Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 8 additions & 5 deletions NEXT_TASKS.md
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# Edge-review next tasks

Last reconciled: **2026-11-15** (America/Los_Angeles), against `main` at `abf0002ab1dea4dc7702f28be1fe49983e42ab8f`, merged PRs #128/#120, open PRs #102/#37, issue #127 and retained review evidence. PR #128 is merged and #129 is closed. The current branch implements #127's abstract replay; its source inspection found a separate BactoTraits strain conflict filed as #130. Independent calibration ratings and correction verification remain pending.
Last reconciled: **2026-11-15** (America/Los_Angeles), against `main` at `a4c8ce937ffd791b7c7b4d05b85121dea1aa09d6`, merged PRs #131/#128/#120 and open PRs #135/#102/#37. PR #131 is merged and #127 is closed. #130's bounded source quarantine has been independently verified against regenerated producer bytes in [KG-Microbe PR #1867](https://github.com/Knowledge-Graph-Hub/kg-microbe/pull/1867); merge and graph-release verification remain pending. Independent scientific calibration has not passed.

Scope: the stratified edge review and related edge-quality follow-ups. Update this file when work starts or ships, checking GitHub and current code before changing status. Record completion dates and PRs. Preserve historical manifests, packets, reports and scientific decisions; a merged engineering PR does not promote an edge's scientific status. Unrelated release/deployment work is outside this backlog.

Expand All @@ -23,14 +23,15 @@ Evidence: [fresh-canary report](artifacts/edge-recanary-20261113/RECANARY-REPORT
| Priority / status | Work | Next action and completion evidence |
| --- | --- | --- |
| 1 — **DONE preparation, PR #128** | Pin calibration cases and decision thresholds | Merged 2026-11-15. The [ten-case protocol](docs/plans/edge-calibration-20261115.md) fixes first-cohort 1/28/30/32/43 and fresh-cohort 4/5/14/32/47, exact occurrence bindings and operational error/coverage thresholds before new scoring. [Preparation receipts](artifacts/edge-calibration-20261115/PREPARATION.md) record 10/10 replayed source packets, 20/20 endpoint roles and two staged authorities after separate canaries. #129's cross-case packet binding fix passed a new canary before all ten were replayed again. These receipts pin the PR #128 runtime; replay under changed profile code requires a new round, not updated historical hashes. Calibration and independent agreement have not passed. |
| 1b — **IMPLEMENTED, awaiting PR review/merge, #127** | Replay primary abstract evidence | Metadata-bound Europe PMC abstract replay and the single retained fresh-05 canonical save/readback canary passed. [Validation](docs/validation/abstract-evidence-20261115.md): two quotations, 16 words, partial completion and six conservative assessments. This technical self-review is excluded from the independent calibration matrices; it cannot resolve #130's conflicting strain join. |
| 1b — **DONE, 2026-11-15, PR #131** | Replay primary abstract evidence | Metadata-bound Europe PMC abstract replay and the single retained fresh-05 canonical save/readback canary passed. [Validation](docs/validation/abstract-evidence-20261115.md): two quotations, 16 words, partial completion and six conservative assessments. All 687 tests, two adversarial reviews and post-merge CI passed. This technical self-review is excluded from the independent calibration matrices; it does not resolve #130's conflicting strain join. |
| 1c — **CANDIDATE VERIFIED; merge pending, #130** | Reconcile the mixed BactoTraits strain row | Four primary catalogs confirm the conflict. [Evidence and independent inspection](docs/validation/bactotraits-identity-20261115.md): actual baseline has 21 assertions including the reviewed nonmotility edge; candidate has zero and one mandatory original-row audit. [KG-Microbe PR #1867](https://github.com/Knowledge-Graph-Hub/kg-microbe/pull/1867) is unmerged. Preserve the original occurrence; producer correction before finalization is distinct from graph-release verification. |
| 2 — **PENDING, actionable bounded audit** | Semantic identity, relation and node-category correctness | Prepare pinned authorities and audit the selected calibration cases, then the existing 100 after gates pass. Prioritize [#68](https://github.com/CultureBotAI/EdgeMech/issues/68)/[#106](https://github.com/CultureBotAI/EdgeMech/issues/106) identity joins and [#75](https://github.com/CultureBotAI/EdgeMech/issues/75)/[#86](https://github.com/CultureBotAI/EdgeMech/issues/86)/[#87](https://github.com/CultureBotAI/EdgeMech/issues/87) relation/type-strain/bin semantics. Review nonempty categories, not only missing fields. All 22 fresh-canary biological-literature exclusions still receive identity/modeling review. |
| 3 — **PENDING, actionable retained-evidence work** | DOI identity, exact snippets and categorical claim judgments | Resolve publication identity [#69](https://github.com/CultureBotAI/EdgeMech/issues/69); replay retained evidence for the selected calibration cases first, then retrieve missing relevant text through appropriate channels. Verify literal quotes and independently judge entailment, conditions and organism/strain scope. Record support, refutation, conflicts, insufficient evidence and assessment blockers distinctly. |
| Parallel — **IN PROGRESS** | General full-text credential protection: [#100](https://github.com/CultureBotAI/EdgeMech/issues/100), [PR #102](https://github.com/CultureBotAI/EdgeMech/pull/102) | The draft has a clean independent review and green CI at `2364161819951b7899ba62cebcd461bf0394c4a7`. Reconcile with current main and validate before a separately authorized merge. Complete this before further authenticated general-helper retrieval; it need not block local modeling or retained-evidence review. |
| 4 — **PENDING, gated execution** | Calibrate, then review the retained 100 | After cases, thresholds and evidence preparation are fixed, independently assess the ten cases with all six axes and report disagreements. Canary any newly introduced worker/launcher or evidence-adapter path end to end before using it in a batch. Cover the existing 100 only after calibration passes. Preserve historical bytes; re-review adds coverage depth, not new reviewed occurrences. |
| 5 — **PENDING; upstream fixes/releases where needed** | Correct and verify affected assertions and nodes | File each finding, identify its maintained owner, retain before/after proposals, add regressions, regenerate a new candidate and independently re-assess affected occurrences. Include [#56](https://github.com/CultureBotAI/EdgeMech/issues/56), [#70](https://github.com/CultureBotAI/EdgeMech/issues/70), [#72](https://github.com/CultureBotAI/EdgeMech/issues/72) and proof of existing fixes [#71](https://github.com/CultureBotAI/EdgeMech/issues/71)/[#73](https://github.com/CultureBotAI/EdgeMech/issues/73). A merged source PR is distinct from a verified corrected graph release. |

