FE-743: Petri parallel execution with resource pools and per-slice sandboxes by kostandinang · Pull Request #149 · hashintel/brunch

kostandinang · 2026-05-22T12:38:17Z

Summary

Adds FiringPolicy (serial | parallel) to the Petri interpreter: parallel mode claims all ready transitions and fires them concurrently with Promise.allSettled, subject to pool limits.
Introduces shared agent resource places so test-writer and code-writer transitions contend under a configurable global concurrency cap.
Isolates each slice in its own sandbox directory under the run tree, which parallel execution and later epic merge depend on.

Context

FE-738 separated mechanical and semantic completion lanes; this PR delivers the main structural advantage of Petri over proc: independent slices can advance concurrently when policy and pools allow.
Planning decision gate: parallel policy must demonstrate measurable benefit on a multi-slice fixture relative to serial before the frontier is considered complete.
Renaming worktreeDir to sandboxDir clarifies that the path is a per-run filesystem root, not necessarily a git worktree (brownfield worktrees land in FE-755).

What changed

FiringPolicy: serial preserves prior single-fire behavior; parallel batches independent ready transitions.
RunPolicy.agentPoolSize: caps simultaneous agent-backed transitions via pool:test-agent and pool:code-agent places.
Per-slice sandboxes: artifacts live under join(sandboxDir, sliceId) so concurrent slices do not stomp the same tree.
CLI: removes dead --engine flag; adds --policy=serial|parallel for explicit policy selection at cook time.

Verification

Engine contract scenarios exercised under both policies (41 contract tests in this area).
Dedicated tests for pool bounds, concurrent firing behavior, and sandbox isolation between slices.
npm run verify green.

Out of scope

Merging slice sandboxes for epic-level verify-epic (FE-745).
Codebase/brownfield cook on real repos (FE-755).

Traceability

Requirements 46–50; SPEC §3, §4; frontier petri-parallel-execution in memory/PLAN.md; umbrella H-6476.

cursor · 2026-05-22T12:38:23Z

PR Summary

Medium Risk
Touches core orchestration concurrency, token claiming, and filesystem layout; mistakes could cause races, partial net state, or cross-slice artifact clashes, though contract tests cover both policies.

Overview
This PR completes Petri Phase 2 (FE-743) by making the orchestrator actually run independent slices in parallel instead of only serially.

The interpreter gains a parallel FiringPolicy: each step claims tokens for every enabled transition and fires them with Promise.allSettled, rolling the batch back on handler failure. Agent capacity moves from per-slice places into shared pool:test-agent / pool:code-agent places, capped by new RunPolicy.agentPoolSize (default: one token per slice). Filesystem isolation renames the run root to sandboxDir, creates sandboxDir/<sliceId> for slice actions and tests, and keeps verify-epic on the parent sandbox (with a noted follow-up to merge slice dirs for epic verification).

CLI exposes --policy=serial|parallel (replacing the unused --engine help text in server/cli.ts). Contract tests now run under both policies and add concurrency, pool-bound, wall-clock, and per-slice sandbox assertions. memory/PLAN.md marks petri-parallel-execution done and records open Phase 3 constraints (declarative output arcs, epic sandbox merge).

^{Reviewed by Cursor Bugbot for commit 421f18b. Bugbot is set up for automated code reviews on this repo. Configure here.}

kostandinang · 2026-05-22T12:38:30Z

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

This stack of pull requests is managed by Graphite. Learn more about stacking.

augmentcode · 2026-05-22T12:42:31Z

🤖 Augment PR Summary

Summary: This PR completes FE-743 by adding true parallel Petri-net execution, shared agent resource pools, and per-slice worktree isolation to reduce wall-clock time on multi-slice plans.

Changes:

Extend FiringPolicy to serial | parallel and implement runParallel() using greedy token claiming + Promise.allSettled.
Add shared pool:test-agent/pool:code-agent places and RunPolicy.agentPoolSize to cap global agent concurrency.
Route action contexts and test runs through join(worktreeDir, sliceId) to isolate slice execution directories.
Update brunch cook CLI: retire --engine, add --policy=serial|parallel, default to serial.
Expand contract tests to cover both policies and add explicit concurrency/pool/worktree isolation checks.
Update memory/PLAN.md to mark petri-parallel-execution done and adjust sequencing.

Technical Notes: Parallel execution aims for serial parity while allowing bounded concurrency via shared pool tokens.

_{🤖 Was this summary useful? React with 👍 or 👎}

augmentcode

Review completed. 2 suggestions posted.

Comment augment review to trigger a new review at any time.

Co-authored-by: Amp <amp@ampcode.com>

Parallel firing policy (petri-net.ts): - FiringPolicy = 'serial' | 'parallel' - runParallel: greedy token claiming + Promise.allSettled concurrent fire - Extracted isEnabled() private helper Shared resource pools (net-compiler.ts): - pool:test-agent / pool:code-agent replace per-slice agent places - agentPoolSize on RunPolicy bounds global concurrency - Worktree-per-slice: join(worktreeDir, sliceId) for all action contexts CLI (cook-cli.ts): - Retired dead --engine=proc|petri flag - Added --policy=serial|parallel wired to createOrchestrator Tests (engine-contract.test.ts): - Parallel added to all engines arrays (serial parity) - Concurrency proof, wall-clock benchmark, pool-bounded tests - Extracted withConcurrencyTracking() helper - Worktree isolation adapter test Decision gate passed: parallel measurably beats serial on wall clock. Co-authored-by: Amp <amp@ampcode.com>

Deposit fulfilled transition outputs even when a sibling sets ctx.halted in the same batch, matching serial commit semantics while still rolling back on handler rejection. Co-authored-by: Cursor <cursoragent@cursor.com>

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 421f18b. Configure here.}

cursor · 2026-05-27T23:37:17Z

  return placeId;
 }

+type TransitionClaim = { transition: TransitionDef; consumed: Token[] };


Stale benign-residual entries mask removed per-slice places

Low Severity

BENIGN_RESIDUAL_PLACES still contains 'test-agent' and 'code-agent', but this PR removed per-slice agent places (e.g. slice:X:test-agent) and replaced them with shared pool:test-agent / pool:code-agent places. Pool places are already excluded by the startsWith('pool:') check in hasWorkBearingTokens, so the set entries are now dead code that can never match any place in the compiled net. Keeping them is confusing for anyone reading the deadlock-detection logic, as it implies per-slice agent places still exist.

^{Reviewed by Cursor Bugbot for commit 421f18b. Configure here.}

kostandinang mentioned this pull request May 22, 2026

FE-738: Petri semantic lanes with two-lane subnet and engine factory #148

Open

kostandinang mentioned this pull request May 22, 2026

FE-730: Greenfield orchestrator POC with pi-agent dispatch #143

Open

kostandinang changed the title ~~plan: petri-parallel-execution in-progress, FE-743 assigned~~ FE-743: Petri parallel execution — concurrent firing, resource pools, worktree-per-slice May 22, 2026