Chat Completions Openai Compatible

An existing OpenAI-SDK codebase is being pointed at DeepSeek. The integrator leaves base_url at the OpenAI default and only swaps the API key, expecting deepseek-chat to respond.

Set base_url to https://api.deepseek.com (the OpenAI SDK reuses the same client; only base_url and api_key change). Requests otherwise keep the OpenAI-compatible /chat/completions shape. Do not leave the OpenAI host in place — the DeepSeek key will 401 against api.openai.com.

A latency-sensitive autocomplete feature is wired to model=deepseek-reasoner for every keystroke 'because it is smarter'.

Route latency-sensitive, low-reasoning tasks to deepseek-chat; reserve deepseek-reasoner for tasks that benefit from chain-of-thought. deepseek-reasoner emits extra reasoning_content tokens and is slower/costlier per call — do not use it as the default for high-frequency lightweight requests.

A long-generation request sets max_tokens too low. The response returns choices[0].finish_reason='length' with an obviously cut-off final sentence.

Detect finish_reason='length' and either raise the output as a partial completion or issue a continuation (append the truncated assistant message and a continue turn). Never present a length-truncated answer as complete.