[PR #6929] [MERGED] fix(environment): bound create learning and invalidate stale write proofs #7454

New issue

Closed

opened 2026-11-01 21:47:58 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-11-01 21:47:58 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/morluto/rea/pull/6929
Author: @earakely-scale
Created: 10/24/2026
Status: ✅ Merged
Merged: 10/24/2026
Merged by: @gensokyo-dev

Base: main ← Head: feat/environment-build

📝 Commits (10+)

da42a23 feat(environment): report what an environment serves with an environment-coverage view
cc290d9 feat(environment): build an environment bundle in one command
045f14a test(environment): cover environment-build and the coverage view
cf7eb70 test(environment): build the real-agentenv lane's bundle with environment-build
1e2c12e docs(environment): document environment-build and the coverage view
093af31 fix(environment): harden environment-build after review
a040f01 docs(environment): state the build's conditions and record its verification
bc35ef3 fix(environment): dispatch a pending interrupt before each build stage check
62368bd Merge remote-tracking branch 'origin/main' into eb
212c998 fix(environment): scan scenario credentials, own the dry-run tarball and serialize Evidence once

📊 Changes

21 files changed (+683 additions, -141 deletions)

View changed files

📝 docs/agent-environments.md (+4 -2)
📝 docs/specs/agent-environments/environment-build/spec.md (+3 -3)
📝 docs/specs/agent-environments/environment-build/verification.md (+11 -0)
📝 docs/specs/agent-environments/write-rules/spec.md (+6 -1)
📝 docs/specs/agent-environments/write-rules/verification.md (+11 -0)
📝 src/application/environment/environmentBuildReport.test.ts (+84 -0)
📝 src/application/environment/environmentStateWriteReview.test.ts (+270 -0)
📝 src/domain/analysisView/EnvironmentStateAccumulator.ts (+14 -10)
📝 src/domain/analysisView/EnvironmentStateExtraction.ts (+9 -3)
📝 src/domain/analysisView/EnvironmentStateWriteSimulation.ts (+89 -25)
📝 src/domain/analysisView/environmentCoverageView.ts (+2 -1)
📝 src/domain/analysisView/environmentStateRules.ts (+5 -2)
📝 src/domain/analysisView/environmentStateWriteBulk.ts (+2 -2)
📝 src/domain/analysisView/environmentStateWriteCreate.ts (+12 -28)
📝 src/domain/analysisView/environmentStateWriteEffects.ts (+4 -0)
📝 src/domain/analysisView/environmentStateWriteExplain.ts (+6 -7)
📝 src/domain/analysisView/environmentStateWriteSorted.ts (+73 -15)
📝 src/domain/analysisView/environmentStateWriteStamps.test.ts (+1 -12)
📝 src/domain/analysisView/environmentStateWriteStamps.ts (+4 -25)
📝 src/domain/analysisView/environmentStateWrites.test.ts (+49 -2)

...and 1 more files

📄 Description

Summary

Create learning buffered future requests to relax guards, although request variation cannot rebuild copied dependent references or state. Withdrawn array-sorting rules also left stale write confirmations or snapshot overrides.

Lease the current parsed request, remove variation lookahead, and keep unsupported dependent values pinned.
Retire inconsistent read views with their confirmations and snapshot overrides. Continue checking later observations for value contradictions without restoring serving proof; distinguish changed values from array order.
Preserve the shared CLI/MCP report and complete retained-event scan already integrated through #6946, with duplicate-content regression coverage.

Verification

Final head: a6c145a21468224a915f73c27af028fbc454a1e5, preserving all author updates. Local: 1,353 source cases; 46 CLI/MCP/process cases with five platform skips; static/typecheck/module boundaries, architecture, docs and packaged build passed. The unchanged SDK/schema contracts passed 16 cases at fc8176694 after regenerating stale test metadata.

Greptile passed; no unresolved review threads. CI passed 18,159 source cases (13 skipped), static/docs/package/Windows-controls builds, Linux native-package baseline and the actual agentenv runner.

