Replace retired models in the AutoGen Studio default gallery - #8310
Open
JingHao-Leon wants to merge 1 commit into
Open
JingHao-Leon wants to merge 1 commit into
JingHao-Leon wants to merge 1 commit into
Conversation
The default gallery shipped two components whose vendors have retired the backing models, so every new Studio user started from failing requests (issue microsoft#8296): - Anthropic model client: claude-3-7-sonnet-20250219 was retired on 2026-02-19. The gallery now uses claude-sonnet-4-6, registered explicitly in the Anthropic model-info registry (with a 200000 token limit; the Claude-4 family was entirely missing from _MODEL_TOKEN_LIMITS, so those models silently fell back to the 100000 default). - Image generation tool: dall-e-3 was shut down on 2026-05-12. The tool now calls gpt-image-1-mini; gpt-image-* always returns base64 data so the response_format parameter is gone, and the size literals move to the sizes gpt-image supports (1024x1024, 1536x1024, 1024x1536, auto). The frontend default_gallery.json is updated in sync (component config, tool description, and the embedded generate_image source).
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
The AutoGen Studio default gallery shipped two components whose vendors have retired the backing models, so every new Studio user started from failing requests (#8296):
claude-3-7-sonnet-20250219— retired on 2026-02-19 per Anthropic's model deprecations page.dall-e-3— shut down on 2026-05-12 per OpenAI's deprecations page.Changes
Anthropic model client (
gallery/builder.py, frontenddefault_gallery.json): the gallery now usesclaude-sonnet-4-6(Anthropic's recommended replacement), with an explicit entry inautogen_ext/models/anthropic/_model_info.py.While verifying, I also found the Claude-4 family was entirely missing from
_MODEL_TOKEN_LIMITS—get_token_limit("claude-sonnet-4-20250514")(andclaude-opus-4-20250514,claude-sonnet-4-6, …) fell back to the 100000 default instead of the true 200000 context window. Those entries are added too.Image generation tool (
gallery/tools/generate_image.py+ embedded source indefault_gallery.json): switched togpt-image-1-mini(a current model —gpt-image-1itself is listed for shutdown on 2026-10-23, so the premium choice would begpt-image-2; happy to switch if maintainers prefer). Parameter changes per the GPT Image API:response_format="b64_json"removed (gpt-image-* always returns base64 data) and the size literals move from DALL-E's512x512/256x256to1024x1024/1536x1024/1024x1536/auto.Notes for reviewers
get_info("claude-sonnet-4-6")already resolved via the fuzzystartswithmatcher in_model_info.get_info, so the reportedValueErrordoesn't reproduce on currentmain— the explicit registry entry (and the token-limit fix) is what this PR contributes on the autogen-ext side. Happy to split that into a separate PR if preferred.default_gallery.jsonis updated in sync with the Python side (component config, tool description, and the embeddedgenerate_imagesource), so frontend and backend galleries stay consistent.Fixes #8296