Bug Description

This issue was reported by Hermes Agent on behalf of @oxngon.

Bug Description

When sending /btw <prompt> (or /background <prompt>) with image attachments on Discord (and likely other platforms), the images are cached to disk by the adapter but never forwarded to the spawned background agent. The agent receives only the text prompt and has no access to the attached images.

Steps to Reproduce

Steps to Reproduce

1. Send a message on Discord: /btw Which one should I buy? with 1-2 images attached
2. The background agent responds without any knowledge of the images
3. The images were cached (visible in logs) but never passed to the agent

Expected Behavior

Expected Behavior

The background agent should receive and process image attachments, just like a regular message does.

Actual Behavior

Actual Behavior

The background agent only receives the text prompt. Images are silently dropped.

Affected Component

Gateway (Telegram/Discord/Slack/WhatsApp)

Messaging Platform (if gateway-related)

Discord

Debug Report

Debug report: https://paste.rs/g9Idk
agent.log: https://paste.rs/w4GOl
gateway.log: https://paste.rs/qTEIe

Operating System

Ubuntu 24.04

Python Version

3.11.15

Hermes Version

v0.13.0 (2026.5.7)

Additional Logs / Traceback (optional)

### Notes

- The `/btw` command is an alias for `/background` — both are affected
- Regular (non-command) text+image messages work correctly — the image processing in `handle_message` (line ~6731) handles them properly
- The fix is backward-compatible: background tasks without images continue to work identically

Root Cause Analysis (optional)

Root Cause

In gateway/run.py, _handle_background_command (line ~10164) extracts only the text via event.get_command_args().strip() and passes it to _run_background_task. The event.media_urls and event.media_types fields are never forwarded.

The _run_background_task method then calls agent.run_conversation(user_message=prompt) with the bare text prompt — no image context.

Proposed Fix (optional)

Proposed Fix

1. Forward event.media_urls and event.media_types from _handle_background_command to _run_background_task
2. In _run_background_task, before calling agent.run_conversation(), enrich the prompt with vision descriptions using the existing _enrich_message_with_vision() method (the same path used for regular messages)

This reuses the existing vision infrastructure and requires minimal code changes.

Are you willing to submit a PR for this?

• I'd like to fix this myself and submit a PR