claude-memory-layer 1.0.27 → 1.0.29
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.env.example +7 -0
- package/AGENTS.md +11 -0
- package/README.md +374 -49
- package/benchmarks/replay/anonymized-real-sessions.json +48 -0
- package/dist/cli/index.js +10097 -6003
- package/dist/cli/index.js.map +4 -4
- package/dist/core/index.js +9745 -5587
- package/dist/core/index.js.map +4 -4
- package/dist/hooks/post-tool-use.js +6545 -5270
- package/dist/hooks/post-tool-use.js.map +4 -4
- package/dist/hooks/semantic-daemon.js +6646 -5354
- package/dist/hooks/semantic-daemon.js.map +4 -4
- package/dist/hooks/session-end.js +6618 -5347
- package/dist/hooks/session-end.js.map +4 -4
- package/dist/hooks/session-start.js +6619 -5354
- package/dist/hooks/session-start.js.map +4 -4
- package/dist/hooks/stop.js +6614 -5325
- package/dist/hooks/stop.js.map +4 -4
- package/dist/hooks/user-prompt-submit.js +6702 -5356
- package/dist/hooks/user-prompt-submit.js.map +4 -4
- package/dist/index.js +13537 -0
- package/dist/index.js.map +7 -0
- package/dist/mcp/index.js +20770 -0
- package/dist/mcp/index.js.map +7 -0
- package/dist/server/api/index.js +6632 -5319
- package/dist/server/api/index.js.map +4 -4
- package/dist/server/index.js +6667 -5340
- package/dist/server/index.js.map +4 -4
- package/dist/services/memory-service.js +6568 -5350
- package/dist/services/memory-service.js.map +4 -4
- package/dist/ui/assets/js/bootstrap.js +244 -0
- package/dist/ui/assets/js/chat.js +373 -0
- package/dist/ui/assets/js/disclosure.js +232 -0
- package/dist/ui/assets/js/modals.js +298 -0
- package/dist/ui/assets/js/overview.js +655 -0
- package/dist/ui/assets/js/state.js +72 -0
- package/dist/ui/assets/js/views.js +468 -0
- package/dist/ui/index.html +43 -1
- package/dist/ui/index.ts +3 -0
- package/dist/ui/style.css +222 -0
- package/docs/ARCHITECTURE_COMPARISON_AND_RECOMMENDATIONS.md +627 -0
- package/docs/HERMES_MEMORY_INGESTION_ANALYSIS.md +440 -0
- package/docs/MEMORY_USEFULNESS_AUDIT.md +371 -0
- package/docs/MEMORY_USEFULNESS_AUDIT_RAW.json +80 -0
- package/docs/MEMSEARCH_PROJECT_STRUCTURE_ANALYSIS.md +333 -0
- package/docs/PRODUCT_VALIDATION_MATRIX.md +82 -0
- package/docs/PROJECT_STRUCTURE_ANALYSIS.md +421 -0
- package/docs/REFACTORING_MILESTONES_AND_ISSUES.md +501 -0
- package/docs/REFACTORING_PLAN_THIN_CORE.md +414 -0
- package/docs/REFERENCE_PROJECT_ANALYSES.md +25 -0
- package/docs/SUPERLOCALMEMORY_PROJECT_STRUCTURE_ANALYSIS.md +452 -0
- package/docs/TARGET_ARCHITECTURE_AND_FOLDER_STRUCTURE.md +446 -0
- package/docs/architecture/comparison-index.md +47 -0
- package/docs/reports/codex-real-data-validation-20260505T040447Z.md +46 -0
- package/package.json +12 -5
- package/scripts/build.ts +25 -8
- package/scripts/generate-session-qrels.ts +126 -0
- package/scripts/postinstall-embedding-backend.cjs +142 -0
- package/scripts/replay-retrieval-benchmark.ts +69 -0
- package/specs/thin-core-refactor/context.md +275 -0
- package/specs/thin-core-refactor/plan.md +536 -0
- package/specs/thin-core-refactor/spec.md +465 -0
- package/src/adapters/claude/capture/index.ts +3 -0
