Add screenshot demo for multimodal LLM analytics testing - #37
Merged
Merged
Conversation
Adds a new tool to demonstrate how screenshot/image data appears in PostHog's LLM Analytics traces and generations view. Useful for testing computer-use style products that send screenshots with AI requests. Features: - Sends screenshots with chat messages to Anthropic Claude and OpenAI GPT-4o - Includes sample UI screenshots for testing - Randomly selects from sample images - Supports tool calls (click, type, scroll) to simulate computer-use - Enables _INTERNAL_LLMA_MULTIMODAL to preserve base64 image data
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
screenshot-demotool to demonstrate how screenshot/image data appears in PostHog's LLM AnalyticsFeatures
_INTERNAL_LLMA_MULTIMODALto preserve base64 image data in PostHogUsage
Test plan
./run.sh --tools- verified both Anthropic and OpenAI successfully analyze screenshots_INTERNAL_LLMA_MULTIMODAL=true)