khive-pack-knowledge 0.10.0

Knowledge verb pack — lore corpus (atoms/domains), TF-IDF retrieval, concept registration
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
//! A5 stored-vector rerank behavior at the knowledge search boundary.

use super::{vamana, KnowledgeHandlers};
use async_trait::async_trait;
use khive_pack_kg::KgPack;
use khive_runtime::{
    AllowAllGate, BackendId, EmbedderProvider, KhiveRuntime, Namespace, RuntimeConfig,
    VerbRegistry, VerbRegistryBuilder,
};
use khive_storage::types::{
    BatchWriteSummary, IndexRebuildScope, StorageResult, VectorRecord, VectorSearchHit,
    VectorSearchRequest, VectorStoreInfo,
};
use khive_storage::{StorageError, VectorStore};
use khive_types::SubstrateKind;
use lattice_embed::{EmbedError, EmbeddingModel, EmbeddingService};
use serde_json::{json, Value};
use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
use std::sync::{Arc, Mutex};
use uuid::Uuid;

const MODEL: &str = "all-minilm-l6-v2";
const DIM: usize = 384;
const ALPHA_CONTENT: &str = "shared lexical rerank evidence explains how an indexed atom participates in retrieval while its persisted vector determines the final semantic ordering across otherwise similar search candidates";
const BETA_CONTENT: &str = "shared lexical rerank evidence explains how another indexed atom participates in retrieval while its persisted vector determines the final semantic ordering across otherwise similar search candidates";
const DOMAIN_DESCRIPTION: &str = "shared lexical domain evidence explains how a curated domain mirror participates in retrieval while its persisted embedding supplies semantic ranking for related knowledge search requests";

fn axis(index: usize) -> Vec<f32> {
    let mut vector = vec![0.0; DIM];
    vector[index] = 1.0;
    vector
}

#[derive(Default)]
struct Calls {
    passages: Mutex<Vec<String>>,
    generic: AtomicUsize,
    queries: AtomicUsize,
    fail_query: AtomicBool,
}

struct RoleService(Arc<Calls>);

#[async_trait]
impl EmbeddingService for RoleService {
    async fn embed(
        &self,
        texts: &[String],
        _model: EmbeddingModel,
    ) -> Result<Vec<Vec<f32>>, EmbedError> {
        self.0.generic.fetch_add(texts.len(), Ordering::SeqCst);
        Ok(texts.iter().map(|_| axis(1)).collect())
    }

    async fn embed_query(
        &self,
        texts: &[String],
        _model: EmbeddingModel,
    ) -> Result<Vec<Vec<f32>>, EmbedError> {
        self.0.queries.fetch_add(texts.len(), Ordering::SeqCst);
        if self.0.fail_query.load(Ordering::SeqCst) {
            return Err(EmbedError::InferenceFailed(
                "controlled query failure".into(),
            ));
        }
        Ok(texts.iter().map(|_| axis(0)).collect())
    }

    async fn embed_passage(
        &self,
        texts: &[String],
        _model: EmbeddingModel,
    ) -> Result<Vec<Vec<f32>>, EmbedError> {
        self.0
            .passages
            .lock()
            .expect("passage recorder")
            .extend(texts.iter().cloned());
        Ok(texts
            .iter()
            .map(|text| axis(usize::from(text.contains("Vector Beta"))))
            .collect())
    }

    fn supports_model(&self, _model: EmbeddingModel) -> bool {
        true
    }

    fn name(&self) -> &'static str {
        "a5-role-test"
    }
}

struct RoleProvider(Arc<Calls>);

#[async_trait]
impl EmbedderProvider for RoleProvider {
    fn name(&self) -> &str {
        MODEL
    }

    fn dimensions(&self) -> usize {
        DIM
    }

    async fn build(&self) -> Result<Arc<dyn EmbeddingService>, khive_runtime::RuntimeError> {
        Ok(Arc::new(RoleService(self.0.clone())))
    }
}

fn runtime(calls: &Arc<Calls>) -> KhiveRuntime {
    let rt = KhiveRuntime::new(RuntimeConfig {
        wal_ceiling_bytes: 0,
        wal_ceiling_configured_bytes: 0,
        wal_ceiling_source: Default::default(),
        wal_ceiling_env_raw: None,
        web: Default::default(),
        telemetry: Default::default(),
        mounts: Vec::new(),
        brain: Default::default(),
        git_write: Default::default(),
        display_timezone: khive_runtime::config::resolve_default_display_timezone(),
        events_split: None,
        db_path: None,
        blob_hydration_bytes: khive_runtime::DEFAULT_BLOB_HYDRATION_BYTES,
        default_namespace: Namespace::local(),
        embedding_model: Some(EmbeddingModel::AllMiniLmL6V2),
        additional_embedding_models: vec![],
        gate: Arc::new(AllowAllGate),
        packs: vec!["kg".into(), "knowledge".into()],
        backend_id: BackendId::main(),
        brain_profile: None,
        visible_namespaces: vec![],
        allowed_outbound_namespaces: vec![],
        actor_id: None,
        exec: Default::default(),
        ..khive_runtime::RuntimeConfig::no_embeddings()
    })
    .expect("in-memory runtime");
    rt.register_embedder(RoleProvider(calls.clone()));
    rt
}

fn registry(rt: &KhiveRuntime) -> VerbRegistry {
    let mut builder = VerbRegistryBuilder::new();
    builder.register(KgPack::new(rt.clone()));
    builder.register(crate::KnowledgePack::new(rt.clone()));
    let registry = builder.build().expect("knowledge registry");
    rt.install_edge_rules(registry.all_edge_rules());
    registry
}

async fn fixture() -> (KhiveRuntime, Arc<Calls>, Uuid, Uuid) {
    let calls = Arc::new(Calls::default());
    let rt = runtime(&calls);
    let registry = registry(&rt);
    let upsert = registry
        .dispatch(
            "knowledge.upsert_atoms",
            json!({"atoms": [
                {"slug": "vector-alpha", "name": "Vector Alpha", "content": ALPHA_CONTENT,
                 "tags": ["first", "tag-one"], "finalized": true},
                {"slug": "vector-beta", "name": "Vector Beta", "content": BETA_CONTENT,
                 "tags": ["second", "tag-two"], "finalized": true}
            ]}),
        )
        .await
        .expect("upsert atoms");
    assert_eq!(upsert["created"], 2, "upsert: {upsert}");
    let indexed = registry
        .dispatch("knowledge.index", json!({"rebuild_ann": false}))
        .await
        .expect("index atoms");
    assert_eq!(indexed["indexed"], 2, "index: {indexed}");
    let listed = registry
        .dispatch("knowledge.list", json!({"kind": "atom", "limit": 10}))
        .await
        .expect("list atom ids");
    let rows = listed["results"].as_array().expect("atom rows");
    let id = |slug: &str| -> Uuid {
        rows.iter()
            .find(|row| row["slug"] == slug)
            .and_then(|row| row["id"].as_str())
            .expect("seed atom id")
            .parse()
            .expect("uuid")
    };
    (rt, calls, id("vector-alpha"), id("vector-beta"))
}

async fn search(rt: &KhiveRuntime, rerank: bool) -> Value {
    let token = rt.authorize(Namespace::local()).expect("authorize");
    KnowledgeHandlers::search(
        rt,
        &token,
        json!({"query": "shared lexical rerank evidence", "kind": "atom", "limit": 2,
               "rerank": rerank, "rerank_alpha": 0.0}),
        &vamana::new_shared(),
    )
    .await
    .expect("knowledge search")
}

fn slugs(result: &Value) -> Vec<&str> {
    result["results"]
        .as_array()
        .expect("result rows")
        .iter()
        .map(|row| row["slug"].as_str().expect("slug"))
        .collect()
}

#[tokio::test]
async fn rerank_follows_mutated_stored_vectors_without_reembedding_candidates() {
    let (rt, calls, alpha, beta) = fixture().await;
    calls.passages.lock().expect("passage recorder").clear();
    let first = search(&rt, true).await;
    assert_eq!(slugs(&first), ["vector-alpha", "vector-beta"], "{first}");
    assert_eq!(
        first["rerank_provenance"]["stored_vector_lookup"],
        "supported"
    );
    assert_eq!(first["rerank_provenance"]["candidates"], 2);
    assert_eq!(first["rerank_provenance"]["from_stored"], 2);
    assert_eq!(first["rerank_provenance"]["embedded_fallback"], 0);
    assert!(first["results"]
        .as_array()
        .expect("result rows")
        .iter()
        .all(|hit| hit["score_provenance"]["embedding_rerank"] == true));
    assert!(calls.passages.lock().expect("passage recorder").is_empty());

    let token = rt.authorize(Namespace::local()).expect("authorize");
    let store = rt
        .vectors_for_model(&token, rt.default_embedder_name())
        .expect("vector store");
    store
        .insert(
            alpha,
            SubstrateKind::Entity,
            "local",
            "knowledge.atom",
            vec![axis(1)],
        )
        .await
        .expect("replace alpha vector");
    store
        .insert(
            beta,
            SubstrateKind::Entity,
            "local",
            "knowledge.atom",
            vec![axis(0)],
        )
        .await
        .expect("replace beta vector");

    let second = search(&rt, true).await;
    assert_eq!(slugs(&second), ["vector-beta", "vector-alpha"], "{second}");
    assert_eq!(second["rerank_provenance"]["from_stored"], 2);
    assert_eq!(second["rerank_provenance"]["embedded_fallback"], 0);
    assert!(calls.passages.lock().expect("passage recorder").is_empty());
    assert_eq!(calls.generic.load(Ordering::SeqCst), 0);
}

#[tokio::test]
async fn missing_vector_falls_back_with_document_intent_and_index_text_order() {
    let (rt, calls, _alpha, beta) = fixture().await;
    let indexed_texts = calls.passages.lock().expect("passage recorder").clone();
    let indexed_beta = indexed_texts
        .iter()
        .find(|text| text.starts_with("Vector Beta\n\n"))
        .expect("index-time beta passage")
        .clone();
    assert_eq!(
        indexed_beta,
        format!("Vector Beta\n\n{BETA_CONTENT}\n\nTags: second, tag-two")
    );
    calls.passages.lock().expect("passage recorder").clear();

    let token = rt.authorize(Namespace::local()).expect("authorize");
    let store = rt
        .vectors_for_model(&token, rt.default_embedder_name())
        .expect("vector store");
    assert!(store.delete(beta).await.expect("delete beta vector"));

    let result = search(&rt, true).await;
    assert_eq!(
        result["rerank_provenance"]["stored_vector_lookup"],
        "supported"
    );
    assert_eq!(result["rerank_provenance"]["candidates"], 2);
    assert_eq!(result["rerank_provenance"]["from_stored"], 1);
    assert_eq!(result["rerank_provenance"]["embedded_fallback"], 1);
    assert_eq!(
        *calls.passages.lock().expect("passage recorder"),
        vec![indexed_beta],
        "fallback must use the same tagged renderer as indexing"
    );
    assert_eq!(calls.generic.load(Ordering::SeqCst), 0);
}

#[tokio::test]
async fn domain_mirror_uses_stored_vector_then_tagged_document_fallback() {
    let calls = Arc::new(Calls::default());
    let rt = runtime(&calls);
    let registry = registry(&rt);
    let upsert = registry
        .dispatch(
            "knowledge.upsert_domains",
            json!({"domains": [{
                "slug": "vector-domain",
                "name": "Vector Domain",
                "description": DOMAIN_DESCRIPTION,
                "tags": ["curated", "domain-tag"]
            }]}),
        )
        .await
        .expect("upsert domain and mirror");
    assert_eq!(upsert["created"], 1, "upsert: {upsert}");
    let indexed = registry
        .dispatch("knowledge.index", json!({"rebuild_ann": false}))
        .await
        .expect("index domain mirror");
    assert_eq!(indexed["indexed"], 1, "index: {indexed}");

    let listed = registry
        .dispatch("knowledge.list", json!({"kind": "domain", "limit": 1}))
        .await
        .expect("list canonical domain");
    let domain_id: Uuid = listed["results"][0]["id"]
        .as_str()
        .expect("domain id")
        .parse()
        .expect("uuid");
    let indexed_text =
        format!("Vector Domain\n\n{DOMAIN_DESCRIPTION}\n\nTags: curated, domain-tag, type:domain");
    assert_eq!(
        *calls.passages.lock().expect("passage recorder"),
        vec![indexed_text.clone()],
        "indexing must embed the domain mirror with its type tag"
    );
    calls.passages.lock().expect("passage recorder").clear();

