Summary

Telegram Bot API 9.3 (Dec 31, 2025) introduced sendMessageDraft — a purpose-built method for streaming partial messages during generation. As of Bot API 9.5 (March 1, 2026), it's available to all bots.

Currently, OpenClaw's Telegram draft streaming uses the editMessageText approach: send a message, then repeatedly edit it as tokens arrive. This works but is throttled to ~1 update/second (DEFAULT_THROTTLE_MS = 1000 in src/telegram/draft-stream.ts) to stay within Telegram's rate limits on edits.

Request

Evaluate adopting sendMessageDraft as the streaming transport for Telegram channels when streaming: "partial" is configured. Since this is a dedicated streaming method (not a message edit), it may support faster update intervals than the current editMessageText workaround.

Additionally, consider exposing throttleMs as a configurable option under channels.telegram (e.g., channels.telegram.streamThrottleMs) so users can tune the tradeoff between responsiveness and rate-limit safety.

Context

• sendMessageDraft params: chat_id, text, parse_mode, entities, link_preview_options, reply_parameters, message_thread_id, reply_markup
• Return type: Message
• Available to all bots since Bot API 9.5 (March 1, 2026)
• Telegram changelog: https://core.telegram.org/bots/api-changelog
• Bot API docs: https://core.telegram.org/bots/api#sendmessagedraft

Current behavior

• src/telegram/draft-stream.ts uses sendMessage + editMessageText loop
• DEFAULT_THROTTLE_MS = 1000 (hardcoded, not configurable)
• TELEGRAM_STREAM_MAX_CHARS = 4096
• Config channels.telegram.streaming: "partial" enables draft streaming but throttle is not tunable

Environment

• OpenClaw v2026.3.1
• Node 22.22.0
• macOS (LaunchAgent gateway)