@bitkyc08/opencodex
Version:
Universal provider proxy for OpenAI Codex & Claude Code — use any LLM with Codex CLI/App/SDK and Claude Code
8 lines (7 loc) • 515 B
text/typescript
/** Proxy-owned refusal text for an exhausted late-attribution budget. A compaction on the same
* Ollama route resends the same history, so the only reliable recovery is a new thread.
*
* Kept in its own module so combo failover can recognize this proxy-owned 413 without pulling
* the Ollama adapter into its import graph. */
export const OLLAMA_LATE_ATTRIBUTION_LIMIT_MESSAGE =
"Late tool-output attribution for this conversation exceeded the 256 KiB per-request limit. Start a new thread to continue.";