camel-ai / crab

CRAB: Cross-environment Agent Benchmark for Multimodal Language Model Agents. https://crab.camel-ai.org/
https://crab.camel-ai.org/
194 stars 28 forks source link

[Feature Request] Add image compression for actions #19

Open dandansamax opened 3 months ago

dandansamax commented 3 months ago

Required prerequisites

Motivation

Current implementation uses uncompressed PNG format image with base64 encoding in the whole process. Each action have to encode and decode image between base64 + PNG and python PIL.Image.

Solution

It's better to only uncompress and serilize the image during network transfer. While in other actions we should keep using PIL.Image format.

Additional context

No response