**Recommended continuation:** finish #127's PR review, reconcile the BactoTraits strain conflict [#130](https://github.com/CultureBotAI/EdgeMech/issues/130), and assemble versioned predicate/node authority and DOI/snippet proofs before freezing a new evidence package for two separate reviewers. Reconcile ontology versions before assessing #75/#87; an old local snapshot cannot stand in for the frozen or maintained version. Prepare #102 in parallel for authenticated retrieval. Reuse the 100 retained occurrences as the first correction-focused cohort; do not expand to the remaining 9,900 until evidence, modeling, correction-verification and reviewer-calibration gates in the plan are met. At least one real finding's correction must be independently verified against regenerated candidate bytes; a dependency explanation cannot substitute for that proof ([#119](https://github.com/CultureBotAI/EdgeMech/issues/119)). Work within the retained 100 can continue while this expansion gate is unmet.
**Recommended continuation:** finish the reviewed #130 correction PRs through separately approved merges, then assemble versioned predicate/node authority and DOI/snippet proofs before freezing a new evidence package for two separate reviewers. Reconcile ontology versions before assessing #75/#87; an old local snapshot cannot stand in for the frozen or maintained version. Prepare #102 in parallel for authenticated retrieval. Reuse the 100 retained occurrences as the first correction-focused cohort; do not expand to the remaining 9,900 until evidence, modeling and reviewer-calibration gates in the plan are met. #130 now supplies the bounded independent candidate-byte correction demonstration required by [#119](https://github.com/CultureBotAI/EdgeMech/issues/119); this does not establish a graph release or satisfy the remaining gates. Work within the retained 100 can continue while expansion remains gated.

