Skip to content

chore(graphify): refresh knowledge graphs - #944

Closed
allyblockcast[bot] wants to merge 1 commit into
masterfrom
bot/graphify-reindex
Closed

chore(graphify): refresh knowledge graphs#944
allyblockcast[bot] wants to merge 1 commit into
masterfrom
bot/graphify-reindex

Conversation

@allyblockcast

@allyblockcast allyblockcast Bot commented Aug 2, 2026

Copy link
Copy Markdown

Automated knowledge-graph refresh from the graphify-reindex CronJob.

AST-only rebuild (graphify update), zero LLM tokens. Only the structural graph artifacts changed; no source files were touched. Safe to merge once CI is green.

@allyblockcast

allyblockcast Bot commented Aug 2, 2026

Copy link
Copy Markdown
Author

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: c69350d

Important Issues (1)

  • [pr-review-toolkit / gstack/review / native-codex] server/src/graphify-out/GRAPH_REPORT.md:462 — The refreshed report claims 458 communities with 46 communities smaller than three nodes omitted, but the exact-head graph has only 33 communities smaller than three nodes and the report renders only 367 community sections. This drops substantive communities: Community 150 has nine nodes, Community 446 has four, and Community 454 has three, yet none has a detailed section. The navigation list and summary counts also disagree (413 links versus “412 shown”), so consumers cannot trust this generated report as a faithful view of graph.json.
    • Regenerate with corrected report logic and add an integrity check reconciling community sizes, navigation links, rendered sections, and the summary before committing generated output.

Strengths

  • Both JSON artifacts parse successfully.
  • All 8,999 node IDs are unique, all 18,798 edges resolve to existing nodes, and every graph community has a matching label.
  • The artifact provenance correctly records the source snapshot cce01e9942404569120c9b875166c549a7700902; the PR commit is the generated-output commit layered on that snapshot.

Recommended Action

  1. Address the Important issue this cycle before merge.

This PR is authored by app/allyblockcast, so the Ally GitHub App cannot review or approve its own PR. Reopen this exact head under an independent author before requesting the App approval required by review/ally-complete.

@kkroo
kkroo force-pushed the bot/graphify-reindex branch from c69350d to 9e03377 Compare August 4, 2026 23:34
@allyblockcast

allyblockcast Bot commented Aug 5, 2026

Copy link
Copy Markdown
Author

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: 9e03377

Prior Findings Dispositioned (1)

  • prior:c69350d important 1 — still-present — server/src/graphify-out/GRAPH_REPORT.md:462 — The exact-head graph has 458 communities, but only 33 contain fewer than three nodes and 425 are therefore reportable. The report still claims 46 thin omissions, links 413 communities, and renders only 367 community sections; communities 150 (9 nodes), 446 (4), and 454 (3) remain missing.

Important Issues (1)

  • [prior:c69350d important 1 / pr-review-toolkit / gstack/review / native-codex] server/src/graphify-out/GRAPH_REPORT.md:462 — The generated report remains internally inconsistent with graph.json, so its community summary, navigation, and detailed sections are not a faithful view of the graph.
    • Correct the report-generation logic, regenerate the artifacts, and add an integrity check that reconciles graph community sizes, summary counts, navigation links, and rendered sections.

Strengths

  • Both JSON artifacts parse successfully.
  • All 8,999 node IDs are unique, all 18,798 edge endpoints resolve, and all 458 graph communities have exactly one label entry.
  • The graph provenance records the full source snapshot cce01e9942404569120c9b875166c549a7700902.

Recommended Action

  1. Address the Important issue before merge.

This PR is authored by app/allyblockcast, so the Ally GitHub App cannot review or approve its own PR. Reopen this exact head under an independent author before requesting the App approval required by review/ally-complete.

@kkroo kkroo left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed generated graphify refresh. Rebuilt server/src graph on current base, preserved directed graph metadata, and verified the staleness predicate locally against origin/master.

@kkroo
kkroo enabled auto-merge August 5, 2026 12:29
@kkroo
kkroo added this pull request to the merge queue Aug 5, 2026
@kkroo
kkroo removed this pull request from the merge queue due to a manual request Aug 6, 2026
@kkroo
kkroo force-pushed the bot/graphify-reindex branch from 0828bf0 to db6cf82 Compare August 6, 2026 02:18
@kkroo
kkroo added this pull request to the merge queue Aug 6, 2026
@allyblockcast

allyblockcast Bot commented Aug 6, 2026

Copy link
Copy Markdown
Author

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: db6cf82

Prior Findings Dispositioned (1)

  • prior:c69350d important 1 — still-present — server/src/graphify-out/GRAPH_REPORT.md:459 — The exact-head graph contains 452 communities: 416 have at least three nodes and only 36 are thin. The report instead claims 412 shown and 40 thin omitted, links only 410 communities, and renders only 370 community sections; reportable communities such as 92 (21 nodes), 163 (14), 246 (8), and 424 (5) are still absent.

