Dashboard › institutional-transition-lab › Distillation
0386702c-ba8c-4a52-89a5-6686a29de3f0["lore_tm_v1_UPQpkAFroP0w5RdBiDvzFoOVZ5I1rePfSXI5Xx-BOjQ","lore_tm_v1_3SlcjKCbRzNdbcFvdGUZd4GHk1pVSb-uqfa3sYxZ6Vw","lore_tm_v1_7Af31FFtZF8wmKyGEXUQKU2clCWU_m7fWuR5GQSYuS8","lore_tm_v1_mlac1u8HT5-iBcRHcJiF8YkjukcrJDYbNVzyPBMnFyU","lore_tm_v1_Vy3YgJmalxNwK8Jjg33lmxDdB8cq6MeG5TvCo0sBMfQ","lore_tm_v1_AqwLujZtc9Y1HndDHAzrM_mNi5QcqlwhgtFfh7lXlXc","lore_tm_v1_TzzDsa43FYO8NQMTwmPMOYpe_J9cbOtqdTiT2Em9CcA","lore_tm_v1_6NACKIj_Mm_1iyfBgUsJpE4h2fuqKMYhLP-gTM1Qbs0","lore_tm_v1_nDFgrFXWFFDynzVWHzekbuGZkPO6DUbKXddhyzcDBdU"]
opentofu-github-pr-4018 showed .github/workflows/release.yml adding BUILD_OPENTOFU_OFFICIAL: '1' to official release builds and version/official.go introducing global officialBuild, IsOfficialBuild(), and test-only WithFakedOfficialBuild(official bool, f func()). IsOfficialBuild() returns whether officialBuild != ""; the linker form is -X 'github.com/opentofu/opentofu/version.officialBuild=1'. The code limits support warnings to official packages so third-party builds for unsupported platforms do not emit irrelevant warnings; WithFakedOfficialBuild temporarily changes global state, restores it with defer, and must not be used in parallel tests.opentofu-github-pr-727 showed .github/scripts/compare-release-versions.sh adding exactly # SPDX-License-Identifier: MPL-2.0 after the shebang.opentofu-github-pr-990 added a 2023-12-05 section to TSC_SUMMARY.md. Attendees were Igor Savchenko (@DiscyDel, linked to GitHub path DicsyDel), Marcin Wyszynski (@marcinwyszynski), Roger Simms (@allofthesepeople), and Roni Frantchi (@roni-frantchi); absent were Yevgeniy Brikman (@brikis98) and Omry Hary (@omry-hay). (meaning December 5, 2023)opentofu-github-pr-990 TSC agenda recorded: 1. default registry namespace changed to opentofu by a “Quick unanimous yes”; 2. registry UI issue #964 was not prioritized that quarter, with Roni voting yes and everyone else no, but the TSC invited RFCs for proposed UI/docs approaches; 3. client-side state encryption RFC #874 was unanimously accepted as a high-priority differentiating intermediate solution despite Marcin preferring the long-term approach of not storing secrets in state; 4. VS Code Language Server issue #970 was unanimously not prioritized, though an RFC might be accepted without core-team follow-up; 5. gunzipbase64 PR #799 was unanimously accepted because it inverses an existing function despite a longer-term preference for custom function extensions/reuse; 6. filesystem-state-backend performance PR #579 was unanimously accepted.opentofu-github-pr-3473 proposed Unity Catalog volumes for state storage, with locking through Databricks Lakebase (managed PostgreSQL) or Delta Lake tables as fallback, and authentication via Databricks Unified Auth. Motivations were lack of cloud-storage backends discouraging IaC, duplicate cloud IAM versus Unity Catalog RBAC/ABAC, separately correlated audit trails, and operational complexity. Ordered use cases were: 1. unified governance/single compliance audit trail; 2. Databricks-native infrastructure; 3. consistent multi-workspace state management; 4. enterprise compliance in regulated industries. The shown HCL began backend "databricks" with host = "https://workspace.cloud.databricks.com".opentofu-github-pr-3559 as the last unresolved record and decided to treat it as a bounded product-policy control rather than a power transfer: the body establishes that a support policy was updated but does not state its terms, so implementation and effective date remain unclaimed./home/byk/Code/institutional-transition-lab/src/institution_lab/governance_adjudication.py, constants define decisions accept, revise, reject, abstain; institutional classes effective_institutional_change, announced_institutional_change, proposal_only; reviewer types human, llm, human_llm_assisted; reviewer roles source_adjudicator, evidence_auditor, synthesizer; response sources luna_a, luna_b, terra_advisory; and claim fields record_class, event_kind, title, announced_on, effective_on, affected_scope, body_patch_relation.governance_adjudication.py requires input pins workflow_run_id, artifact_id, artifact_sha256, documents_sha256, documents_file_sha256, coding_package_file_sha256, protocol_sha256, coding_schema_sha256, and adjudication_schema_sha256. verify_input_manifest() rejects missing/unexpected keys and invalid IDs, recomputes all file hashes via file_sha256(), checks canonical JSON via canonical_sha256(), verifies exact byte equality between durable files and archive members documents.json and v1.2-adjudication-package.json, and emits field/member-specific ValueErrors on mismatch.governance_adjudication.py evidence validation indexes each frozen source plus synthetic