## Source, ontology and regression follow-ups

Expand All @@ -39,7 +40,8 @@ All entries below remain open as of the reconciliation date. Local evidence audi
| Issues | What remains | Closure evidence |
| --- | --- | --- |
| [#68](https://github.com/CultureBotAI/EdgeMech/issues/68), [#106](https://github.com/CultureBotAI/EdgeMech/issues/106) | Discordant MicrobeDecoder organism crosswalks, including frozen records 10743 and 20904 | Trace the exact joins, independently resolve identity or prove an existing quarantine, and prevent unrelated NCBI fields from entering the measurement/reference binding. Edge47 currently excludes the discordant fields. |
| [#130](https://github.com/CultureBotAI/EdgeMech/issues/130) | BactoTraits row 8521 lists DSM 19353/JCM 14586 while its BacDive 17851 link and nominal taxon identify DSM 18772 | Independently resolve collection accessions and the maintained BactoTraits identity join, verify any existing fix/quarantine, and test regenerated candidate bytes before claiming correction. Preserve the source conflict; the nominal-strain paper does not reconcile it. |
| [#130](https://github.com/CultureBotAI/EdgeMech/issues/130) | Exact-row quarantine and regenerated producer bytes independently verified; owner merge and graph-release verification remain pending | [Identity investigation](docs/validation/bactotraits-identity-20261115.md), [owner PR #1867](https://github.com/Knowledge-Graph-Hub/kg-microbe/pull/1867). No replacement subject or trait cells are inferred. The partial canonical identity review predates correction validation and remains unchanged. |
| [#132](https://github.com/CultureBotAI/EdgeMech/issues/132) | Neighboring BactoTraits row 8517 also has conflicting collection codes; wider source alignment remains unassessed | Independently resolve the additional accessions and trace complete row/column provenance before bounding or repairing other rows. Do not broaden the single-row quarantine automatically. |
| [#69](https://github.com/CultureBotAI/EdgeMech/issues/69) | Inherited dataset DOIs identify unrelated publications | Verify exact titles/identifiers in the maintained assignment table; prove corrections with a bounded regression before any bulk repair. Database provenance is not biological claim evidence. |
| [#70](https://github.com/CultureBotAI/EdgeMech/issues/70), [#72](https://github.com/CultureBotAI/EdgeMech/issues/72) | Maximum salinity tolerance mapped to halophily; generic sporulation mapped to endospore formation | Establish the correct trait semantics and a source-grounded mapping or explicit exclusion. Preserve range/optimum/tolerance and spore-type distinctions. |
| [#75](https://github.com/CultureBotAI/EdgeMech/issues/75), [#86](https://github.com/CultureBotAI/EdgeMech/issues/86), [#87](https://github.com/CultureBotAI/EdgeMech/issues/87) | Obsolete observation predicate; taxon/type-strain conflation; opaque BactoTraits quantitative classes | Obtain maintained predicate/type-strain representations and auditable bin definitions with bounds, endpoint conventions, units and range-versus-optimum meaning. Preserve API-kit outcome context. Do not invent local ontology definitions. |
Expand All @@ -52,14 +54,15 @@ Related edge-quality work, **PENDING source/release follow-up**:

## LBNL/library service usage and next access check

Usage below is scoped to the two canaries and the edge47 follow-up. The LBNL literature/review skills were used for the follow-up, but successful licensed-library full-text retrieval was not demonstrated in these runs.
Usage below is scoped to the two canaries, the edge47 follow-up and the BactoTraits identity investigation. The LBNL literature/review skills were used, but successful licensed-library full-text retrieval was not demonstrated in these runs.

| Run / service | Observed use and result |
| --- | --- |
| First 50 | 103 public Europe PMC requests, all HTTP 200. The report explicitly records Valency 0 and publisher TDM 0. |
| Fresh 50 | 425 total paper-retrieval HTTP attempts. Five credential-backed Springer **open-access JATS** API attempts all returned 401. Seven Wiley TDM attempts returned four 403s and three 400s. These publisher attempts supplied no successful full text. Public discovery/retrieval also used Europe PMC, Crossref, OpenAlex and NCBI. |
| Edge47 follow-up | The authenticated Wiley attempt for the 1993 paper returned 403. Local Zotero search found neither the priority 1983 nor 1993 paper. The 1983 access check skipped the LibKey/proxy route because it was unconfigured at that time. The complete 2024 paper actually inspected was retrieved anonymously through Europe PMC, with 149,817 XML bytes saved and verified. |
| CBORG tools | No Valency, Tavily or Context7 execution is recorded in the examined fresh-canary/edge47 transport, packet and receipt metadata. This is a statement about retained traces, not proof that no conversation-level tool call ever occurred. Skill names alone do not establish service execution. |
| BactoTraits identity investigation, 2026-11-15 | Three successful CBORG Tavily calls (one search, two extracts) and four anonymous primary-catalog fetches. Request-ID provenance and raw catalog/text/hash receipts were retained privately. DSMZ/JCM/BacDive establish the accession conflict; no Valency, Context7, publisher TDM or full-paper retrieval was used. |

Public provenance is in the reports linked above. Detailed private receipts remain outside Git: `edge-sample-review-20261112-data/phase2/literature/http-attempts.jsonl` (Springer entries 17, 44, 165, 223, 397; Wiley entries 93, 95, 103, 105, 124, 126, 345), and `retained-pr107-20261115/.temp/fulltext/salt-evidence-47-20261115/` (`local-search.json`, `kandler1983-access.json`, Wiley and primary2024 receipts). These are relative archive locators under the private EdgeMech agent-work directory, not repository files. Do not commit raw article bodies, private source packets or credentials.

Expand Down
Loading
Loading