openclaw - 💡(How to fix) Fix [Bug]: WhatsApp outbound media sends falsely succeed on 2026.4.5, attachment is dropped and only caption/text is delivered [4 comments, 4 participants]

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

Summary

On OpenClaw 2026.4.5, outbound WhatsApp media sends can return a successful tool result with a real-looking WhatsApp messageId, but the recipient only receives the caption/text. The actual attachment never arrives.

This affects direct message tool sends on WhatsApp.

Environment

• OpenClaw: 2026.4.5
• OS: macOS (Apple Silicon)
• Channel: WhatsApp Web / Baileys
• Gateway mode: local LaunchAgent
• WhatsApp account: linked and healthy
• Text-only WhatsApp sends still work

What I tested

I tested outbound WhatsApp sends using the message tool with:

1. an image
2. a normal audio file
3. an audio send with asVoice=true

All 3 returned successful results with:

• channel: "whatsapp"
• via: "gateway"
• mediaUrl: <local file path>
• a runId
• a WhatsApp messageId
• a toJid

Example behavior:

• tool returns success
• gateway logs show message sent
• recipient only gets the caption/text like:
  • whatsapp image test from lana
  • whatsapp audio file test from lana
  • whatsapp voice-note test from lana
• no actual attachment appears in WhatsApp

Expected behavior

If media is provided, WhatsApp should deliver the actual attachment plus optional caption.

Actual behavior

The send is reported as successful, but only the text/caption is delivered. The media payload is silently dropped.

Important details

• This is not a total WhatsApp outage.
• Text-only sends still work.
• This is not just a FishAudio or voice-note generation problem.
• This is not just the PTT lane.
• Image sends, plain audio sends, and asVoice=true sends can all degrade into text-only delivery.
• WhatsApp remains linked and healthy in openclaw status --deep.

Investigation so far

I diffed the npm tarballs for 2026.4.2 and 2026.4.5.

Findings:

• The low-level WhatsApp/Baileys payload builder appears basically unchanged between those versions.
• The existing audio -> ptt: true behavior was already present before 2026.4.5, so that alone does not explain this regression.
• 2026.4.5 includes significant refactoring in the higher-level WhatsApp outbound plumbing, especially around:
  • send-lXVwx8YQ.js
  • attached channel result adapters
  • outbound send dependency wiring
  • WhatsApp channel setup/runtime packaging

Current suspicion:

• the regression is likely in the higher-level outbound media mapping or delivery path before the final WhatsApp socket send
• the runtime appears to falsely acknowledge success while the attachment is lost and only the caption survives

Related issue

This looks related in spirit to:

• #54131

But this issue seems broader, because it is happening on 2026.4.5 and appears to affect direct message tool sends too, not just MEDIA: token agent replies.

Repro idea

Using message tool or equivalent CLI send:

• send image with caption on WhatsApp
• send audio file on WhatsApp
• send audio with asVoice=true

Observe:

• success result with messageId
• recipient receives caption text only
• no media attachment

Steps to reproduce

1. Install or update OpenClaw to 2026.4.5.
2. Make sure WhatsApp is linked and shows healthy in openclaw status --deep.
3. Send a WhatsApp message with the message tool using a local image file and a caption.
4. Observe the tool result:
   • it returns success
   • includes channel: "whatsapp"
   • includes via: "gateway"
   • includes a real-looking messageId
5. Check the recipient phone.
6. Notice that only the caption text arrives, and the image attachment is missing.

Also reproduces with audio

Repeat the same test with:

• a normal audio file
• an audio file with asVoice=true

Observed result is the same:

• tool reports success
• WhatsApp recipient receives only the caption/text
• no media attachment is delivered

Expected behavior

In previous version, media attachment (audio, image, etc) would successfully send to Whatsapp.

Actual behavior

Caption text received via Whatsapp without the actual attachment.

OpenClaw version

2026.4.5

Operating system

macOS 26.4

Install method

npm

Model

openai-codex/gpt-5.4

Provider / routing chain

openclaw -> gpt-5.4

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response