Critical Issues (0)

Important Issues (2)

  • [prior:c69350d important 1 / pr-review-toolkit / gstack/review / native-codex] server/src/graphify-out/GRAPH_REPORT.md:459 — The generated report remains internally inconsistent with graph.json, so its summary, navigation, and detailed sections are not a faithful view of the graph. Correct the report-generation logic, regenerate the artifacts, and add an integrity check that reconciles graph community sizes, summary counts, navigation links, and rendered sections.
  • [gstack/review / native-codex] server/src/graphify-out/graph.json:292854 — This refresh was built from 64e0d44ce022bf736a6c36de406ea0d1e28853ae, which is 12 commits behind the PR base da92fbf6dc830e28e9ed267e16e98e2295623b24; meanwhile GRAPH_REPORT.md:13 claims it was built from unrelated artifact commit 9e033770. The PR therefore omits graph changes from the current base and provides contradictory freshness metadata. Rebuild from the exact current base and ensure both artifacts record the same full source SHA.

Strengths

  • Both JSON artifacts parse successfully, all 9,141 node IDs are unique, all 19,059 edge endpoints resolve, and every graph community has a label.
  • Edge confidence metadata is well-formed and the report's node, edge, and inferred-edge totals match graph.json.
  • Current CI checks are green.

Recommended Action

  1. Fix both Important issues before merge.
  2. Re-run an exact-head artifact integrity check after regeneration.

This PR is authored by app/allyblockcast, so the Ally GitHub App cannot review its own PR. The exact head must be reopened under an independent author before an App approval is possible; the shared User credential is not a substitute for the gate-authorizing App review.

@kkroo
kkroo removed this pull request from the merge queue due to a manual request Aug 6, 2026
kkroo added a commit that referenced this pull request Aug 8, 2026
…bution audit

Addresses the three Important findings on #1097.

1. pr.yml: drop the `github.head_ref != 'bot/graphify-reindex'` exemption
   entirely rather than hardening it. Measurement, not assumption: the
   graphify-reindex bot commits via `git push` as
   `allyblockcast[bot]@users.noreply.github.com`, which is NOT the REST
   write-path stamp this gate matches
   (`290875700+allyblockcast[bot]@users.noreply.github.com`). Running the
   gate over the real graphify PRs #789 and #944 passes both, unmodified.
   The exemption was guarding a rejection that never happens — while being
   a fork bypass (`head_ref` carries no repository identity, so any fork
   can claim the branch name) and a merge-queue false-reject (`head_ref`
   is empty on `merge_group`). Removing it fixes both halves at once and
   restores gate coverage in the merge queue. Pinned by a regression test
   on the bot's real identity.

2. Audit window: `gh pr list --state merged` orders by creation time, so
   the "last N merged PRs" were not the last N merged. Over-fetch and sort
   by `mergedAt` locally (selectRecentlyMergedPrs), with the residual
   window limitation documented on AUDIT_OVERFETCH_FACTOR.

3. Commit-list truncation: `GET /pulls/{n}/commits` hard-caps at 250 even
   with --paginate, so an offense past that point read as a pass. Detect
   and fail closed, reported as INCOMPLETE rather than VIOLATION so the
   two outcomes stay distinguishable.

Verified: 15/15 unit tests pass; both new guards confirmed by negative
control (reverting each fails exactly its own tests); pr.yml re-parsed with
ruby -ryaml; live --audit-merged exits 1 on a real in-the-wild violation
(#1104 bb6fefe) while both graphify PRs pass.
@allyblockcast
allyblockcast Bot force-pushed the bot/graphify-reindex branch from db6cf82 to 09d69ca Compare August 9, 2026 05:56

@allyblockcast allyblockcast Bot left a comment

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: 09d69ca

Looks good. This generated-artifact refresh is internally consistent: the report build commit matches the reviewed head parent, and the graph contains 9,343 nodes and 19,415 links with labels for every referenced community.

Critical Issues (0)

Important Issues (0)

Suggestions (0)

Strengths

  • Limits the change to generated knowledge-graph artifacts.
  • Keeps the report counts, graph structure, and community-label mapping consistent.

Recommended Action

  1. Merge once required checks complete.

kkroo pushed a commit that referenced this pull request Aug 10, 2026
…-23894)

CTO measured that 11 of 71 sampled checkouts under /paperclip/work had
local git config already stamped with the shared allyblockcast[bot] App
identity (and 18 more had no local identity at all), so a git-push commit
can still trip this gate — the failure message and AGENTS.md's forward
guidance both asserted the opposite. Widening the match to the bare
allyblockcast[bot]@users.noreply.github.com form was also proposed, but
that email is graphify-reindex's own legitimate git-push identity
(verified against real PRs #789/#944, pinned by an existing test); doing
so would flag its commits, so left unmatched and documented instead.
@kkroo
kkroo added this pull request to the merge queue Aug 11, 2026
@kkroo
kkroo removed this pull request from the merge queue due to a manual request Aug 11, 2026
@kkroo

kkroo commented Aug 11, 2026

Copy link
Copy Markdown

Dequeued from the merge queue because this generated graph snapshot was built from base 48d51211f5f8550f03d15e79704b4728bf9b6167, far behind current master (1eab22741e87db5646598df64e3848510645fd1e). Regenerate from current master and obtain exact-head reviews before requeueing; landing this snapshot would overwrite newer graph state.

@kkroo

kkroo commented Aug 11, 2026

Copy link
Copy Markdown

Superseded by #1215, merged as 7441c03 (chore(graphify): refresh from current master). That commit regenerated the same three graphify outputs from a newer master snapshot; rebuilding or merging this older 48d5121 snapshot would regress generated graph data.

@kkroo kkroo closed this Aug 11, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant