OSINT family-adapter edges + Callcenter DataFusion/Gremlin + aiwar POC (codex roll-up)#560
Merged
Merged
Conversation
…iwar POC Follow-up to merged #557; rolls in both codex P1 review findings + the operator's 16-family-adapter edge model + the Callcenter slice + an aiwar OSINT POC on the real graph. soa_graph (codex P1 #1 + #2, operator model): - classid IS the class (exact): project_snapshot / nearest_anchor include only rows where classid == domain.classid — a mixed-class board can't leak one domain's nodes/edges into another's view. - 16 x 8-bit family-node adapters: the canonical EdgeBlock is read as 16 family adapters (12 in-family + 4 out-of-family), each non-zero byte -> a FAMILY node by `family & 0xFF`, collision-aware (ambiguous low byte skipped, never mis-routed). Member-by-identity resolution removed -> the >255-member aliasing dissolves (resolution is family-level only). Trade: mixin dependency for extreme render stability + flexibility (E-FAMILY-ADAPTER-EDGES-ARE-RENDER-STABLE). lance-graph-callcenter (the slice): - graph_table (query-lite): GraphSnapshot -> `nodes` + `edges` arrow MemTable TableProviders + register_graph(SessionContext). The DataFusion / SQL / Cypher->SQL path (mirrors transcode::ontology_table). - graph_gremlin (always-on, pure contract types): g(&snap).v().out()/ .in_()/.out_e(label)/.values_kind() — the Gremlin POC = SurrealQL `->edge->` traversal kernel. contract::aiwar + example (the POC the operator asked for): - AiwarClassView (entity category ⇒ family id) + aiwar_node_rows ingest the real AdaWorldAPI/aiwar-neo4j-harvest/data/aiwar_graph.json into OSINT NodeRows; project_snapshot gives a Gotham graph whose family nodes ARE the categories. Example run: 221 entities/326 edges -> 281 nodes (221 members + 60 family hubs) + 481 edges. q2 wires it to the Quadro-2 visual. Tests: contract 703 lib (+5), clippy --all-targets -D warnings clean. callcenter 10 graph tests (--features query, incl. live SQL roundtrip), default build compiles graph_gremlin; new files clippy-clean (pre-existing callcenter query-clippy debt logged TD-CALLCENTER-QUERY-CLIPPY). Board updated (LATEST_STATE, AGENT_LOG, EPIPHANIES, TECH_DEBT, plan). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CcpLeEC3XK8Eye53GKBVvi
|
Warning Review limit reached
More reviews will be available in 46 minutes and 59 seconds. Learn how PR review limits work. Your organization has used up its prepaid credits, and credit purchases are no longer available. Enable the review add-on in the billing tab to keep reviews running — you're only billed for reviews past your plan's rate limits ($0.25/file). ⌛ How to resolve this issue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based credits. 🚦 How do rate limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan refill rate. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, the refill rate gradually slows as usage increases. The highest same-day bursts are limited more strictly. Please see our Fair Usage Limits Policy for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (5)
📝 WalkthroughWalkthroughThe PR reworks the SoA→Gotham graph projection to use collision-aware 16×8-bit family-node adapter resolution with classid filtering, adds a feature-gated GUID v2 tail layout (3×u16) with NiblePath lowering, introduces callcenter Gremlin traversal and DataFusion table providers over ChangesFamily-node adapter edges, GUID v2 tail, callcenter surfaces, aiwar OSINT POC
Sequence Diagram(s)sequenceDiagram
participant CLI as aiwar_family_poc
participant Contract as lance-graph-contract
participant SOA as soa_graph::project_snapshot
participant Callcenter as lance-graph-callcenter
CLI->>Contract: ingest_aiwar_json(path) → LiteralGraph
CLI->>Contract: AiwarClassView::from_graph(&graph)
Contract-->>CLI: category→family_id map
CLI->>Contract: aiwar_node_rows(&graph) → Vec<NodeRow>
Note over Contract: EdgeBlock slots = out-of-family target-category bytes
CLI->>SOA: project_snapshot(OSINT_GOTHAM, rows)
Note over SOA: filter by classid, collision-aware family-node adapter resolution
SOA-->>CLI: GraphSnapshot (family + member nodes, family: edges)
CLI->>Callcenter: g(&snap).v(&[ids]).out_e(label).to_vec()
Callcenter-->>CLI: reached vertex ids
CLI->>Callcenter: register_graph(&ctx, &snap)
Callcenter-->>CLI: nodes/edges MemTable in SessionContext
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![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: 1a4b76ac01
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".
