Repository navigation
fix(adk-bridge): preserve tool event Agent identities - #267
Merged
krisztianfekete merged 2 commits intoNov 3, 2026
Merged
krisztianfekete merged 2 commits into
krisztianfekete merged 2 commits into
Conversation
erauner12
marked this pull request as ready for review
November 1, 2026 23:38
erauner12
force-pushed
the
fix/preserve-tool-agent-identities
branch
from
November 2, 2026 21:11
1973116 to
35bbe13
Compare
krisztianfekete
approved these changes
Nov 3, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #268.
Trace conversion currently flattens tool calls before the evaluation adapter assigns event authors. That can replace a captured Agent identity with
agent, causing native judges to fail whenapp_detailsuses the captured name.Preserve each tool call/result owner as native
InvocationEvents. Missing metadata receives a names-only fallback without replacing captured declarations, and model-call accounting stays idempotent. The existing flat ADK eval-set/API format keeps its current exports and single-owner adaptation; this is an ADK data-model contract, separate from deprecated producer telemetry. Ambiguous flat multi-Agent evidence fails explicitly.This repair targets upstream
mainindependently. #264 is stacked on it and also targets upstreammainbecause its dependency branch lives in my fork. Merge #267 first; then I will rebase #264 onto updated upstreammainto leave only its tool-definition changes.The replay test now uses the pure-semconv
v2proj_span_only/S2.jsonfixture. A fake-judge regression completes native final-response evaluation both with captured Agent metadata and with names-only enrichment, assertingresult.error is Noneand a complete score. Synthetic tests also cover multiple Agents, correlated pairs, both ADK data variants, unchanged exports and accounting.Local validation: 2,014 model-free unit tests and 29 HTTP integration tests passed; Ruff lint, format and diff checks passed. No live model calls.
The in-memory
intermediate_datavariant changes; callers should use ADK’s variant-neutral readers.AI assistance: non-trivial code and test assistance.