Unrelated CI failure: PR shard 2 has the same two Windows-path error-category assertions as main 798450638 shard 2. Normalized failure sections are byte-identical; test blob and PR ownership scope are unchanged. Seven other shards were cancelled by fail-fast and remain unverified in this run. The user explicitly authorized bypassing confirmed unrelated main failures.

Current-head real workflows:

Four pinned public recordings (NocoDB, Ghost, Metabase, Mailpit): generated-bundle HTTP MCP completed 412 SDK calls with shared-engine-equivalent outcomes. Recorded read bodies match in 302/358 comparisons, preserving known recording/model limits. Original third-party apps were not rerun.
Real Chromium → owned SQLite HTTP app → one-command bundle → HTTP MCP: seven successful native response bodies and an atomic missing-target refusal matched an independently reset database.
Actual browser-scenario CLI → build CLI with an untyped URL-encoded request: three credential-name pointers, byte-identical Evidence copy, verified owned-browser cleanup.
Three write regressions fail at frozen 0895c89bc; the later-retired-observation regression fails at fc8176694 and passes at this head.

Post-merge 41cf869949b82bd0f24d425f6557432ae3a63545: rebuilt the actual merged runtime and bundles. All four recordings retained identical write plans and per-record read/write outcomes; generated-bundle HTTP MCP and independent Chromium/SQLite qualification passed again.

Measured workflow costs

Three alternating rounds per input, 15 matched pairs; complete bundles were byte-identical. Evidence workflows take two CLI calls → one; HAR takes three → one. One-CPU medians:

Input	Elapsed seconds	CPU seconds	Peak process-family MiB
NocoDB	6.26 → 3.45	6.06 → 3.28	416.1 → 418.7
Ghost	5.97 → 3.65	5.80 → 3.41	420.8 → 424.4
Metabase	6.37 → 4.15	6.21 → 3.90	430.0 → 429.7
Mailpit	5.56 → 3.24	5.37 → 3.01	413.8 → 416.3
NativeSQLite	8.18 → 3.75	7.91 → 3.43	399.1 → 402.9

Fresh CLI processes; OS caches uncontrolled; persistent-process warm runs unmeasured. Includes orchestration/hashing, excludes install/download/package preparation. These compare equivalent workflows on the same implementation. Payload-exit counters and cleanup verified; final whole-scope counters unavailable. Maximum build-workflow peak: 433.7 MiB; no general memory-saving claim.

cl100k_base estimates: coverage replies 1,892–2,031 tokens versus full state 8,907–64,282, preserving parent/artifact/coverage/unknowns and retained Evidence. Already-redirected workflows' arguments + final stdout increase from 521–591 to 630–986 tokens because build reports include coverage/warnings. No blanket token-saving claim: actual model usage, schemas, prompts, orchestration and retries are outside that estimate. The affected SDK entry separately measures 114,557 tokens (20,088 for name/description/input).