…ation plan Capture the operator's what-if before it dilutes: repartition the 48-bit basin tail family(u24)|identity(u24) -> leaf(u16)|family(u16)|identity(u16) (uniform 8×u16 key) + scope codebooks per family (family -> Codebook, finer sibling of classid -> ClassView). Blast radius measured (grep): CONTAINED in lance-graph — q2/smb-office-rs/ medcare-rs = 0 (no downstream touches NodeGuid); routing prefix (from_guid_prefix / mailbox_scan / NiblePath<->entity_type) is tail-agnostic; ~3 layout files (canonical_node, soa_graph, aiwar) + ~35 mostly-test NodeGuid::new call sites. Unrelated 0xFFFF.. hits are cycle counters / fingerprint masks, not the GUID tail. PROPOSED — gated on operator sign-off (canon version bump per I-LEGACY-API-FEATURE-GATED) + two capacity numbers (<=65536 identities per (leaf,family); <=256 codebook entries per family before split). Ships feature-gated guid-v2-tail (default OFF) with field-isolation matrix + version gate. INTEGRATION_PLANS prepended; D-GV2-1..5. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CcpLeEC3XK8Eye53GKBVvi
… mixin
Operator design lock (2026-06-20) refining the guid-v2-tail plan:
- leaf is the 4th HHTL routing tier ("a natural HHTL"): the cascade is
HEEL·HIP·TWIG·leaf = 4 tiers × 4 nibbles = 16 nibbles = a full u64
NiblePath. from_guid_prefix_v2 = HEEL·HIP·TWIG·leaf (classid is the
separate codebook prefix).
- family = the basin / episodic hub; identity = instance. basin-local
key = family·identity (4 bytes, was 6). u16 offsets: leaf 10..12,
family 12..14, identity 14..16.
- Family node = episodic basin: connections accumulated on the basin ARE
the supporting edges of every member. Mixin = O(1) address reference
(a byte), shared state stored once on the basin, distance = HHTL hop
arithmetic on the address — never O(n) edge materialization or BFS.
EPIPHANY E-MIXIN-IS-AN-ADDRESS-REFERENCE-NOT-A-COPY captures it; the
open leaf/family decision is RESOLVED (leaf=routing, family=basin).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CcpLeEC3XK8Eye53GKBVvi
…feature-gated Operator greenlit the guid-v2-tail plan. Ships D-GV2-1: the v2 basin tail behind feature `guid-v2-tail` (default OFF), ADDITIVE + NON-breaking (v1 new/family()/identity() untouched). canonical_node (under #[cfg(feature="guid-v2-tail")]): - new_v2(classid,heel,hip,twig,leaf,family,identity) — leaf/family/identity as full u16 at fixed offsets 10..12 / 12..14 / 14..16 (no u24). - leaf() (4th HHTL tier), family_v2() (basin/episodic hub), identity_v2() (instance), local_key_v2() (family++identity, 4B), decode_v2/GuidPartsV2, to_hex_v2 (uniform 4-hex Display), GUID_TAIL_LAYOUT_VERSION_V2 = 2. hhtl: from_guid_prefix_v2 = HEEL·HIP·TWIG·leaf (16 nibbles). leaf IS in the routing path; family/identity are the basin tail (NOT in the path); classid is the separate codebook prefix. Per I-LEGACY-API-FEATURE-GATED: distinct v2 names (no function silently changes semantics under the flag), field-isolation matrix test (vary one tier → only that accessor changes), v1/v2 coexistence test, leaf-in-path test, version-gate const. Cutover (rename v2→canonical, deprecate v1, ENVELOPE_LAYOUT_VERSION bump) = D-GV2-5, after D-GV2-2/3/4 adopt the v2 accessors. Verified BOTH configs: default lib tests 703 (unchanged, non-breaking); --features guid-v2-tail 706 (+3 v2 tests); clippy -D warnings clean both. Plan D-GV2-1 marked SHIPPED; AGENT_LOG updated. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CcpLeEC3XK8Eye53GKBVvi
