Sprint 3: Action vocab (ActionDef + ActionInvocation) + ogar-adapter + SPO+TeKaMoLo emission#3
Merged
Merged
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: be571ec1f0
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Comment on lines
+664
to
+667
| triples.push(Triple::new(&id, "ogar:methodBody", body.clone())); | ||
| } | ||
| for d in &def.decorators { | ||
| triples.push(Triple::new(&id, "ogar:decoratorName", d.clone())); |
There was a problem hiding this comment.
When an ActionDef has a body or decorators, these triples reuse ogar:methodBody / ogar:decoratorName, but the canonical vocabulary declares those predicates with rdfs:domain ogar:MethodDecl in vocab/ogar.ttl lines 201-204. Under RDFS reasoning, every decorated/body-carrying ActionDef will also be inferred as a MethodDecl, polluting method queries and violating the vocabulary-as-contract rule; add ActionDef-specific predicates or widen the TTL domains before emitting them here.
Useful? React with 👍 / 👎.
Comment on lines
+737
to
+744
| v.push(Triple::new(action_subject, "ogar:dependsPath", p.clone())); | ||
| } | ||
| v | ||
| } | ||
| KausalSpec::ContextDepends { keys } => { | ||
| let mut v = vec![Triple::new(action_subject, "ogar:kausalKind", "ogar:ContextDepends")]; | ||
| for k in keys { | ||
| v.push(Triple::new(action_subject, "ogar:dependsContext", k.clone())); |
There was a problem hiding this comment.
For KausalSpec::Depends / ContextDepends, emitting ogar:dependsPath and ogar:dependsContext from an ActionDef conflicts with the TTL contract: those predicates are declared with rdfs:domain ogar:ComputedField in vocab/ogar.ttl lines 187-188. Any action with dependency guards will therefore be inferred as a ComputedField, which corrupts downstream graph queries that enumerate computed fields; define action/kausal dependency predicates or adjust the vocabulary domain before using these terms here.
Useful? React with 👍 / 👎.
…d on UPSTREAM-DEPS) Rebased onto main after the UPSTREAM-DEPS.md doc landed. Sprint 3 content unchanged; adds an EPIPHANIES entry noting lance-graph #461 (Quasicryth + COW Adaptive Radix Tree) as the future binding point for migrating the BTreeMapAdapter placeholder to a COW radix structure (per B3 YAGNI exit condition). NEW crate: crates/ogar-adapter - Adapter trait with map() direction (unmap deferred to Sprint 4.5) - BTreeMapAdapter with iter_prefix() filter - OdooAdapter with 15 default leaves for OGAR↔Odoo identity renames - 7 unit tests ogar-vocab — Action vocabulary - ActionDef (declaration) + ActionInvocation (per (S, P, O, context)) — B1 fix - KausalSpec sum type (StateGuard / LifecycleTrigger / Depends / ContextDepends / External) — B1 fix - B2 provenance: trace_id, parent_invocation, idempotency_key, emitted_at_millis, failure_reason - ActionState lifecycle: Pending / Committed / Failed / Cancelled — B2 fix - LokalSpec with tenant + actor + company ogar-emitter — emit_action_def + emit_action_invocation - Full SPO+TeKaMoLo triple emission - KausalSpec sum-type dispatch produces variant-specific triples - 4 new tests — 49 tests total vocab/ogar.ttl — 24 new predicates + 4 enumeration classes EPIPHANIES updates - Sprint 3 brutal-review cycle synthesis (5 research + 3 brutal — B1+B2 critical fixes LANDED, B3 selective YAGNI cuts applied) - lance-graph #461 (COW Adaptive Radix Tree) noted as the future NibleHHTL substrate; current BTreeMapAdapter placeholder per B3 will migrate when leaf count / cross-tenant sharing demands a COW radix structure Tests - cargo check --workspace clean - cargo test --workspace: 49 tests passing
AdaWorldAPI
force-pushed
the
claude/sprint-3
branch
from
be571ec to
7ca5809
Compare
…tion Codex review on the Sprint 3 emitter caught two RDFS domain-pollution gaps: emit_action_def reused predicates whose canonical domain is a DIFFERENT class, so under RDFS/OWL reasoning an ActionDef would be inferred as a MethodDecl or ComputedField — and, worse, plain SPARQL queries enumerating those classes would catch ActionDefs too. The real problem is query pollution, not just type inference: a query `?cf ogar:dependsPath ?p` enumerating computed-field dependencies would return dependency-guarded actions if they shared the predicate. Union domains would NOT fix this (the predicate is still shared); distinct predicates are required. Two reuses fixed with action/kausal-scoped predicates: 1. ActionDef body + decorators (Codex finding #1) ogar:methodBody → ogar:actionBody (domain ogar:ActionDef) ogar:decoratorName → ogar:actionDecorator (domain ogar:ActionDef) The MethodDecl emitter keeps ogar:methodBody / ogar:decoratorName (domain ogar:MethodDecl) — unchanged. 