- package/src/adapters/claude/context/index.ts +3 -0
- package/src/adapters/claude/hooks/index.ts +21 -0
- package/src/adapters/claude/hooks/post-tool-use.ts +239 -0
- package/src/adapters/claude/hooks/prompt-injection-policy.ts +104 -0
- package/src/adapters/claude/hooks/semantic-daemon-client.ts +209 -0
- package/src/adapters/claude/hooks/semantic-daemon.ts +283 -0
- package/src/adapters/claude/hooks/session-end.ts +59 -0
- package/src/adapters/claude/hooks/session-start.ts +73 -0
- package/src/adapters/claude/hooks/stop.ts +128 -0
- package/src/adapters/claude/hooks/user-prompt-submit.ts +361 -0
- package/src/adapters/claude/index.ts +4 -0
- package/src/adapters/claude/transcript/index.ts +4 -0
- package/src/adapters/claude/transcript/transcript-reader.ts +57 -0
- package/src/adapters/claude/transcript/turn-reconstructor.ts +65 -0
- package/src/apps/cli/claude-settings-hooks.ts +138 -0
- package/src/apps/cli/codex-import-runner.ts +125 -0
- package/src/apps/cli/codex-validation-output.ts +95 -0
- package/src/apps/cli/hermes-import-runner.ts +130 -0
- package/src/apps/cli/hermes-validation-output.ts +91 -0
- package/src/apps/cli/index.ts +1731 -0
- package/src/apps/cli/mcp-install.ts +106 -0
- package/src/apps/cli/retrieval-disclosure-output.ts +196 -0
- package/src/apps/dashboard/assets/js/bootstrap.js +244 -0
- package/src/apps/dashboard/assets/js/chat.js +373 -0
- package/src/apps/dashboard/assets/js/disclosure.js +232 -0
- package/src/apps/dashboard/assets/js/modals.js +298 -0
- package/src/apps/dashboard/assets/js/overview.js +655 -0
- package/src/apps/dashboard/assets/js/state.js +72 -0
- package/src/apps/dashboard/assets/js/views.js +468 -0
- package/src/{ui → apps/dashboard}/index.html +43 -1
- package/src/apps/dashboard/index.ts +3 -0
- package/src/{ui → apps/dashboard}/style.css +222 -0
- package/src/apps/index.ts +5 -0
- package/src/apps/server/api/chat.ts +244 -0
- package/src/apps/server/api/citations.ts +105 -0
- package/src/apps/server/api/events.ts +137 -0
- package/src/apps/server/api/health.ts +53 -0
- package/src/apps/server/api/index.ts +26 -0
- package/src/apps/server/api/projects.ts +74 -0
- package/src/apps/server/api/search.ts +184 -0
- package/src/apps/server/api/sessions.ts +115 -0
- package/src/apps/server/api/stats.ts +723 -0
- package/src/apps/server/api/turns.ts +143 -0
- package/src/apps/server/api/utils.ts +65 -0
- package/src/apps/server/index.ts +111 -0
- package/src/cli/index.ts +2 -1311
- package/src/cli/retrieval-disclosure-output.ts +2 -0
- package/src/compat/index.ts +5 -0
- package/src/core/derive/fact-deriver.ts +170 -0
- package/src/core/derive/index.ts +2 -0
- package/src/core/derive/summary-deriver.ts +76 -0
- package/src/core/embedder.ts +4 -152
- package/src/core/engine/embedding-maintenance-service.ts +187 -0
- package/src/core/engine/endless-memory-services.ts +4 -0
- package/src/core/engine/index.ts +19 -0
- package/src/core/engine/memory-engine-services.ts +170 -0
- package/src/core/engine/memory-ingest-service.ts +317 -0
- package/src/core/engine/memory-query-service.ts +173 -0
- package/src/core/engine/memory-runtime-service.ts +162 -0
- package/src/core/engine/memory-service-composition.ts +231 -0
- package/src/core/engine/retrieval-analytics-service.ts +181 -0
- package/src/core/engine/retrieval-disclosure-service.ts +420 -0
- package/src/core/engine/retrieval-orchestrator.ts +377 -0
- package/src/core/engine/retrieval-services.ts +176 -0
- package/src/core/engine/shared-memory-services.ts +4 -0
- package/src/core/entity-repo.ts +1 -3
- package/src/core/event-store.ts +3 -3
- package/src/core/evidence-aligner.ts +2 -2
- package/src/core/external-market-context.ts +582 -0
- package/src/core/graduation.ts +2 -3
- package/src/core/index.ts +21 -0
- package/src/core/matcher.ts +2 -4
- package/src/core/model/memory-fact.ts +30 -0
- package/src/core/model/memory-rule.ts +14 -0
- package/src/core/model/memory-summary.ts +21 -0
- package/src/core/model/raw-event.ts +28 -0
- package/src/core/model/retrieval-result.ts +35 -0
- package/src/core/privacy/filter.ts +21 -10
- package/src/core/product-validation-matrix.ts +314 -0
- package/src/core/progressive-retriever.ts +1 -2
- package/src/core/registry/project-path.ts +54 -0
- package/src/core/registry/session-registry.ts +69 -0
- package/src/core/replay-evaluator.ts +625 -0
- package/src/core/retrieval-benchmark.ts +117 -0
- package/src/core/retrieval-quality.ts +109 -0
- package/src/core/retriever.ts +53 -15
- package/src/core/session-qrels.ts +360 -0
- package/src/core/shared-event-store.ts +1 -1
- package/src/core/sqlite-event-store.ts +35 -11
- package/src/core/task/blocker-resolver.ts +2 -2
- package/src/core/task/task-resolver.ts +0 -1
- package/src/core/vector-outbox.ts +1 -10
- package/src/core/vector-worker.ts +1 -1
- package/src/extensions/endless-memory/endless-memory-services.ts +350 -0
- package/src/extensions/endless-memory/index.ts +1 -0
- package/src/extensions/index.ts +5 -0
- package/src/extensions/mcp/handlers.ts +960 -0
- package/src/extensions/mcp/index.ts +48 -0
- package/src/extensions/mcp/tools.ts +252 -0
- package/src/extensions/shared-memory/index.ts +1 -0
- package/src/extensions/shared-memory/shared-memory-services.ts +211 -0
- package/src/extensions/vector/embedder.ts +197 -0
- package/src/extensions/vector/index.ts +1 -0
- package/src/hooks/post-tool-use.ts +3 -236
- package/src/hooks/semantic-daemon-client.ts +1 -208
- package/src/hooks/semantic-daemon.ts +6 -271
- package/src/hooks/session-end.ts +4 -79
- package/src/hooks/session-start.ts +4 -73
- package/src/hooks/stop.ts +3 -173
- package/src/hooks/user-prompt-submit.ts +3 -338
- package/src/index.ts +13 -0
- package/src/mcp/handlers.ts +2 -212
- package/src/mcp/index.ts +3 -46
- package/src/mcp/tools.ts +2 -78
- package/src/server/api/chat.ts +2 -244
- package/src/server/api/citations.ts +2 -105
- package/src/server/api/events.ts +2 -137
- package/src/server/api/health.ts +2 -53
- package/src/server/api/index.ts +2 -26
- package/src/server/api/projects.ts +2 -74
- package/src/server/api/search.ts +2 -102
- package/src/server/api/sessions.ts +2 -115
- package/src/server/api/stats.ts +2 -724
- package/src/server/api/turns.ts +2 -143
- package/src/server/api/utils.ts +2 -46
- package/src/server/index.ts +2 -100
- package/src/services/bootstrap-organizer.ts +46 -26
- package/src/services/codex-session-history-importer.ts +521 -29
- package/src/services/hermes-session-history-importer.ts +733 -0
- package/src/services/memory-service-config.ts +36 -0
- package/src/services/memory-service-registry.ts +150 -0
- package/src/services/memory-service.ts +211 -1325
- package/src/services/session-history-importer.ts +58 -14
- package/tests/README.md +23 -0
- package/tests/adapters/claude/claude-semantic-daemon-adapter.test.ts +54 -0
- package/tests/adapters/claude/claude-transcript-reconstructor.test.ts +98 -0
- package/tests/adapters/claude-hook-prompt-injection-policy.test.ts +99 -0
- package/tests/apps/app-layer-boundary.test.ts +48 -0
- package/tests/apps/claude-settings-hooks.test.ts +107 -0
- package/tests/apps/cli-disclosure-output.test.ts +212 -0
- package/tests/apps/codex-import-runner.test.ts +99 -0
- package/tests/apps/codex-validation-output.test.ts +100 -0
- package/tests/apps/hermes-import-runner.test.ts +99 -0
- package/tests/apps/mcp-install-command.test.ts +59 -0
- package/tests/apps/package-build-entrypoints.test.ts +30 -0
- package/tests/apps/postinstall-embedding-backend.test.ts +167 -0
- package/tests/apps/search-api-disclosure.test.ts +162 -0
- package/tests/apps/stats-api-lightweight.test.ts +67 -0
- package/tests/apps/ui-disclosure-output.test.ts +140 -0
- package/tests/{bootstrap-organizer.test.ts → core/bootstrap-organizer.test.ts} +1 -1
- package/tests/{canonical-key.test.ts → core/canonical-key.test.ts} +1 -1
- package/tests/core/codex-session-history-importer-validation.test.ts +185 -0
- package/tests/{consolidation-worker.test.ts → core/consolidation-worker.test.ts} +2 -2
- package/tests/core/embedding-maintenance-service.test.ts +282 -0
- package/tests/{evidence-aligner.test.ts → core/evidence-aligner.test.ts} +1 -1
- package/tests/core/external-market-context.test.ts +209 -0
- package/tests/core/fact-deriver.test.ts +79 -0
- package/tests/core/hermes-session-history-importer-validation.test.ts +609 -0
- package/tests/{ingest-interceptor.test.ts → core/ingest-interceptor.test.ts} +1 -1
- package/tests/{markdown-mirror.test.ts → core/markdown-mirror.test.ts} +2 -2
- package/tests/{matcher.test.ts → core/matcher.test.ts} +1 -1
- package/tests/{md-mirror.test.ts → core/md-mirror.test.ts} +2 -2
- package/tests/core/memory-engine-services.test.ts +240 -0
- package/tests/core/memory-ingest-service.test.ts +296 -0
- package/tests/core/memory-query-service.test.ts +129 -0
- package/tests/core/memory-runtime-service.test.ts +201 -0
- package/tests/core/memory-service-composition.test.ts +192 -0
- package/tests/core/memory-service-config.test.ts +41 -0
- package/tests/core/memory-service-facade.test.ts +30 -0
- package/tests/core/memory-service-registry.test.ts +206 -0
- package/tests/core/product-validation-matrix.test.ts +61 -0
- package/tests/core/project-registry.test.ts +78 -0
- package/tests/core/replay-evaluator.test.ts +181 -0
- package/tests/core/retrieval-analytics-service.test.ts +210 -0
- package/tests/core/retrieval-benchmark.test.ts +93 -0
- package/tests/core/retrieval-disclosure-service.test.ts +264 -0
- package/tests/core/retrieval-orchestrator.test.ts +403 -0
- package/tests/core/retrieval-quality.test.ts +31 -0
- package/tests/core/retrieval-services.test.ts +185 -0
- package/tests/{retriever-fallback-chain.test.ts → core/retriever-fallback-chain.test.ts} +3 -3
- package/tests/{retriever-strategy-scope.test.ts → core/retriever-strategy-scope.test.ts} +70 -3
- package/tests/{retriever.memu-adoption.test.ts → core/retriever.memu-adoption.test.ts} +3 -3
- package/tests/core/session-history-importer-filter.test.ts +78 -0
- package/tests/core/session-qrels.test.ts +250 -0
- package/tests/{sqlite-event-store-replication.test.ts → core/sqlite-event-store-replication.test.ts} +36 -1
- package/tests/core/summary-deriver.test.ts +66 -0
- package/tests/extensions/embedder-warning-suppression.test.ts +53 -0
- package/tests/extensions/endless-memory-extension-boundary.test.ts +17 -0
- package/tests/extensions/endless-memory-services.test.ts +325 -0
- package/tests/extensions/mcp-context-tools.test.ts +905 -0
- package/tests/extensions/mcp-extension-boundary.test.ts +21 -0
- package/tests/extensions/mcp-package-build.test.ts +22 -0
- package/tests/extensions/mcp-project-aware-tools.test.ts +102 -0
- package/tests/extensions/shared-memory-extension-boundary.test.ts +24 -0
- package/tests/extensions/shared-memory-services.test.ts +309 -0
- package/tests/extensions/vector-extension-boundary.test.ts +21 -0
- package/.claude/settings.local.json +0 -25
- package/.npm-cache/_cacache/content-v2/sha512/04/76/c098f88dfe584a2b80870bff7421b05d17d3d9ee1027f77772332a22d3f93a9a57101a2855107f6ad82077a818bba912b2bc317f2361b5ddb09ad284d9ce +0 -0
- package/.npm-cache/_cacache/content-v2/sha512/60/25/d2ecd39cfc7cab58351162814be77f935c6d6491c10c3745d456da7ddb2117ffd90c10e53fe3c0f1ed16b403307841543634504398b16ee4e6b6dd8e0c45 +0 -0
- package/.npm-cache/_cacache/index-v5/2b/9a/7f8f40206ed8a2e0a84efaa953ccaed1f5d001e14b931083f2e7a0738007 +0 -2
- package/.npm-cache/_cacache/index-v5/2e/d9/fcfa5c6a6abdc2a3644ab84a95936047298c465a2f47ee03db8f7fe1e946 +0 -3
- package/.npm-cache/_cacache/index-v5/a9/42/e519633356d12d3d2f19da66a8301016d496c8f5c3e0554124aaa62dc043 +0 -2
- package/.npm-cache/_logs/2026-02-26T12_04_52_729Z-debug-0.log +0 -256
- package/.npm-cache/_logs/2026-02-26T12_05_36_835Z-debug-0.log +0 -18
- package/.npm-cache/_logs/2026-02-26T12_05_45_982Z-debug-0.log +0 -32
- package/.npm-cache/_logs/2026-02-26T12_05_48_515Z-debug-0.log +0 -260
- package/.npm-cache/_logs/2026-02-26T12_05_53_567Z-debug-0.log +0 -69
- package/.npm-cache/_update-notifier-last-checked +0 -0
- package/bootstrap-kb/decisions/decisions.md +0 -244
- package/bootstrap-kb/glossary/glossary.md +0 -46
- package/bootstrap-kb/modules/.claude-plugin.md +0 -22
- package/bootstrap-kb/modules/agents.md.md +0 -15
- package/bootstrap-kb/modules/claude.md.md +0 -15