…on tile pyramid Operator design lock: making every tier the same size (u16) makes the KEY, the per-family CODEBOOK, the VALUE tile, and the PERTURBATION pyramid all the SAME 2bit×2bit 4×4 Morton-tile primitive. - 1 nibble = 2bit×2bit = a 4×4 Morton tile = FAN_OUT-16 (one HHTL level; the same morton4 the domino AMX kernel runs on the value side). - 1 u16 tier = 4 nibbles = a 256×256 Morton tile (256 = 4⁴, the OGAR centroid tile per tier; nibble-interleave = alternating-axis refinement). - 8 u16 tiers = one stacked pyramid. One kernel (Morton + AMX 4×4 BF16 GEMM sweeps tiers/codebooks/values uniformly), one distance (Morton common-prefix = HHTL hop = family_hop_count), one codebook shape (256×256 per tier). The 24+24 tail broke this (u24 ≠ clean tile); 16+16+16 restores it. EPIPHANY E-UNIFORM-MORTON-TILE-PYRAMID; plan section added. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CcpLeEC3XK8Eye53GKBVvi
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
🧹 Nitpick comments (1)
crates/lance-graph-callcenter/src/graph_gremlin.rs (1)
110-120: ⚡ Quick winAvoid repeated linear scans in
values_kind().
values_kind()does a full node scan per current ID. Building an ID→kind index once per call removes the O(|current|×|nodes|) behavior.⚡ Proposed refactor
-use std::collections::HashSet; +use std::collections::{HashMap, HashSet}; ... pub fn values_kind(&self) -> Vec<String> { + let kinds_by_id: HashMap<&str, &str> = self + .snap + .nodes + .iter() + .map(|n| (n.id.as_str(), n.kind.as_str())) + .collect(); self.current .iter() - .filter_map(|id| { - self.snap - .nodes - .iter() - .find(|n| &n.id == id) - .map(|n| n.kind.clone()) - }) + .filter_map(|id| kinds_by_id.get(id.as_str()).map(|k| (*k).to_string())) .collect() }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@crates/lance-graph-callcenter/src/graph_gremlin.rs` around lines 110 - 120, The `values_kind()` method currently performs a full linear scan of `self.snap.nodes` for each ID in `self.current`, resulting in O(|current|×|nodes|) complexity. Instead, build a HashMap index that maps node IDs to their kinds once at the start of the method, then iterate through `self.current` and use the HashMap to look up each kind in constant time. This will change the complexity to O(|nodes| + |current|) by trading space for time.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@crates/lance-graph-callcenter/src/graph_gremlin.rs`:
- Around line 36-41: The v method currently accepts arbitrary vertex IDs without
validating they exist in the graph, allowing operations like count() to report
non-existent vertices. In the else branch of the v method where custom IDs are
provided, filter the input IDs to only include those that actually exist in
self.snap.nodes. Modify the mapping logic to check each ID against the available
nodes in the graph before adding it to the current vector.
In `@crates/lance-graph-callcenter/src/graph_table.rs`:
- Around line 197-204: The test currently only verifies that the query result
has one row but does not validate the actual aggregate count value returned by
the COUNT(*) query. Extract the count value from the first batch (batches[0])
and add an assertion to verify it equals the expected node count of 4 (as
mentioned in the comment: 2 OSINT members plus 2 family nodes). This ensures the
test fails if the wrong total is returned, not just if the result shape is
incorrect.
In `@crates/lance-graph-contract/src/aiwar.rs`:
- Around line 39-43: The enumerate index cast to u32 in the families mapping
lacks bounds validation, which allows silent truncation via the masking
operation at lines 106-107 that could produce duplicate NodeGuids. Add explicit
validation to ensure the enumerated index does not exceed the 24-bit limit
(0xFF_FFFF) before the cast to u32, and raise an error if this limit is breached
instead of silently wrapping or masking the value. Apply the same validation fix
to the other location mentioned at lines 101-107 where similar index-to-guid
conversion occurs.
In `@crates/lance-graph-contract/src/soa_graph.rs`:
- Around line 392-397: The ambiguity test filter on line 395 uses an incorrect
source-id suffix in the ends_with call. Change the suffix from "010000000001" to
"000100000001" which is the canonical tail for family=0x0100 and identity=1.
This ensures the filter correctly identifies the intended node with the
ambiguous low byte and prevents the test from passing vacuously.
---
Nitpick comments:
In `@crates/lance-graph-callcenter/src/graph_gremlin.rs`:
- Around line 110-120: The `values_kind()` method currently performs a full
linear scan of `self.snap.nodes` for each ID in `self.current`, resulting in
O(|current|×|nodes|) complexity. Instead, build a HashMap index that maps node
IDs to their kinds once at the start of the method, then iterate through
`self.current` and use the HashMap to look up each kind in constant time. This
will change the complexity to O(|nodes| + |current|) by trading space for time.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 365a4a8b-f78e-4469-bc7d-04f184b812ce
📒 Files selected for processing (17)
.claude/board/AGENT_LOG.md.claude/board/EPIPHANIES.md.claude/board/INTEGRATION_PLANS.md.claude/board/LATEST_STATE.md.claude/board/TECH_DEBT.md.claude/plans/guid-v2-tail-per-family-codebook-v1.md.claude/plans/unified-soa-rubikon-integration-v1.mdcrates/lance-graph-callcenter/src/graph_gremlin.rscrates/lance-graph-callcenter/src/graph_table.rscrates/lance-graph-callcenter/src/lib.rscrates/lance-graph-contract/Cargo.tomlcrates/lance-graph-contract/examples/aiwar_family_poc.rscrates/lance-graph-contract/src/aiwar.rscrates/lance-graph-contract/src/canonical_node.rscrates/lance-graph-contract/src/hhtl.rscrates/lance-graph-contract/src/lib.rscrates/lance-graph-contract/src/soa_graph.rs
Comment on lines
+36
to
+41
| pub fn v(&self, ids: &[&str]) -> Traversal<'a> { | ||
| let current: Vec<String> = if ids.is_empty() { | ||
| self.snap.nodes.iter().map(|n| n.id.clone()).collect() | ||
| } else { | ||
| ids.iter().map(|s| s.to_string()).collect() | ||
| }; |
There was a problem hiding this comment.
Filter v(ids) seeds to vertices that actually exist.
Right now Line 40 accepts arbitrary IDs, so count()/to_vec() can report non-existent vertices (e.g., v(&["missing"]).count() == 1).
💡 Proposed fix
pub fn v(&self, ids: &[&str]) -> Traversal<'a> {
+ let node_ids: HashSet<&str> = self.snap.nodes.iter().map(|n| n.id.as_str()).collect();
let current: Vec<String> = if ids.is_empty() {
self.snap.nodes.iter().map(|n| n.id.clone()).collect()
} else {
- ids.iter().map(|s| s.to_string()).collect()
+ ids.iter()
+ .copied()
+ .filter(|id| node_ids.contains(*id))
+ .map(str::to_string)
+ .collect()
};
Traversal {
snap: self.snap,
current,
}
}🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/lance-graph-callcenter/src/graph_gremlin.rs` around lines 36 - 41, The
v method currently accepts arbitrary vertex IDs without validating they exist in
the graph, allowing operations like count() to report non-existent vertices. In
the else branch of the v method where custom IDs are provided, filter the input
IDs to only include those that actually exist in self.snap.nodes. Modify the
mapping logic to check each ID against the available nodes in the graph before
adding it to the current vector.
Comment on lines
+197
to
+204
| // GROUP BY over node kinds: 2 OSINT members + 2 family nodes. | ||
| let df = ctx | ||
| .sql("SELECT count(*) AS n FROM nodes") | ||
| .await | ||
| .unwrap(); | ||
| let batches = df.collect().await.unwrap(); | ||
| assert_eq!(batches[0].num_rows(), 1); | ||
| } |
There was a problem hiding this comment.
Assert the aggregate value, not only result shape.
This currently verifies one output row, but not that count(*) is the expected node count. A wrong total could still pass.
