DeepSeek Chat Completions Require Full History Replay

Migrated with permission from DeepSeek V4 Pro, a related editorial property operated by our team.

Checked on June 27, 2026: DeepSeek's official multi-round conversation guide says /chat/completions does not store context server-side, so each follow-up call must resend the prior conversation history, including earlier assistant output, before the next user turn.

What we verified on June 27, 2026

DeepSeek's official multi-round conversation guide says /chat/completions is stateless: the server does not keep prior turn context for you.

The official fix is explicit rather than implied: each later request must concatenate all previous conversation history and send it back to the API.

DeepSeek's own example appends the assistant response from round one before sending round two, then adds the next user question after that.

The published sample still uses the ordinary OpenAI-compatible route with base_url=https://api.deepseek.com and model=deepseek-v4-pro, which means this is not a special SDK-only edge case.

The guide makes the support boundary clearer for production apps: if a follow-up answer looks forgetful, the first debugging target is usually the caller's messages replay, not a hidden DeepSeek memory failure.