bug(codex): `invalid_request_error` after auto compact

[dyxushuai]

Is it a request payload issue?
[ ] Yes, this is a request payload issue. I am using a client/cURL to send a request payload, but I received an unexpected error.
[x] No, it's another issue.

Describe the bug
Use codex with websocket

supports_websockets = true

invalid_request_error always occurs after codex auto compact.

Context compacted

⚠ Heads up: Long threads and multiple compactions can cause the model to be less accurate. Start a new thread when possible to keep threads small and targeted.

■ {"type":"error","status":400,"error":{
    "message": "No tool output found for function call call_rPQoj8lo6TtbE0VmFZMzOdMT.",
    "type": "invalid_request_error",
    "param": "input",
    "code": null
  }}

[KajiMaCN]

I ran into the same problem.

[DragonFSKY]

@dyxushuai Fixed in #2266.

After compact, Codex CLI sends the full compacted history as input without previous_response_id. The proxy was unconditionally merging this with stale pre-compact state (lastRequest + lastResponseOutput), which introduced orphaned function_call items that no longer had matching function_call_output — hence the 400.

The fix adds a check (inputContainsFullTranscript) that detects compaction / compaction_summary items or assistant messages in the input. When present, the stale merge is skipped and the compacted input is forwarded as-is.

Deployed and verified — compact followed by continued conversation works without errors.