✅ Tighten test assertion
- let df = ctx
- .sql("SELECT count(*) AS n FROM nodes")
+ let df = ctx
+ .sql("SELECT count(*) AS n FROM nodes HAVING count(*) = 4")
.await
.unwrap();
let batches = df.collect().await.unwrap();
- assert_eq!(batches[0].num_rows(), 1);
+ let rows: usize = batches.iter().map(|b| b.num_rows()).sum();
+ assert_eq!(rows, 1, "expected exactly 4 projected nodes");📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| // GROUP BY over node kinds: 2 OSINT members + 2 family nodes. | |
| let df = ctx | |
| .sql("SELECT count(*) AS n FROM nodes") | |
| .await | |
| .unwrap(); | |
| let batches = df.collect().await.unwrap(); | |
| assert_eq!(batches[0].num_rows(), 1); | |
| } | |
| // GROUP BY over node kinds: 2 OSINT members + 2 family nodes. | |
| let df = ctx | |
| .sql("SELECT count(*) AS n FROM nodes HAVING count(*) = 4") | |
| .await | |
| .unwrap(); | |
| let batches = df.collect().await.unwrap(); | |
| let rows: usize = batches.iter().map(|b| b.num_rows()).sum(); | |
| assert_eq!(rows, 1, "expected exactly 4 projected nodes"); | |
| } |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/lance-graph-callcenter/src/graph_table.rs` around lines 197 - 204, The
test currently only verifies that the query result has one row but does not
validate the actual aggregate count value returned by the COUNT(*) query.
Extract the count value from the first batch (batches[0]) and add an assertion
to verify it equals the expected node count of 4 (as mentioned in the comment: 2
OSINT members plus 2 family nodes). This ensures the test fails if the wrong
total is returned, not just if the result shape is incorrect.
Comment on lines
+39
to
+43
| let families = labels | ||
| .into_iter() | ||
| .enumerate() | ||
| .map(|(i, l)| (l, (i as u32) + 1)) | ||
| .collect(); |
There was a problem hiding this comment.
Avoid silent u24 truncation when minting GUID family/identity.
(i as u32) plus masking at Lines 106-107 (& 0x00FF_FFFF) can silently wrap large inputs, producing duplicate NodeGuids and mislinked snapshots instead of failing fast.
Suggested patch
- let families = labels
+ let families = labels
.into_iter()
.enumerate()
- .map(|(i, l)| (l, (i as u32) + 1))
+ .map(|(i, l)| {
+ let fam = u32::try_from(i + 1).expect("category index exceeds u32");
+ assert!(fam <= 0x00FF_FFFF, "family id exceeds u24");
+ (l, fam)
+ })
.collect();
@@
- NodeRow {
+ let identity = u32::try_from(i).expect("node index exceeds u32");
+ assert!(identity <= 0x00FF_FFFF, "identity exceeds u24");
+ NodeRow {
key: NodeGuid::new(
NodeGuid::CLASSID_OSINT,
0,
0,
0,
- fam & 0x00FF_FFFF,
- (i as u32) & 0x00FF_FFFF,
+ fam,
+ identity,
),Also applies to: 101-107
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/lance-graph-contract/src/aiwar.rs` around lines 39 - 43, The enumerate
index cast to u32 in the families mapping lacks bounds validation, which allows
silent truncation via the masking operation at lines 106-107 that could produce
duplicate NodeGuids. Add explicit validation to ensure the enumerated index does
not exceed the 24-bit limit (0xFF_FFFF) before the cast to u32, and raise an
error if this limit is breached instead of silently wrapping or masking the
value. Apply the same validation fix to the other location mentioned at lines
101-107 where similar index-to-guid conversion occurs.
Comment on lines
+392
to
+397
| let from_0100 = snap | ||
| .edges | ||
| .iter() | ||
| .filter(|e| e.label == "linked" && e.source.ends_with("010000000001")) | ||
| .count(); | ||
| assert_eq!(from_0100, 0, "ambiguous low byte must be skipped"); |
There was a problem hiding this comment.
Fix the source-id suffix in the ambiguity test filter.
On Line 395, ends_with("010000000001") does not match the canonical tail for family=0x0100, identity=1 (000100000001), so this check can miss the intended node and pass vacuously.
