on model input and output, and return a refusal identifier when a risk is hit.

Among these, DeepSeek particularly reminds that developers providing application services to external users should enable security verification parameters and at the same time fulfill obligations related to content security review, large model registration, and AI labeling.

Conclusion: DeepSeek wants to make reinventing the wheel simpler and innovation greater

This time, DeepSeek directly integrates these capabilities into the DeepSeek API. The relevant tool invocation and context stitching are automatically completed by the server side, which can lower the threshold for developers to build complex AI applications.

In the announcement, DeepSeek mentioned that existing conversational products are only a basic form of interaction between humans and models. Opening up the various capabilities behind the official product this time is intended to hope that developers reduce repetitive basic work and devote more energy to product innovation.

For developers, this means that the DeepSeek API is no longer just a model invocation entry point, but can also become the capability foundation for building complete AI applications.

Source: DeepSeek

Source

Zhidx