Summary

Environment

• Project: openclaw

• Model: openai-codex / gpt-5.3-codex

• Platform: Feishu (works fine), Telegram (problem occurs)

• Streaming config: streaming: partial

Description

When using the openai-codex / gpt-5.3-codex model:

• In Feishu: The model works correctly. It can respond with multiple messages, maintain typing status, and execute the actual operations properly.

• In Telegram: Only a single message is returned. The Telegram typing status disappears immediately, and no subsequent responses or actual operations are executed.

The streaming configuration in my openclaw settings is set to partial.

Steps to reproduce

using the openai-codex / gpt-5.3-codex model both in feishu and telegram

Expected behavior

Telegram bot should behave the same as Feishu:

• Maintain typing status during response

• Return multiple messages if needed

• Execute the actual operation

Actual behavior

• Only one reply is sent

• Typing status stops instantly

• No further response or operation execution

OpenClaw version

2026.2.24

Operating system

opencloudos

Install method

npm

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response