Repository navigation
Conversation
…estart-safety-and-replay.md)
…tive/scripts/benchmark.js)
…tive/src/evaluate.js)
…tive/src/store.js)
…tive/test/persistence.test.js)
…ows/proactive.yml)
…roactive-two-loop-runtime.md)
…tive/package.json)
…tive/src/index.js)
…tive/test/runtime.test.js)
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds @ruvbot/proactive alongside the existing chatbot, turning Build Your Own Duet guidance into three ADRs and executable contracts.
Attention: bounded polling, source-qualified deduplication, persistent cursors, severity floors, stale-event filtering and quiet-hour deferral. P0 and imminent deadlines still alert.
Work: immutable expiring approval packets; edits revoke prior approval; host permissions and outbound checks gate dispatch. Optional private single-writer snapshots persist executing receipts before dispatch. Interrupted actions recover as uncertain and cannot automatically retry. Storage faults halt mutation. Stale locks require explicit operator reconciliation.
MetaHarness: pinned npm kernel 0.1.3, truthful backend reporting, separate reproducible 86-byte WASM promotion predicate, proposal-only learning and labelled replay evaluation.
Validation: 23 tests pass locally and GitHub CI passes on Node 20, 22 and 24. CI verifies WASM rebuild parity, tests, synthetic benchmark, demo, doctor and npm packaging. Tests include actual process exit during dispatch, corrupt approval recovery rejection, storage failure and concurrent execution. Fresh tarball install exposes the new store and evaluation modules. The original production dependency audit reported zero vulnerabilities.
Synthetic 1000-scenario replay (100 critical, 900 routine) returned 100% encoded-rule recall/precision/accuracy in approximately 13 ms locally. This is regression evidence, not real-world quality evidence.
Production boundaries: no deployed scheduler, chatbot UI, live connectors, authenticated approval API, durable notification outbox or distributed execution. Snapshots contain plaintext action payloads and require protected host storage. MetaHarness's published WASM artifact is absent, so doctor reports JS; ruvbot's own WASM predicate executes in tests. No publication, production activation or merge performed.