UNPKG

@bitkyc08/opencodex

Version:

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM with Codex CLI/App/SDK and Claude Code

13 lines (12 loc) • 719 B
/** * Microsoft 365 Copilot refusals raised while collecting an answer, after the chat request was * already sent. They carry HTTP 413 because this adapter's buffers overflowed, but the provider * has processed the turn, so combo failover must not treat them as a pre-processing intake refusal * and send the turn to another target. Kept dependency-free so failover can read them. */ export const M365_RESPONSE_LIMIT_MESSAGES = { answer: "Microsoft 365 Copilot answer exceeded the buffer limit", reasoning: "Microsoft 365 Copilot reasoning exceeded the buffer limit", messages: "Microsoft 365 Copilot returned too many messages", frame: "Microsoft 365 Copilot frame buffer exceeded its limit", } as const;