Release v0.8.0#167
Merged
conradbzura merged 15 commits intomasterfrom Apr 7, 2026
Merged
Conversation
Add BackpressureLike protocol and BackpressureContext dataclass that let users configure a per-worker hook to reject incoming tasks. The hook receives a snapshot of active task count and the incoming task, returning True to reject (triggering gRPC RESOURCE_EXHAUSTED) or False to accept. The load balancer already treats RESOURCE_EXHAUSTED as transient and skips to the next worker. The hook is threaded through LocalWorker -> WorkerProcess -> WorkerService, serialized via cloudpickle across the spawn boundary to support lambdas and closures.
Each SDLC skill now checks for a knowledge graph at .understand-anything/knowledge-graph.json and invokes /understand-chat as a sub-step during context gathering. The graph query logic stays in /understand-chat — no duplication across skills. Skills degrade gracefully when no graph exists.
Cover BackpressureContext (instantiation, frozen immutability, equality, Hypothesis property test), BackpressureLike protocol (sync, async, callable class, negative case), WorkerService dispatch with backpressure (sync/async accept and reject, context field validation, active task counting), and LocalWorker construction with the hook. Update public API completeness test for the new exports.
LocalWorker, WorkerProcess, and WorkerService were missing docstring entries for parameters introduced by the backpressure feature. WorkerProcess was also missing uid, tags, and extra params that predated this PR. Removes a stale :param tasks: entry from WorkerService._cancel which takes no arguments.
Extends the pairwise covering array and Hypothesis strategy with a BackpressureMode dimension (NONE, SYNC, ASYNC) to verify that backpressure hooks survive cloudpickle serialization through real subprocesses. Adds rejection composition tests validating the end-to-end RESOURCE_EXHAUSTED path and load balancer fallback to an accepting worker.
Each SDLC pipeline skill now executes inside a dedicated subagent rather than inline in the parent conversation context. Dispatcher files in llms/dispatchers/ spawn a general-purpose subagent that reads the corresponding skill definition from llms/skills/ and executes its workflow, returning a structured summary to the parent. This keeps the parent context clean for multi-step pipeline sessions.
Add a detect-changes job that checks out the repo and invokes the existing get-touched-files composite action to determine whether source code, tests, build config, protos, or CI files were modified. The lint, unit-tests, and integration-tests jobs now depend on this gate and only run when has_code_changes is true, avoiding wasted CI minutes on documentation-only or skill-definition-only PRs.
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.
Auto-generated by the cut release workflow.