auth fixes, readme updates - #5
Merged
Merged
Conversation
alex-struk
pushed a commit
that referenced
this pull request
May 14, 2026
…move failing-samples refs, fix synth-typed claims Six targeted updates per user feedback: 1. Removed the "High variance run-to-run" bullet from the E03 deep-dive section and the "Replication check" paragraph below it — the replication-check discussion was distraction from the main story. 2. Removed the "Replication checks" section at the end of the E08 deep-dive (E05 + E03 re-run side-by-side commentary). 3. Removed point #5 of the E08-vs-E03 mechanism breakdown (CU's higher run-to-run variance) — it depended on the replication data the user wanted gone, so the supporting evidence was no longer in the report. 4. Fixed the E04 synth-samples failure-mode prose. The previous version claimed that density wasn't the issue because HR0081 (10) is also fully filled. That's factually wrong — HR0081 (10) is not densely filled; the synth-* cluster is the densest set in the dataset. The rewrite says E04 struggles on dense images and HR0081 (10) doesn't reproduce the problem because it has many blank cells. 5. Removed all pass_rate references throughout the report (headline observations, E03 deep-dive headline, E08 deep-dive bullets, the same-pipeline bake-off table, the failure-mode-samples manual sample (6) bullet, the Reflection's first point). The 0.8 cutoff was arbitrary and not meaningful for downstream consumers. 6. Removed the "E06 reproducibility" sub-section in Appendix A and the entire "Appendix B — Reproducing the comparison report" section. Also removed the dangling "Appendix B has the reproduction commands" line from the introduction. Net diff: −71 / +12. Report is now noticeably shorter and more focused on the actual cross-engine comparison story without the meta-discussion of how the runs were validated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
alex-struk
added a commit
that referenced
this pull request
Jun 23, 2026
…om history, sandbox-escape tests Review findings #3, #4, #5, #6, #7 (PR 6 — dynamic nodes / Deno sandbox), plus #13: - #3: pass --no-remote to deno check + run so static remote imports can't bypass the --allow-net allowlist at module-load time (verified no template uses them). - #4 (cross-cutting): remove the caller x-api-key from the entire workflow/activity input chain (GraphWorkflowInput, ExecutionState, dyn.run input); the platform key is now sourced server-side in the activity from PLATFORM_API_KEY (env added to the temporal-worker manifests). SECURITY: it is no longer persisted in Temporal history. Semantic change: dyn.run callbacks now act as a service account, not the caller. Also resolves #2 (upload-and-Try is now symmetric with /runs). - #5: add sandbox-escape negative tests (read/write/subprocess/ffi/non-allowlisted fetch/remote import). They require the deno-runner up in CI to actually execute. - #6: scope the egress NetworkPolicy DNS rule to the openshift-dns namespace. - #7: dedicated local docker network + documented local-dev egress gap. - #13 (PR 4): collectOutputTopLevelKeys uses the namespace-aware getCtxRootKey, so doc.*/segment.* outputs are snapshotted and restored from cache correctly. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
alex-struk
added a commit
that referenced
this pull request
Jun 23, 2026
…out; start deno-runner in VSCode - #5 escape test: Deno rejects a remote import with "a remote specifier was requested ... but --no-remote is specified" — broaden the stderr assertion to match "remote" (the fix from #3 works; the regex was just too narrow). Verified against the live deno-runner: 14/14 integration tests pass. - graph-workflow.dynamic-nodes beforeAll: raise the createLocal() timeout 30s->180s. The real reason these "couldn't run" was a one-time ~500MB Temporal test-server download exceeding the 30s hook timeout (which also left a stale .downloading lock). With a warm cache the suite passes in ~7s. - .vscode/tasks.json: add "deno-runner: docker up" + health-wait to Dev: prerequisites so the sandbox sidecar starts with the rest of the stack. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
alex-struk
pushed a commit
that referenced
this pull request
Jul 2, 2026
…move failing-samples refs, fix synth-typed claims Six targeted updates per user feedback: 1. Removed the "High variance run-to-run" bullet from the E03 deep-dive section and the "Replication check" paragraph below it — the replication-check discussion was distraction from the main story. 2. Removed the "Replication checks" section at the end of the E08 deep-dive (E05 + E03 re-run side-by-side commentary). 3. Removed point #5 of the E08-vs-E03 mechanism breakdown (CU's higher run-to-run variance) — it depended on the replication data the user wanted gone, so the supporting evidence was no longer in the report. 4. Fixed the E04 synth-samples failure-mode prose. The previous version claimed that density wasn't the issue because HR0081 (10) is also fully filled. That's factually wrong — HR0081 (10) is not densely filled; the synth-* cluster is the densest set in the dataset. The rewrite says E04 struggles on dense images and HR0081 (10) doesn't reproduce the problem because it has many blank cells. 5. Removed all pass_rate references throughout the report (headline observations, E03 deep-dive headline, E08 deep-dive bullets, the same-pipeline bake-off table, the failure-mode-samples manual sample (6) bullet, the Reflection's first point). The 0.8 cutoff was arbitrary and not meaningful for downstream consumers. 6. Removed the "E06 reproducibility" sub-section in Appendix A and the entire "Appendix B — Reproducing the comparison report" section. Also removed the dangling "Appendix B has the reproduction commands" line from the introduction. Net diff: −71 / +12. Report is now noticeably shorter and more focused on the actual cross-engine comparison story without the meta-discussion of how the runs were validated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
alex-struk
pushed a commit
that referenced
this pull request
Jul 2, 2026
…move failing-samples refs, fix synth-typed claims Six targeted updates per user feedback: 1. Removed the "High variance run-to-run" bullet from the E03 deep-dive section and the "Replication check" paragraph below it — the replication-check discussion was distraction from the main story. 2. Removed the "Replication checks" section at the end of the E08 deep-dive (E05 + E03 re-run side-by-side commentary). 3. Removed point #5 of the E08-vs-E03 mechanism breakdown (CU's higher run-to-run variance) — it depended on the replication data the user wanted gone, so the supporting evidence was no longer in the report. 4. Fixed the E04 synth-samples failure-mode prose. The previous version claimed that density wasn't the issue because HR0081 (10) is also fully filled. That's factually wrong — HR0081 (10) is not densely filled; the synth-* cluster is the densest set in the dataset. The rewrite says E04 struggles on dense images and HR0081 (10) doesn't reproduce the problem because it has many blank cells. 5. Removed all pass_rate references throughout the report (headline observations, E03 deep-dive headline, E08 deep-dive bullets, the same-pipeline bake-off table, the failure-mode-samples manual sample (6) bullet, the Reflection's first point). The 0.8 cutoff was arbitrary and not meaningful for downstream consumers. 6. Removed the "E06 reproducibility" sub-section in Appendix A and the entire "Appendix B — Reproducing the comparison report" section. Also removed the dangling "Appendix B has the reproduction commands" line from the introduction. Net diff: −71 / +12. Report is now noticeably shorter and more focused on the actual cross-engine comparison story without the meta-discussion of how the runs were validated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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
1 participant
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.
No description provided.