Author readiness

Human owner: @earakely-scale

I reviewed the final diff and relevant existing code, understand the change and its edge cases, and take responsibility for the complete PR, including AI-assisted work.
I checked the evidence above, confirmed it supports the current change, and recorded meaningful validation gaps.
I declare this PR ready for initial maintainer review.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/morluto/rea/pull/6929 **Author:** [@earakely-scale](https://github.com/earakely-scale) **Created:** 10/24/2026 **Status:** ✅ Merged **Merged:** 10/24/2026 **Merged by:** [@gensokyo-dev](https://github.com/gensokyo-dev) **Base:** `main` ← **Head:** `feat/environment-build` --- ### 📝 Commits (10+) - [`da42a23`](https://github.com/morluto/rea/commit/da42a23ceec4b82d836eb298b2bff930bc6c0b5f) feat(environment): report what an environment serves with an environment-coverage view - [`cc290d9`](https://github.com/morluto/rea/commit/cc290d9f7770126dfb7eca74e081309417db5515) feat(environment): build an environment bundle in one command - [`045f14a`](https://github.com/morluto/rea/commit/045f14a3ae907df9878ea060413377b4cd4111de) test(environment): cover environment-build and the coverage view - [`cf7eb70`](https://github.com/morluto/rea/commit/cf7eb709e19f61bc5e1b23a2b905f2c31b2c3305) test(environment): build the real-agentenv lane's bundle with environment-build - [`1e2c12e`](https://github.com/morluto/rea/commit/1e2c12e5c3bda69a952e33f3c4eeddc90d1deb11) docs(environment): document environment-build and the coverage view - [`093af31`](https://github.com/morluto/rea/commit/093af3144853d28fec554b1c0d507c8bca00574b) fix(environment): harden environment-build after review - [`a040f01`](https://github.com/morluto/rea/commit/a040f01e34351cc7de25e2757e580273e27fe674) docs(environment): state the build's conditions and record its verification - [`bc35ef3`](https://github.com/morluto/rea/commit/bc35ef3cc095213ce8adeaf4cabcab3c945aee66) fix(environment): dispatch a pending interrupt before each build stage check - [`62368bd`](https://github.com/morluto/rea/commit/62368bdae95311dccb1d3b538103d17089965c4b) Merge remote-tracking branch 'origin/main' into eb - [`212c998`](https://github.com/morluto/rea/commit/212c998cc067fcea452f89aa6eb5d26aec27abf6) fix(environment): scan scenario credentials, own the dry-run tarball and serialize Evidence once ### 📊 Changes **21 files changed** (+683 additions, -141 deletions) <details> <summary>View changed files</summary> 📝 `docs/agent-environments.md` (+4 -2) 📝 `docs/specs/agent-environments/environment-build/spec.md` (+3 -3) 📝 `docs/specs/agent-environments/environment-build/verification.md` (+11 -0) 📝 `docs/specs/agent-environments/write-rules/spec.md` (+6 -1) 📝 `docs/specs/agent-environments/write-rules/verification.md` (+11 -0) 📝 `src/application/environment/environmentBuildReport.test.ts` (+84 -0) 📝 `src/application/environment/environmentStateWriteReview.test.ts` (+270 -0) 📝 `src/domain/analysisView/EnvironmentStateAccumulator.ts` (+14 -10) 📝 `src/domain/analysisView/EnvironmentStateExtraction.ts` (+9 -3) 📝 `src/domain/analysisView/EnvironmentStateWriteSimulation.ts` (+89 -25) 📝 `src/domain/analysisView/environmentCoverageView.ts` (+2 -1) 📝 `src/domain/analysisView/environmentStateRules.ts` (+5 -2) 📝 `src/domain/analysisView/environmentStateWriteBulk.ts` (+2 -2) 📝 `src/domain/analysisView/environmentStateWriteCreate.ts` (+12 -28) 📝 `src/domain/analysisView/environmentStateWriteEffects.ts` (+4 -0) 📝 `src/domain/analysisView/environmentStateWriteExplain.ts` (+6 -7) 📝 `src/domain/analysisView/environmentStateWriteSorted.ts` (+73 -15) 📝 `src/domain/analysisView/environmentStateWriteStamps.test.ts` (+1 -12) 📝 `src/domain/analysisView/environmentStateWriteStamps.ts` (+4 -25) 📝 `src/domain/analysisView/environmentStateWrites.test.ts` (+49 -2) _...and 1 more files_ </details> ### 📄 Description ## Summary Create learning buffered future requests to relax guards, although request variation cannot rebuild copied dependent references or state. Withdrawn array-sorting rules also left stale write confirmations or snapshot overrides. - Lease the current parsed request, remove variation lookahead, and keep unsupported dependent values pinned. - Retire inconsistent read views with their confirmations and snapshot overrides. Continue checking later observations for value contradictions without restoring serving