    let token = rt.authorize(Namespace::local()).expect("authorize");
    let ann = vamana::new_shared();
    let params = || {
        json!({"query": "shared lexical domain evidence", "kind": "domain",
               "limit": 1, "rerank": true, "rerank_alpha": 0.0})
    };
    let stored = KnowledgeHandlers::search(&rt, &token, params(), &ann)
        .await
        .expect("search indexed domain");
    assert_eq!(
        stored["results"][0]["id"],
        domain_id.to_string(),
        "{stored}"
    );
    assert_eq!(stored["results"][0]["kind"], "domain", "{stored}");
    assert_eq!(
        stored["rerank_provenance"]["stored_vector_lookup"],
        "supported"
    );
    assert_eq!(stored["rerank_provenance"]["candidates"], 1);
    assert_eq!(stored["rerank_provenance"]["from_stored"], 1);
    assert_eq!(stored["rerank_provenance"]["embedded_fallback"], 0);
    assert!(calls.passages.lock().expect("passage recorder").is_empty());

    let store = rt
        .vectors_for_model(&token, rt.default_embedder_name())
        .expect("vector store");
    assert!(store.delete(domain_id).await.expect("delete mirror vector"));
    let fallback = KnowledgeHandlers::search(&rt, &token, params(), &ann)
        .await
        .expect("search domain without stored vector");
    assert_eq!(
        fallback["results"][0]["id"],
        domain_id.to_string(),
        "{fallback}"
    );
    assert_eq!(fallback["rerank_provenance"]["candidates"], 1);
    assert_eq!(fallback["rerank_provenance"]["from_stored"], 0);
    assert_eq!(fallback["rerank_provenance"]["embedded_fallback"], 1);
    assert_eq!(
        *calls.passages.lock().expect("passage recorder"),
        vec![indexed_text],
        "a mirror domain missing its vector needs the indexed document-intent text"
    );
    assert_eq!(calls.generic.load(Ordering::SeqCst), 0);
}

#[tokio::test]
async fn provenance_is_absent_without_rerank_or_after_query_embedding_failure() {
    let (rt, calls, _alpha, _beta) = fixture().await;
    calls.passages.lock().expect("passage recorder").clear();

    let disabled = search(&rt, false).await;
    assert_eq!(disabled["total"], 2, "{disabled}");
    assert!(disabled.get("rerank_provenance").is_none(), "{disabled}");
    assert!(calls.passages.lock().expect("passage recorder").is_empty());

    calls.fail_query.store(true, Ordering::SeqCst);
    let queries_before_failure = calls.queries.load(Ordering::SeqCst);
    let failed = search(&rt, true).await;
    assert_eq!(failed["total"], 2, "{failed}");
    assert!(failed.get("rerank_provenance").is_none(), "{failed}");
    assert_eq!(
        calls.queries.load(Ordering::SeqCst),
        queries_before_failure + 1
    );
    assert!(calls.passages.lock().expect("passage recorder").is_empty());
}

// The runtime owns a concrete SQLite vector backend. This fake also feeds the
// search rerank's store-injection seam to assert unsupported provenance.
pub(super) struct NoReadStore;

#[async_trait]
impl VectorStore for NoReadStore {
    async fn insert(
        &self,
        _subject_id: Uuid,
        _kind: SubstrateKind,
        _namespace: &str,
        _field: &str,
        _vectors: Vec<Vec<f32>>,
    ) -> StorageResult<()> {
        unreachable!("unused")
    }

    async fn insert_batch(&self, _records: Vec<VectorRecord>) -> StorageResult<BatchWriteSummary> {
        unreachable!("unused")
    }

    async fn delete(&self, _subject_id: Uuid) -> StorageResult<bool> {
        unreachable!("unused")
    }

    async fn count(&self) -> StorageResult<u64> {
        unreachable!("unused")
    }

    async fn search(&self, _request: VectorSearchRequest) -> StorageResult<Vec<VectorSearchHit>> {
        unreachable!("unused")
    }

    async fn info(&self) -> StorageResult<VectorStoreInfo> {
        unreachable!("unused")
    }

    async fn rebuild(&self, _scope: IndexRebuildScope) -> StorageResult<VectorStoreInfo> {
        unreachable!("unused")
    }
}

#[tokio::test]
async fn unsupported_vector_read_remains_an_explicit_capability() {
    let store = NoReadStore;
    assert!(!store.capabilities().supports_vector_read);
    let error = store
        .get_vectors(&[Uuid::new_v4()], "local", "knowledge.atom")
        .await
        .expect_err("default vector read is unsupported");
    assert!(matches!(
        error,
        StorageError::Unsupported { operation, .. } if operation == "get_vectors"
    ));
}