- package/bootstrap-kb/modules/context.md.md +0 -15
- package/bootstrap-kb/modules/docs.md +0 -18
- package/bootstrap-kb/modules/handoff.md.md +0 -15
- package/bootstrap-kb/modules/package-lock.json.md +0 -15
- package/bootstrap-kb/modules/package.json.md +0 -15
- package/bootstrap-kb/modules/plan.md.md +0 -15
- package/bootstrap-kb/modules/readme.md.md +0 -15
- package/bootstrap-kb/modules/scripts.md +0 -26
- package/bootstrap-kb/modules/spec.md.md +0 -15
- package/bootstrap-kb/modules/specs.md +0 -20
- package/bootstrap-kb/modules/src.md +0 -51
- package/bootstrap-kb/modules/tests.md +0 -42
- package/bootstrap-kb/modules/tsconfig.json.md +0 -15
- package/bootstrap-kb/modules/vitest.config.ts.md +0 -15
- package/bootstrap-kb/overview/overview.md +0 -40
- package/bootstrap-kb/sources/manifest.json +0 -950
- package/bootstrap-kb/sources/manifest.md +0 -227
- package/bootstrap-kb/timeline/timeline.md +0 -57
- package/claude-memory-layer-1.0.14.tgz +0 -0
- package/d.sh +0 -3
- package/deploy.sh +0 -3
- package/dist/ui/app.js +0 -2101
- package/memory/.claude-plugin/commands/2026-02-25.md +0 -263
- package/memory/_index.md +0 -419
- package/memory/agent_response/uncategorized/2026-02-26.md +0 -176
- package/memory/agent_response/uncategorized/2026-03-03.md +0 -14
- package/memory/agent_response/uncategorized/2026-03-04.md +0 -1421
- package/memory/agent_response/uncategorized/2026-03-05.md +0 -157
- package/memory/default/uncategorized/2026-02-25.md +0 -4839
- package/memory/session_summary/uncategorized/2026-02-26.md +0 -13
- package/memory/session_summary/uncategorized/2026-03-03.md +0 -5
- package/memory/session_summary/uncategorized/2026-03-04.md +0 -50
- package/memory/specs/20260207-dashboard-upgrade/2026-02-25.md +0 -142
- package/memory/specs/citations-system/2026-02-25.md +0 -1121
- package/memory/specs/endless-mode/2026-02-25.md +0 -1392
- package/memory/specs/entity-edge-model/2026-02-25.md +0 -1263
- package/memory/specs/evidence-aligner-v2/2026-02-25.md +0 -1028
- package/memory/specs/mcp-desktop-integration/2026-02-25.md +0 -1334
- package/memory/specs/post-tool-use-hook/2026-02-25.md +0 -1164
- package/memory/specs/private-tags/2026-02-25.md +0 -1057
- package/memory/specs/progressive-disclosure/2026-02-25.md +0 -1436
- package/memory/specs/task-entity-system/2026-02-25.md +0 -924
- package/memory/specs/vector-outbox-v2/2026-02-25.md +0 -1510
- package/memory/specs/web-viewer-ui/2026-02-25.md +0 -1709
- package/memory/tool_observation/uncategorized/2026-02-26.md +0 -209
- package/memory/tool_observation/uncategorized/2026-03-03.md +0 -21
- package/memory/tool_observation/uncategorized/2026-03-04.md +0 -1033
- package/memory/tool_observation/uncategorized/2026-03-05.md +0 -33
- package/memory/user_prompt/uncategorized/2026-02-26.md +0 -25
- package/memory/user_prompt/uncategorized/2026-03-04.md +0 -634
- package/memory/user_prompt/uncategorized/2026-03-05.md +0 -6
- package/specs/optional-duckdb/context.md +0 -77
- package/specs/optional-duckdb/plan.md +0 -142
- package/specs/optional-duckdb/spec.md +0 -35
