An application using DeepSeek's thinking mode needs to preserve more than the final answer when it supplies tools. The API returns generated reasoning separately, in reasoning_content. Its current documentation requires applications to send that field back from every previous turn when the request includes tools, even if a particular turn made no tool call. Omitting it can produce a 400 error.
Without tools in the request, previous reasoning content is ignored. The integration therefore needs different conversation-history handling for these two cases; a client that keeps only visible answers can discard information the tool-enabled API requires.
Terms in this piece · Glossary
chain-of-thought — Having a model write out intermediate reasoning steps before its answer, which markedly improves performance on hard problems.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.