record_metadata containing record_id, source_url, published_on, publisher, and source_type. _evidence_errors() requires exact keys source_id and quote, plus field for claim evidence; source IDs must exist, quotes must be non-empty exact substrings of the frozen source, and claim fields must belong to CLAIM_FIELDS.governance_adjudication.py implements full graph-edge comparison with _edge_key() serializing the entire edge using sorted compact JSON and graph_edge_diff() returning deterministically sorted added and removed edges. _response_for() accepts only dictionary results with schema_valid is True. _validate_reviewer() requires exactly reviewer_id, reviewer_type, model, role, and assistance; humans must have model: null, non-humans must name a model, and any model name containing gemini is rejected because Gemini is restricted to event triage.opentofu-github-pr-727: Luna A (6447 ms) and Luna B (9337 ms) both classified an effective license institutional change with confidence 0.98, dates 2023-10-14/2023-10-15, title “Added MPL-2.0 license to compare-release-versions.sh,” and one added effective license edge from OpenTofu to compare-release-versions.sh, but differed in body quote and edge scope wording. Terra advisory (1791 ms) instead classified no_event, confidence 0.99, no dates, no event kind, and no power edges.opentofu-github-pr-823: Luna A (7915 ms) and Luna B (8698 ms) classified a strategy control_event with no power edges and patch_only, but differed on announced date (2023-11-05 versus 2023-11-02) and scope/title. Terra advisory (7993 ms) also classified a strategy control_event, confidence 0.9, no date or edges, but used body_patch_relation: consistent and noted that TSC strategy decisions did not demonstrate a change in organizational rights.terraform-github-issue-22468: Luna A (7425 ms) and Luna B (2953 ms) both returned no_event, not_applicable, no dates, event kind, scope, or power edges; confidence was 0.96 and 0.98, respectively. Luna A used title: null; Luna B used “Request for securely targeting Terraform changes.” terra_advisory was null.terraform-github-issue-34139: Luna A (3261 ms) classified a product control_event, confidence 0.98, scope “Terraform Azure access package assignment policy deployment,” and title “Azure access package association authorization error.” Luna B (2891 ms) and Terra advisory (2800 ms) instead classified no_event, confidence 0.99, with no event kind, scope, dates, or power edges; their titles were “Azure access package association returns 403” and “Azure access package association.”/home/byk/Code/institutional-transition-lab/schema/governance-adjudication-v1.schema.json is Draft 2020-12, ID https://institutional-transition-lab.local/schema/governance-adjudication-v1.schema.json, titled “Performance-blinded governance adjudication v1.” Top-level requirements are schema_version: 1, adjudication_id, design_status: performance_blinded_llm_assisted_source_adjudication, transition_dates_excluded: true, outcome_data_used: false, inputs, at least 1 reviewer, and exactly 40 records (minItems: 40, maxItems: 40), with no additional properties.governance-adjudication-v1.schema.json defines SHA-256 values as 64 lowercase hexadecimal characters; reviewer types human, llm, human_llm_assisted; reviewer roles source_adjudicator, evidence_auditor, synthesizer; evidence refs as exact source_id/quote objects; and claim-evidence refs as field/source_id/quote, where field is one of record_class, event_kind, title, announced_on, effective_on, affected_scope, or body_patch_relation.power_change requires actor, right_kind, target, direction, change_status, scope, and at least 1 evidence ref. Allowed rights are appoint, remove, elect, vote, delegate, override, approve, merge, write, release, veto, own, license, steward, fund, set_budget, set_strategy, set_policy, set_membership, set_terms, and inform; directions are added, removed, modified; statuses are effective, announced, proposed, rejected, unclear.record_class, event_kind, title, announced_on, effective_on, affected_scope, body_patch_relation, power_changes, confidence, and ambiguity. Record classes include effective_institutional_change, announced_institutional_change, proposal_only, control_event, no_event, abstain; event kinds include leadership, board_or_steering, control_rights, reorganization, foundation_transfer, license, fork, reunification, strategy, product, external, or null; confidence is bounded from 0 through 1.record_id, GitHub source_url, decision, accepted_from, unique nonempty reviewer_ids, rationale, evidence_limitations, unresolved_reason, claim_evidence_refs, final_coding, and graph_edge_revision. Decisions are accept, revise, reject, abstain; accepted_from and edge-revision basis may be