Repository navigation
Conversation
ADK 2 streams every partial chunk of a model response under the id of its consolidated final event. The assistant projector deduplicated all events by id, so every chunk after the first and the final event were dropped and Studio showed no reply. Skip deduplication for partial events that carry content; complete events and content-free partials (A2A status) still dedupe, keeping the replay guard from volcengine#1156.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Fix Studio dropping streamed assistant replies from ADK 2 backends.
ADK 2.7 gives every partial chunk of one model response the same id as its
consolidated final event (
google/adk/flows/llm_flows/base_llm_flow.py:"Partial chunks of one streaming response share the base id"). Since #1156,
AssistantEventProjector.project()ignores any event whose id was alreadyseen, so with
/run_ssestreaming: trueonly the first chunk renders; theremaining chunks and the final event are discarded and Studio reports no
displayable model reply.
Observed against an AgentKit Runtime (
AgentkitAgentServerApp, veadk 1.1.17,google-adk 2.7.1): HTTP 200, 55 SSE frames sharing one id with a complete
answer, while Studio kept only the first thinking chunk. Affects 1.1.15–1.1.20.
Changes
frontend/src/blocks.ts: skip id deduplication for partial events thatcarry content (streamed chunks). Complete events and content-free partials
(e.g. A2A status) are still deduplicated, preserving the feat(studio): integrate managed MPA runtime lifecycle #1156 replay guard.
replaces the preview, and a replayed final event is ignored.
veadk/webui(Node 22.17.0).Validation
node --test tests/parallelStreamAggregation.test.mjs(new case fails before the fix)npm testnpm run buildnpm run test:webui-assetsRisk
Low. Only the projector's dedup condition changes. A replayed partial content
chunk (e.g. after a reconnect) may briefly duplicate preview text until the
consolidated event replaces it.