feat(agent): introduce generalized reverse proxy bidirectional streaming evaluation - #575
Draft
saurabh-net wants to merge 3 commits into
Draft
feat(agent): introduce generalized reverse proxy bidirectional streaming evaluation#575saurabh-net wants to merge 3 commits into
saurabh-net wants to merge 3 commits into
Conversation
Implements generalized reverse proxy bidirectional streaming architecture for autonomous multi-turn agent evaluation in Evalbench. Key Changes: - evalproto/eval_agent.proto: Defined typed AgentStreamMessage envelopes for turn requests/responses, in-flight remote scoring, remote workspace archival, and session summary delivery. - evalproto/eval_service.proto: Added rpc AgentInteract streaming endpoint. - evalbench/eval_service.py: Implemented AgentInteract servicer bridging streaming clients with in-memory AGENT_PROXY_QUEUES. - evalbench/generators/models/agentic_reverse_proxy.py: Added AgenticReverseProxyGenerator implementing AgentCliGenerator. - evalbench/scorers/remote_scorer.py: Added RemoteScorerProxy and transparent `remote: true` delegation in score.py. - evalbench/reporting/remote_artifact_reporter.py: Added RemoteArtifactReporter and transparent `reporting.gcs.remote: true` delegation. - evalbench/test/agentic_reverse_proxy_test.py: Added comprehensive unit tests.
saurabh-net
requested review from
IsmailMehdi,
helloeve and
prernakakkar-google
as code owners
August 18, 2026 22:30
saurabh-net
marked this pull request as draft
August 18, 2026 22:30
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
This PR introduces generalized reverse proxy bidirectional streaming evaluation into Evalbench, enabling multi-turn autonomous agents running in remote/sandboxed environments to connect via an outbound gRPC stream (
rpc AgentInteract).Key Features
evalproto/eval_agent.proto):TurnRequest/TurnResponse: Carries prompts, multi-turn tool call trajectories (tool_name,parameters_json,output,duration_ms), token statistics, and status codes.ScoringRequest/ScoringResponse: Generic container for named in-flight remote scorers.ArtifactRequest/ArtifactResponse: Generic container for remote workspace archival to GCS.SessionSummaryMessage: Delivers final job score rollups to the connected client upon run completion.evalproto/eval_service.proto&eval_service.py):rpc AgentInteract(stream AgentStreamMessage) returns (stream AgentStreamMessage).AGENT_PROXY_QUEUESand correlation inboxes.AgenticReverseProxyGenerator(evalbench/generators/models/agentic_reverse_proxy.py):AgentCliGeneratorto execute multi-turn agent conversations over the reverse stream.evalbench/scorers/score.py):remote: trueis automatically routed viaRemoteScorerProxyacross the reverse queue.evalbench/reporting/__init__.py):remote: truedelegates workspace archival to the client overArtifactRequest.evalbench/test/agentic_reverse_proxy_test.py):Testing
python -m unittest evalbench/test/agentic_reverse_proxy_test.pypasses (3/3 tests).