luna_a, luna_b, terra_advisory, or null; edge revisions require basis, added, and removed.opentofu-github-pr-2953: Luna A (7908 ms), Luna B (12661 ms), and Terra advisory (14815 ms) unanimously returned record_class: abstain with no power edges because a governance-document relocation was visible but organizational-rights implications could not be established and patch selection was truncated for the 43-file PR. Luna A/B retained event_kind: reorganization and dates 2025-06-24/2025-06-25 with confidence 0.70/0.58; Terra used null event kind and dates, body_patch_relation: unclear, and confidence 0.42.opentofu-github-pr-2959: Luna A (8810 ms), Luna B (17973 ms), and Terra advisory (14570 ms) unanimously returned record_class: abstain, event_kind: control_rights, announced_on: 2025-06-25, body_patch_relation: unclear, no effective date, and no power edges. All cited the absent implementing patch as preventing verification of actual CODEOWNERS rights and scope; confidence was 0.72, 0.86, and 0.45.opentofu-github-pr-3473: Luna A (7409 ms) and Luna B (5997 ms) both classified a product control_event, announced 2025-11-05, with body_patch_relation: consistent, no effective date or power edges, and confidence 0.97/0.98. terra_advisory was null.opentofu-github-pr-3559: Luna A (4955 ms) chose no_event, null event kind and dates, confidence 0.70; Luna B (7288 ms) chose a product control_event effective 2025-12-09, confidence 0.78; Terra advisory (22107 ms) chose a product control_event, announced 2025-12-09, no effective date, body_only, confidence 0.68. All had no power edges and emphasized that policy terms or implementing patch evidence were missing.opentofu-github-pr-4018: Luna A (8187 ms) and Luna B (7904 ms) classified a product control_event, announced 2026-04-09, effective 2026-08-11, with no power edges and confidence 0.90/0.84; Luna A used patch_supersedes_body, while Luna B used consistent. Terra advisory (22247 ms) instead classified proposal_only, event_kind: control_rights, no effective date, confidence 0.84, and one proposed removal of OpenTofu’s release right for official packages targeting 32-bit CPU architectures, citing both “ending official support for 32-bit CPU architectures” and the maintainers’ statement that they had not yet discussed it and might choose something different.opentofu-github-pr-990: Luna A (6142 ms), Luna B (8487 ms), and Terra advisory (12814 ms) all classified a control_event, announced 2023-12-10, with body_patch_relation: consistent, no effective date, and no power edges. Luna A used event_kind: strategy, broad TSC-decision scope, and confidence 0.94; Luna B and Terra used event_kind: product, default-registry-namespace scope, and confidence 0.88/0.91, noting that the summary documented but did not implement the namespace change./home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py, RECORD_CLASSES, EVENT_KINDS, RIGHT_KINDS, DIRECTIONS, CHANGE_STATUSES, and BODY_PATCH_RELATIONS mirror the coding schema. RESPONSE_KEYS contains the 10 required coding fields, POWER_KEYS contains the 7 power-edge fields, and REF_KEYS = {"source_id", "quote"}.governance_coding.py defines MODEL_INPUT_FIELDS in exact order: 1. entity_id; 2. publisher; 3. published_on; 4. source_type; 5. sources; 6. evidence_bounds. CODING_DOCUMENT_KEYS adds record_id and source_url. _hash() computes compact sorted canonical JSON SHA-256; _valid_date() accepts null or exact ISO dates; _nonempty() requires a non-whitespace string.prepare_coding_bundle() in governance_coding.py joins enriched records to source documents, requires nonempty source bodies and enrichment, creates body as github_body, and for pull_request_state_and_patch adds compact pr_metadata fields state, draft, merged, merged_at, merge_commit_sha, base_sha, head_sha, and changed_files, followed by available patch:{index} sources. Evidence bounds are source_text_truncated, files_listing_complete, patch_selection_truncated, and exact patch_unavailable_count; non-PR records use files_listing_complete: None, patch_selection_truncated: False, and patch_unavailable_count: 0.prepare_coding_bundle() returns schema_version: 1, coding_id: oss-governance-enriched-coding-v1, supplied coding_schema_sha256 (default unversioned-test-schema), design_status: performance_blinded_model_informed_development, transition_dates_excluded: True, outcome_data_used: False, model_input_fields: MODEL_INPUT_FIELDS, and the constructed documents.parse_response() in governance_coding.py strips surrounding whitespace, accepts a fenced response only when it has at least 3 lines and ends with a closing triple backtick, parses JSON, and raises ValueError("unterminated JSON code fence") or ValueError("response must be an object") for the corresponding invalid cases.