- package/src/ui/app.js +0 -2101
|
@@ -0,0 +1,185 @@
|
|
|
1
|
+
import { describe, expect, it } from 'vitest';
|
|
2
|
+
import {
|
|
3
|
+
createRetrievalServices,
|
|
4
|
+
type RetrievalEventStore
|
|
5
|
+
} from '../../src/core/engine/retrieval-services.js';
|
|
6
|
+
import type { Embedder } from '../../src/core/embedder.js';
|
|
7
|
+
import type { Matcher } from '../../src/core/matcher.js';
|
|
8
|
+
import type { Retriever, UnifiedRetrievalResult } from '../../src/core/retriever.js';
|
|
9
|
+
import type { MemoryEvent } from '../../src/core/types.js';
|
|
10
|
+
import type { VectorStore } from '../../src/core/vector-store.js';
|
|
11
|
+
|
|
12
|
+
function event(id: string): MemoryEvent {
|
|
13
|
+
return {
|
|
14
|
+
id,
|
|
15
|
+
sessionId: 's1',
|
|
16
|
+
eventType: 'agent_response',
|
|
17
|
+
content: 'retrieval factory keeps MemoryService thin',
|
|
18
|
+
canonicalKey: `test/${id}`,
|
|
19
|
+
dedupeKey: `s1:${id}`,
|
|
20
|
+
timestamp: new Date('2026-02-25T00:00:00.000Z'),
|
|
21
|
+
metadata: {}
|
|
22
|
+
};
|
|
23
|
+
}
|
|
24
|
+
|
|
25
|
+
function retrievalResult(id = 'e1'): UnifiedRetrievalResult {
|
|
26
|
+
const memoryEvent = event(id);
|
|
27
|
+
return {
|
|
28
|
+
memories: [{ event: memoryEvent, score: 0.92, sessionContext: 'nearby context' }],
|
|
29
|
+
matchResult: {
|
|
30
|
+
match: { event: memoryEvent, score: 0.92 },
|
|
31
|
+
confidence: 'high'
|
|
32
|
+
},
|
|
33
|
+
totalTokens: 10,
|
|
34
|
+
context: memoryEvent.content,
|
|
35
|
+
fallbackTrace: ['stage:primary:fast'],
|
|
36
|
+
selectedDebug: [{ eventId: id, score: 0.92, semanticScore: 0.8, lexicalScore: 0.4, recencyScore: 0.1 }],
|
|
37
|
+
candidateDebug: [{ eventId: id, score: 0.92, semanticScore: 0.8, lexicalScore: 0.4, recencyScore: 0.1 }]
|
|
38
|
+
};
|
|
39
|
+
}
|
|
40
|
+
|
|
41
|
+
describe('createRetrievalServices', () => {
|
|
42
|
+
it('builds one retriever and wires the retrieval service bundle to shared ports', async () => {
|
|
43
|
+
let initializeCalls = 0;
|
|
44
|
+
const traceInputs: Array<Record<string, unknown>> = [];
|
|
45
|
+
const retrieveCalls: Array<{ query: string; options: Record<string, unknown> }> = [];
|
|
46
|
+
const traceStats = {
|
|
47
|
+
totalQueries: 2,
|
|
48
|
+
avgCandidateCount: 1,
|
|
49
|
+
avgSelectedCount: 1,
|
|
50
|
+
selectionRate: 1
|
|
51
|
+
};
|
|
52
|
+
const vectorStore = { marker: 'vector' } as unknown as VectorStore;
|
|
53
|
+
const embedder = { marker: 'embedder' } as unknown as Embedder;
|
|
54
|
+
const matcher = { marker: 'matcher' } as unknown as Matcher;
|
|
55
|
+
const store = {
|
|
56
|
+
getHelpfulnessStats: async () => ({
|
|
57
|
+
avgScore: 0.8,
|
|
58
|
+
totalEvaluated: 25,
|
|
59
|
+
totalRetrievals: 30,
|
|
60
|
+
helpful: 20,
|
|
61
|
+
neutral: 3,
|
|
62
|
+
unhelpful: 2
|
|
63
|
+
}),
|
|
64
|
+
recordRetrievalTrace: async (input: Record<string, unknown>) => { traceInputs.push(input); },
|
|
65
|
+
incrementAccessCount: async (_eventIds: string[]) => {},
|
|
66
|
+
recordRetrieval: async (_eventId: string, _sessionId: string, _score: number, _query: string) => {},
