zeroclaw-labs/zeroclaw

[Bug]: Agent request appends each subsequent image in each request when sending two or more images on Telegram.

Open

#5514 opened on Apr 8, 2026

View on GitHub
 (1 comment) (0 reactions) (1 assignee)Rust (31,341 stars) (4,614 forks)batch import
bugchannelchannel:telegramhelp wantedpriority:p2risk: mediumruntimestatus:acceptedstatus:no-stale

Description

Affected component

runtime/daemon

Severity

S3 - minor issue

Current behavior

Currently when you send two or more images into Zeroclaw on Telegram, the gateway perceives each image as one request to the LLM, causing the agent to output two or more output messages, equal to the amount of messages you send. Additionally, the first message would be about the first image, second message will be with the first and second image, and so on.

Expected behavior

The agent should be able to accept all the images as one request while abiding to the max_images limit in config.toml, and output as one message.

Steps to reproduce

1. Onboard on Zeroclaw with a fresh Telegram bot.
2. Use any model with built in multimodality as your LLM, preferably on Openrouter (will be important for step 5).
3. Enable multimodality, and send two images to the LLM.
4. The agent would show two ellipsis ("...") depending on how many messages you send. Note how the first message may describe or talk about the first image, and the second message may or may not describe both, but will still output as a message as well.
5. (optional) Check Openrouter logs, and notice how the first request has the first image, while the second has both the first and second image.

Impact

Affected users: Telegram users. It's more of a QoL thing, since you could just send images one by one, but who's got the time? This may also compound in API costs, since if you send 10 images for example, your agent would have processed 55 images. Logs may differ from how each model processes the images, but the same outcome will happen.

Logs / stack traces

ZeroClaw version

v0.6.9

Rust version

rustc 1.94.1

Operating system

Ubuntu 24.04.3 LTS

Regression?

Unknown

Pre-flight checks

  • I reproduced this on the latest master branch or latest release.
  • I redacted secrets, tokens, and personal data from all submitted content.

Contributor guide

Tech stack
rust
Domain
backend
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-2 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
blocked
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
clear
Prerequisites
Rust programmingUnderstanding of agent request flowFamiliarity with Telegram bot API
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
30
Research direction
Examine the gateway code in the runtime/daemon component that handles incoming Telegram messages with multiple images. Trace how images are batched into LLM requests and ensure that all images are grouped into a single request, respecting the max images limit in config.toml. Check the request building logic to avoid appending images from previous requests. Look for any associated tests or comments that might clarify expected behavior.