proof; distinguish changed values from array order. - Preserve the shared CLI/MCP report and complete retained-event scan already integrated through #6946, with duplicate-content regression coverage. ## Verification Final head: `a6c145a21468224a915f73c27af028fbc454a1e5`, preserving all author updates. Local: 1,353 source cases; 46 CLI/MCP/process cases with five platform skips; static/typecheck/module boundaries, architecture, docs and packaged build passed. The unchanged SDK/schema contracts passed 16 cases at `fc8176694` after regenerating stale test metadata. Greptile passed; no unresolved review threads. [CI](https://github.com/morluto/rea/actions/runs/40108214026) passed 18,159 source cases (13 skipped), static/docs/package/Windows-controls builds, Linux native-package baseline and the actual agentenv runner. **Unrelated CI failure:** [PR shard 2](https://github.com/morluto/rea/actions/runs/40108214026/job/120914596815) has the same two Windows-path error-category assertions as [main `798450638` shard 2](https://github.com/morluto/rea/actions/runs/40106164357/job/120911483741). Normalized failure sections are byte-identical; test blob and PR ownership scope are unchanged. Seven other shards were cancelled by fail-fast and remain unverified in this run. The user explicitly authorized bypassing confirmed unrelated main failures. Current-head real workflows: - Four pinned public recordings (NocoDB, Ghost, Metabase, Mailpit): generated-bundle HTTP MCP completed 412 SDK calls with shared-engine-equivalent outcomes. Recorded read bodies match in 302/358 comparisons, preserving known recording/model limits. Original third-party apps were not rerun. - Real Chromium → owned SQLite HTTP app → one-command bundle → HTTP MCP: seven successful native response bodies and an atomic missing-target refusal matched an independently reset database. - Actual browser-scenario CLI → build CLI with an untyped URL-encoded request: three credential-name pointers, byte-identical Evidence copy, verified owned-browser cleanup. - Three write regressions fail at frozen `0895c89bc`; the later-retired-observation regression fails at `fc8176694` and passes at this head. Post-merge `41cf869949b82bd0f24d425f6557432ae3a63545`: rebuilt the actual merged runtime and bundles. All four recordings retained identical write plans and per-record read/write outcomes; generated-bundle HTTP MCP and independent Chromium/SQLite qualification passed again. ## Measured workflow costs Three alternating rounds per input, 15 matched pairs; complete bundles were byte-identical. Evidence workflows take two CLI calls → one; HAR takes three → one. One-CPU medians: | Input | Elapsed seconds | CPU seconds | Peak process-family MiB | | --- | ---: | ---: | ---: | | NocoDB | 6.26 → 3.45 | 6.06 → 3.28 | 416.1 → 418.7 | | Ghost | 5.97 → 3.65 | 5.80 → 3.41 | 420.8 → 424.4 | | Metabase | 6.37 → 4.15 | 6.21 → 3.90 | 430.0 → 429.7 | | Mailpit | 5.56 → 3.24 | 5.37 → 3.01 | 413.8 → 416.3 | | NativeSQLite | 8.18 → 3.75 | 7.91 → 3.43 | 399.1 → 402.9 | Fresh CLI processes; OS caches uncontrolled; persistent-process warm runs unmeasured. Includes orchestration/hashing, excludes install/download/package preparation. These compare equivalent workflows on the same implementation. Payload-exit counters and cleanup verified; final whole-scope counters unavailable. Maximum build-workflow peak: 433.7 MiB; no general memory-saving claim. `cl100k_base` estimates: coverage replies 1,892–2,031 tokens versus full state 8,907–64,282, preserving parent/artifact/coverage/unknowns and retained Evidence. Already-redirected workflows' arguments + final stdout increase from 521–591 to 630–986 tokens because build reports include coverage/warnings. No blanket token-saving claim: actual model usage, schemas, prompts, orchestration and retries are outside that estimate. The affected SDK entry separately measures 114,557 tokens (20,088 for name/description/input). ## Author readiness Human owner: @earakely-scale - [x] I reviewed the final diff and relevant existing code, understand the change and its edge cases, and take responsibility for the complete PR, including AI-assisted work. - [x] I checked the evidence above, confirmed it supports the current change, and recorded meaningful validation gaps. - [x] I declare this PR ready for initial maintainer review. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-11-01 21:47:58 +00:00