Suggested patch
- .filter(|e| e.label == "linked" && e.source.ends_with("010000000001"))
+ .filter(|e| e.label == "linked" && e.source.ends_with("000100000001"))📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| let from_0100 = snap | |
| .edges | |
| .iter() | |
| .filter(|e| e.label == "linked" && e.source.ends_with("010000000001")) | |
| .count(); | |
| assert_eq!(from_0100, 0, "ambiguous low byte must be skipped"); | |
| let from_0100 = snap | |
| .edges | |
| .iter() | |
| .filter(|e| e.label == "linked" && e.source.ends_with("000100000001")) | |
| .count(); | |
| assert_eq!(from_0100, 0, "ambiguous low byte must be skipped"); |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/lance-graph-contract/src/soa_graph.rs` around lines 392 - 397, The
ambiguity test filter on line 395 uses an incorrect source-id suffix in the
ends_with call. Change the suffix from "010000000001" to "000100000001" which is
the canonical tail for family=0x0100 and identity=1. This ensures the filter
correctly identifies the intended node with the ambiguous low byte and prevents
the test from passing vacuously.
…mily edges (1) graph_gremlin: step() silently deduped reached targets via a `seen` set, breaking Gremlin bag/multiset semantics — g.v(["A","C"]).out().count() gave 1 not 2 when both reach B. Rewrote to per-traverser emission (duplicates preserved); set semantics is now the explicit `dedup()` step. +test out_preserves_bag_multiplicity. (2) aiwar: aiwar_node_rows wrote cross-category adapter bytes into the first 12 in_family slots (labeled `linked` by project_snapshot), so `references` queries missed them and the label flipped with sorted fan-out count. aiwar edges are ALL cross-family (built from tf != fam) → write them to the 4 out_family slots (`references`), cap 4. Test asserts `references` present and no `linked`. contract aiwar 3/3; callcenter gremlin 8/8 (+1 bag test); clippy clean (new files; pre-existing TD-CALLCENTER-QUERY-CLIPPY untouched). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CcpLeEC3XK8Eye53GKBVvi
AdaWorldAPI
pushed a commit
that referenced
this pull request
Jun 20, 2026
…-tail) Continues the v2 arc after #560 merged. The type + in-memory registry tier of the family→Codebook scoping (the Lance-backed/OntologyRegistry tier is deferred). Zero-dep, feature-gated (default OFF). contract::codebook: - Codebook — insertion-ordered index↔label interning, 1-byte index, CODEBOOK_CAP=256; intern() returns None on overflow (the split-the-family signal, never widen the byte). - FamilyCodebookRegistry — family(u16) → Codebook; intern(family,label), resolve(family,index) for cross-family decode. Per-family scoping: the SAME label gets INDEPENDENT indices in different families (no global codebook contamination — this is what dissolves the aiwar "60 noisy families"). The finer sibling of classid→ClassView; the family node's episodic-basin content (E-MIXIN-IS-AN-ADDRESS-REFERENCE-NOT-A-COPY); the 256×256 Morton tile (≤256 leaves for the 1-byte in-family index, E-UNIFORM-MORTON-TILE-PYRAMID). 3 tests (dedup/sequential, overflow-split, per-family scoping). --features guid-v2-tail green; default build clean (codebook absent); clippy -D warnings clean both. Plan D-GV2-2 marked PARTIAL; AGENT_LOG updated. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CcpLeEC3XK8Eye53GKBVvi
AdaWorldAPI
pushed a commit
that referenced
this pull request
Jun 20, 2026
…+ canon conflict Operator asked to DO the ontology-schema migration documentation. Grounded in OGAR crates/ogar-vocab/src/lib.rs (the real codebook): it already defines CODEBOOK (domain-encoded 0xDDCC), ConceptDomain + canonical_concept_domain, source_domain_concept(project|erp), canonical_concept_id, and LabelDTO — and its own note says LabelDTO "long-term belongs in lance-graph-contract; codebook id == NodeGuid.classid low u16." Surfaces a canon CONFLICT: merged CLASSID_OSINT=0x0007 routes to OGAR's Reserved domain (OSINT is 0x07XX); CLASSID_FMA=0x0008 sits in OGAR's OCR block (FMA/anatomy is clinical → Health 0x09XX). Root cause: 0x0007 minted from the early "OSINT is 0x0007" guess before ogar-vocab's 0xDDCC layout was consulted. New .claude/plans/ogar-vocab-contract-codebook-migration-v1.md (D-OVC-1..4): host the codebook/ConceptDomain/LabelDTO in contract, classids follow 0xDDCC (mint project 0x01XX + ERP 0x02XX; realign OSINT→0x0700, FMA→Health). The per-family codebook (D-GV2-2) is the finer scope of the same idea. NO code minted/rewritten: realigning merged OSINT/FMA rewrites canon (#557/#560 + CLAUDE.md canon block) → operator sign-off required (plan §5, three decisions). INTEGRATION_PLANS prepended; ISSUES ISS-CLASSID-OGAR-DRIFT filed; AGENT_LOG updated. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CcpLeEC3XK8Eye53GKBVvi
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
2 participants
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.
