https://docs.all-hands.dev/ logo
Join Slack
Powered by
Can OpenHands (via Web UI) analyze the content of ...
# general
b
Can OpenHands (via Web UI) analyze the content of an image? I'm using a vision model and set OPENHANDS_FORCE_VISION=1
Does this mean that the image content will be appended to the current context? The correct way to read images (tested with Hermes) is to start a separate LLM inference request and only inject the resulting text into the conversation context.
I think I will have to tell OH to access the endpoint directly.
c
We send the request in-context to the main LLM. I think we may want to look into sending it to a different LLM when the main LLM doesn’t support images.
b
🫢😊
it was super easy to create the request with OH, but you have to provide the LLM credentials
c
Left a little comment
b
which one? I only see the solved review
👀 1
c
Sorry, the agent solved, I unsolved. Mainly my overall concern is for the code design to be understandable, and that it works for all it needs to work
b
Yes. I made multiple corrections
And the code got cleaner in upstream
(
inferKindFromHost()
is not called anymore in the OH backend worfklow, so the commented change was reverted)
and I also added support for all endpoint arguments
7 Views
Open in Slack
PreviousNext
Powered by