2. Kausal dependency paths (Codex finding #2) ogar:dependsPath → ogar:kausalDependsPath (domain ogar:ActionDef) ogar:dependsContext → ogar:kausalDependsContext (domain ogar:ActionDef) The ComputedField emitter keeps ogar:dependsPath / ogar:dependsContext (domain ogar:ComputedField) — unchanged. vocab/ogar.ttl - Added ogar:actionBody, ogar:actionDecorator, ogar:kausalDependsPath, ogar:kausalDependsContext, all with rdfs:domain ogar:ActionDef. - Gave the previously domain-less kausal predicates (kausalKind, guardField, guardValue, triggerEvent) explicit rdfs:domain ogar:ActionDef so the full kausal sub-graph is contract-scoped to ActionDef. Tests - kausal_spec_variants_emit_distinct_kinds now asserts on ogar:kausalDependsPath AND asserts ogar:dependsPath does NOT appear on an ActionDef (negative test guarding against regression). - 49 tests still passing. Verified by grep: ogar:methodBody / ogar:decoratorName appear only in emit_method_decl_indexed; ogar:dependsPath / ogar:dependsContext only in emit_computed_field_indexed. No cross-domain predicate reuse remains.
AdaWorldAPI
added a commit
that referenced
this pull request
Jun 4, 2026
…; Sprint 5b unblocked bardioc responded to OGAR's 3 surfaced decisions + mirrored the coordination record (symmetric, with the Lance-sub bus consumer API + ownership table). Net state change recorded: - Decision #1 (registry append API) RESOLVED: the Box::leak interning workaround (ogar_proposal::boundary, PR #5) is accepted. Sprint 5b proceeds without waiting for an upstream SchemaOwned variant (which stays a nice-to-have-cleaner-later, not a blocker). Sprint 5b UNBLOCKED — now gated only on the cross-repo build (protoc), not a decision. - Decision #2 (mailbox home) RESOLVED (grill #9) — already absorbed. - Decision #3 (Lance-sub bus API) SURFACED: consumer API documented by bardioc, but the upstream symbol layout (concrete Rust types) hasn't shipped. Sprint 7 STAYS BLOCKED until it lands — binding against a documented-but-unshipped symbol layout is the guess-the-contract rework this discipline avoids. Also reconciled Sprint 5's stale deliverables (same defect class Codex flagged on Sprint 7): the old "wire contract layer / NiblePath dictionary / Lance write path" deliverables were pre-correction and are DROPPED (upstream's job). Sprint 5 now split 5a (shipped, PR #5) + 5b (unblocked). EPIPHANIES updated append-only (new dated entry citing the prior cross-session entry). Docs only.
AdaWorldAPI
added a commit
that referenced
this pull request
Jun 4, 2026
…print 7 unblocked + corrected; #4 surfaced The parallel sessions shipped the Lance-subscription bus and placed the temporal-epistemology framework. Two things land on OGAR's side; one is a correction to OGAR's own doc. NEW docs/TEMPORAL-TIME-TRAVEL.md - Decision #3 SHIPPED: the bus is lance-graph-callcenter::version_watcher::LanceVersionWatcher, built on std::sync::{Arc,RwLock,Mutex,Condvar}, NOT tokio (upstream I-2 invariant: tokio is Layer-3 outbound only; hot loop never uses tokio::sync). Hot path: subscribe() -> WatchReceiver -> wait_changed() (Condvar park) -> current() returns Arc<CognitiveEventRow>. - This CORRECTS OGAR's design: SOA-IMPLEMENTATION §5.2's tokio::sync KanbanMailbox violates I-2 and is superseded by std::sync (Mutex<VecDeque> + Condvar) on the hot path. WIP/pull/backpressure policy stands; mechanism changes. (Sprint 7 was held — building the tokio version would have been exactly this rework.) - SoA bridge ownership: ontology owns identity/classes/codebooks; callcenter owns LanceMembrane (sole writer) + watcher + CognitiveEventRow; ogar-runtime is a std::sync SUBSCRIBER, never a writer. - Temporal-epistemology is a PLANNER-layer query annotation, NOT OGAR: the parallel session mapped the Python framework onto Lance versions (KnowledgeHorizon = checkout_version; EpistemicMode = planner query annotation; cross-server = HLC tick on CognitiveEventRow). OGAR does NOT build QueryReference/EpistemicMode (planner) or the HLC stamp (callcenter). OGAR consumes. - Decision #4 surfaced (NOT blocking): ActionInvocation.emitted_at_millis is wall-clock i64; cross-server causal ordering needs an HLC tuple. Keep emitted_at an Option on the non_exhaustive struct so an HLC variant is a non-breaking add; conform to callcenter's HLC when the cross-server workload lands. Single-server causal order IS the Lance version sequence, so no change now. CORRECTED docs/SOA-IMPLEMENTATION.md §5 - Correction banner at the top of the runtime layer: the tokio::sync sketches are superseded by the std::sync Condvar LanceVersionWatcher pattern per I-2. Policy kept, mechanism replaced. PLAN.md Sprint 7 - UNBLOCKED + CORRECTED: std::sync Condvar subscriber over LanceVersionWatcher, not tokio/Ractor on the hot path. Consume lance-graph-callcenter::version_watcher as reference. Remaining gate: cross-repo build (protoc) + the user's signal. Decision #4 noted. EPIPHANIES: one entry recording decision #3 shipped + the std::sync correction + the epistemology boundary + decision #4 surfaced. Posture matches the other session: FYI absorbed, not building yet. Sprint 7 holds for the signal. Docs only; tests green.