What
Follow-up to merged #557. Rolls in both codex P1 review findings, the operator's 16×8-bit family-node-adapter edge model, the Callcenter DataFusion/Gremlin slice, and an aiwar OSINT POC on the real graph. q2 wires the resulting
GraphSnapshotto the Quadro-2 visual.codex P1 fixes (rolled in)
project_snapshot/nearest_anchornow include only rows whereclassid == domain.classid, so a mixed-class SoA board can't leak FMA/default nodes (or cross-domain edges) into an OSINT projection. New testmixed_class_board_excludes_other_domains.EdgeBlockis read as 16 family-node adapters, each byte → a FAMILY byfamily & 0xFF(collision-aware: a low byte shared by two families is skipped, never mis-routed). Member-by-identity resolution is gone, so the >255-member aliasing dissolves (resolution is only ever family-level). New testambiguous_family_low_byte_is_skipped_not_misrouted.E-FAMILY-ADAPTER-EDGES-ARE-RENDER-STABLE: edges→families trades a mixin dependency (a referenced family must exist) for extreme render stability (family hubs are fixed anchors; the layout doesn't churn) + flexibility (a node mixes in up to 16 family adjacencies). 8×16-bit / 32×4-residue / member→member second-hop are deferred richer flavors; helix-basin-anchor (CLAM ⇄ Louvain turbovec residue) is the operator's named later enhancement.Callcenter slice (
lance-graph-callcenter)graph_table(query-lite) —GraphSnapshot→nodes+edgesarrow MemTableTableProviders +register_graph(SessionContext). The DataFusion / SQL / Cypher→SQL path (mirrorstranscode::ontology_table). Live SQL roundtrip test:SELECT target FROM edges WHERE source=… AND label='references'.graph_gremlin(always-on, pure contract types) —g(&snap).v(&[…]).out()/.in_()/.out_e(label)/.values_kind()/.count(). The basic Gremlin POC, doubling as the SurrealQL->edge->traversal kernel.aiwar OSINT POC (
contract::aiwar+ example)AiwarClassView(entity category ⇒ family id) +aiwar_node_rowsingest the realaiwar-neo4j-harvest/data/aiwar_graph.json(via the existingliteral_graph::ingest_aiwar_json) into OSINTNodeRows →project_snapshotgives a Gotham graph whose family nodes ARE the categories.examples/aiwar_family_poc.rson the real 174 KB graph: 221 entities / 326 edges → 281 nodes (221 members + 60 family hubs) + 481 edges.typefield (FacialRecognition, ComputerVisionvs…; …split); coarseN_*-bucket grouping is a one-line granularity knob q2 can choose — the mechanism is correct either way.Tests
cargo test -p lance-graph-contract --lib— 703 (+5: aiwar ×3, soa_graph ambiguity + mixed-class ×2).cargo clippy -p lance-graph-contract --all-targets -- -D warnings— clean.cargo test -p lance-graph-callcenter --features query— 10 graph tests (incl. live SQL). Default (no-feature) build compilesgraph_gremlin.graph_table.rs,graph_gremlin.rs) are clippy-clean (verified zero errors originate in them). The crate's--features query -D warningsclippy fails on pre-existing oxrdf-deprecation/doc debt in unrelated modules — logged asTD-CALLCENTER-QUERY-CLIPPY, out of scope here.Board updated in-commit (LATEST_STATE, AGENT_LOG, EPIPHANIES, TECH_DEBT, plan).
🤖 Generated with Claude Code
Generated by Claude Code
Summary by CodeRabbit
Release Notes
New Features
Documentation