#how to send images to openclaw via telegram

1 messages · Page 1 of 1 (latest)

obsidian topaz
#

I am using gemma-4-31b-it and openclaw version 5.6

pliant patioBOT
little duneBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

pliant patioBOT
sleek palmBOT
# obsidian topaz I am using gemma-4-31b-it and openclaw version 5.6

For Telegram, you normally don’t need a special command: send the image/photo to the Telegram bot, with your prompt as the caption or a follow-up message. In a group, mention the bot unless you configured requireMention: false.

The two common failure points are:

  1. Telegram access/routing: make sure the sender/group is allowed and the bot can see the message. In groups, Telegram Privacy Mode can hide messages unless the bot is mentioned, privacy is disabled, or the bot is admin. Docs: https://docs.openclaw.ai/channels/telegram

  2. Vision model routing: OpenClaw only passes/analyzes images if the active model is known image-capable, or if you configure agents.defaults.imageModel. With local/OpenAI-compatible providers like LM Studio, you may need to mark the model with input: ["text", "image"] in models.providers.<provider>.models[].

Quick check from your machine:

openclaw infer image describe --file ./photo.jpg --model <provider/model> --json

If that fails, Telegram isn’t the main issue yet; the model/provider is not being treated as image-capable. For Ollama vision models, the docs show configuring agents.defaults.imageModel: https://docs.openclaw.ai/providers/ollama#vision-and-image-description

Also, please paste the exact output of:

sleek palmBOT