This was referenced Jun 4, 2026
AdaWorldAPI
pushed a commit
that referenced
this pull request
Jun 10, 2026
Born from the session's own regrounding audit: six turns of synthesis,
zero probes run — the workspace rule applies verbatim ('if the probe is
NOT RUN, the next deliverable is the probe, not more synthesis'). This
plan orders ALL the work probe-first:
§0 The three shaky joints (J1 TILE256<->PREFIXBOOK circularity;
J2 'lossless for synthesis' scope drift; J3 Parseval !=
gaussian-preservation) promoted to first-class falsification
targets — each with its resolving probe AND its kill condition.
A probe without a kill condition is theater.
§1 Wave 0 — ground in 48 hours: seven probes runnable against
SHIPPED code only (~170 LOC of tests, zero new impl): PHASE-1
(CurveRuler determinism), WHP-1a/1b (Parseval split — transform
vs quantized envelope), PHI-1 (phi-stride vs uniform), F2
(quasicryth paper_theorems run), F13 (parity suite verify), F8
(helix fidelity — lance-graph's oldest owed probe, #459).
§2 Wave 1 — first impl bricks, each gated: C (SoaEnvelope keystone,
F6) · O1 (Class traversal API) · PrefixShapeTable (ROUTE-1) ·
WHP-3/4 (incl. the permanent two-algebra CI guard) · HILBERT-L4
fix (the PP-13 P0-4 red that blocks every L4 claim).
§3 Wave 2 — the conjecture tower gets numbers: 4^4 codebook builder
(CODEBOOK-44 + BOOT-1 -> J1) · perturbation encoder v0 (PERT-RHO
with reported escalation rate -> J2) · F3 spectral anti-moire ·
F10-F12 anchor re-measures.
§4 Wave 3 — identity-arc phases gated per INTEGRATION-MAP (D/B/O7/
R1-R3 incl. the NodeGuid group-3/4 audit addendum); §5 Wave 4 —
consumers; §6 standing rules (every red is a same-day ledger
event; thresholds measured never optimism); §7 ownership.
CLAUDE.md doc-family updated (the plan is read #3).
Docs-only. PII abort-guard: CLEAN. No model identifier.
https://claude.ai/code/session_01PBTGaPCSnnt6u3pjXpbLwY
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.
Summary
Sprint 3 of OGAR: behavior arm lands.
ActionDef+ActionInvocation+ogar-adaptercrate ship with the SPO+TeKaMoLo emission perdocs/ADAPTERS-AND-ACTORS.md.What's new
New crate
crates/ogar-adapterAdaptertrait +BTreeMapAdapterdefault +OdooAdapterfirst concrete implogit-erp/move → odoo:transportetc.)ogar-vocab— Action vocabularyActionDef(declaration) +ActionInvocation(per-context firing) — B1 fixKausalSpecsum type (StateGuard / LifecycleTrigger / Depends / ContextDepends / External) — B1 fixtrace_id,parent_invocation,idempotency_key,emitted_at_millis,failure_reasonActionStatelifecycle: Pending / Committed / Failed / Cancelled — B2 fixLokalSpecwith tenant + actor + companyogar-emitter—emit_action_def+emit_action_invocationvocab/ogar.ttl— 24 new predicates + 4 enumeration classesBrutal-review cycle 3 outcomes
YAGNI cuts (deferred)
unmap()direction — Sprint 4.5 (SurrealQL bidirectional) introducesTest plan
cargo check --workspacecleancargo test --workspace— 49 tests passing (was 45)EPIPHANIES.mdSprint 3 cycle entry — confirms the synthesis