Chat Completions Api

Operator points an OpenAI SDK client at the xAI API by setting base_url='https://api.x.ai/v1' and model='grok-4'. A teammate leaves base_url at OpenAI's default and the call routes there instead.

xAI exposes an OpenAI-compatible /v1/chat/completions endpoint at https://api.x.ai/v1. Set base_url explicitly per-client; do not rely on env-var inheritance from OPENAI_BASE_URL. Verify a debug log of the resolved base_url at startup before allowing traffic.

Agent posts to /v1/chat/completions with messages=[{role:'system',...}, {role:'user',...}, {role:'assistant',...}, {role:'user',...}] for a multi-turn Grok session.

Honor the OpenAI-compatible roles: system, user, assistant, tool. system may appear once (typically first) for instructions. Maintain user/assistant alternation after system. Do not stuff system content into a user turn — that changes how Grok weighs instructions vs user input.

Client calls /v1/chat/completions with stream=true. SSE chunks arrive as data: {choices:[{delta:{content:'...'}}]} lines terminated by data: [DONE].

Parse SSE line-by-line: split on '\n\n', strip 'data: ' prefix, JSON-decode each chunk, accumulate choices[0].delta.content, and stop on the literal string '[DONE]'. Do not feed '[DONE]' to JSON.parse.