Repository navigation
agentplane: give sampler Runner execution its own package - #16230
Open
agentydragon wants to merge 1 commit into
Open
agentydragon wants to merge 1 commit into
agentydragon wants to merge 1 commit into
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Move the concrete Runner execution adapter and its tests into
internal/executor: session handshake/rebind, incarnation fencing, process/file operations, output cursors, reconnection and outcome evidence stay together. The runtime consumes the concrete Executor API rather than reaching into its fields.The existing caller-visible tool failure becomes
toolcontract.Rejection, shared with Code Mode input validation. Executor-ownedToolOutcomeErrorretains remote-outcome evidence and unwraps to the same caller-visible classification, including when effects are unknown. Image evidence keeps digest redaction. No replay, cursor, timeout, dispatch, auth or process-lifecycle policy changes.Runtime drain tests now use a loopback generated Executor gRPC server and the real constructor instead of importing private fake-client helpers or mutating Executor fields. Existing drain gates and ownership-release assertions remain. Tests and their PostgreSQL setup follow the behavior they verify; no production test hooks or generic port are added.
Validation: Executor tests, remaining sampler integration/unit targets and both command targets compile on RBE. Formatting, Gazelle and commit hooks pass. Executor and sampler unit suites pass on the extraction branch. Its older-base sampler integration run encountered the known Store request-read fixture omission already corrected in #16228/#16229; the combined final runtime/RPC suites pass in #16232. Hosted CI checks the published head separately.
Proposed, ready successor to #16229 in native stack #16208. RPC and Executor are independently specified after Store; this is their review order. The final resident-runtime package move follows separately.
Test/binary compilation: https://www.buildbuddy.io/invocation/62624ae1-ad44-4c13-9e34-ed129ab279ae
Executor and unit suites (passed; older-base integration fixture failure described above): https://www.buildbuddy.io/invocation/12f0534d-deaa-4d0e-b39a-3ac594c00ded
Parent review: all eight production-file diffs, remote-outcome unwrapping, image evidence and real transport fixture; 141 SQL literals unchanged, all 290 existing test functions retained and all test files registered.
Review commit:
890f935a90.Combined current runtime/RPC suites: https://www.buildbuddy.io/invocation/d954c215-8bdb-4ee5-9238-2a6127ac7350