|
|
67
|
+
getEvent: async (id: string) => event(id),
|
|
68
|
+
getSessionEvents: async (_sessionId: string) => [event('e1')],
|
|
69
|
+
getRecentEvents: async (_limit?: number) => [event('e1')],
|
|
70
|
+
getRetrievalTraceStats: async () => traceStats,
|
|
71
|
+
getRecentRetrievalTraces: async () => [],
|
|
72
|
+
getMostAccessed: async () => [],
|
|
73
|
+
evaluateSessionHelpfulness: async (_sessionId: string) => {},
|
|
74
|
+
getUnevaluatedSessions: async () => [],
|
|
75
|
+
getHelpfulMemories: async () => []
|
|
76
|
+
};
|
|
77
|
+
let createArgs: {
|
|
78
|
+
eventStore: unknown;
|
|
79
|
+
vectorStore: unknown;
|
|
80
|
+
embedder: unknown;
|
|
81
|
+
matcher: unknown;
|
|
82
|
+
} | null = null;
|
|
83
|
+
let registeredRewriter: ((query: string) => Promise<string | null>) | null = null;
|
|
84
|
+
const fakeRetriever = {
|
|
85
|
+
setQueryRewriter(rewriter: (query: string) => Promise<string | null>) {
|
|
86
|
+
registeredRewriter = rewriter;
|
|
87
|
+
},
|
|
88
|
+
async retrieve(query: string, options: Record<string, unknown>) {
|
|
89
|
+
retrieveCalls.push({ query, options });
|
|
90
|
+
return retrievalResult('e1');
|
|
91
|
+
}
|
|
92
|
+
} as unknown as Retriever;
|
|
93
|
+
|
|
94
|
+
const services = createRetrievalServices({
|
|
95
|
+
initialize: async () => { initializeCalls += 1; },
|
|
96
|
+
eventStore: store,
|
|
97
|
+
vectorStore,
|
|
98
|
+
embedder,
|
|
99
|
+
matcher,
|
|
100
|
+
getProjectHash: () => 'project-1',
|
|
101
|
+
hasSharedStore: () => false,
|
|
102
|
+
createRetriever: (eventStore, vectorStoreArg, embedderArg, matcherArg) => {
|
|
103
|
+
createArgs = {
|
|
104
|
+
eventStore,
|
|
105
|
+
vectorStore: vectorStoreArg,
|
|
106
|
+
embedder: embedderArg,
|
|
107
|
+
matcher: matcherArg
|
|
108
|
+
};
|
|
109
|
+
return fakeRetriever;
|
|
110
|
+
}
|
|
111
|
+
});
|
|
112
|
+
|
|
113
|
+
expect(services.retriever).toBe(fakeRetriever);
|
|
114
|
+
expect(createArgs).toEqual({ eventStore: store, vectorStore, embedder, matcher });
|
|
115
|
+
expect(registeredRewriter).toEqual(expect.any(Function));
|
|
116
|
+
|
|
117
|
+
await services.retrievalOrchestrator.retrieveMemories('thin core', { sessionId: 's1', topK: 2 });
|
|
118
|
+
|
|
119
|
+
expect(retrieveCalls[0]).toMatchObject({
|
|
120
|
+
query: 'thin core',
|
|
121
|
+
options: {
|
|
122
|
+
topK: 2,
|
|
123
|
+
projectHash: 'project-1',
|
|
124
|
+
projectScopeMode: 'strict'
|
|
125
|
+
}
|
|
126
|
+
});
|
|
127
|
+
expect(traceInputs[0]).toMatchObject({
|
|
128
|
+
sessionId: 's1',
|
|
129
|
+
projectHash: 'project-1',
|
|
130
|
+
queryText: 'thin core',
|
|
131
|
+
selectedEventIds: ['e1'],
|
|
132
|
+
candidateEventIds: ['e1'],
|
|
133
|
+
confidence: 'high'
|
|
134
|
+
});
|
|
135
|
+
|
|
136
|
+
const disclosure = await services.retrievalDisclosureService.search('thin core disclosure', { topK: 1 });
|
|
137
|
+
expect(disclosure.results[0]).toMatchObject({
|
|
138
|
+
id: 'event:e1',
|
|
139
|
+
sourceRef: 'event:e1',
|
|
140
|
+
sessionId: 's1'
|
|
141
|
+
});
|
|
142
|
+
|
|
143
|
+
await expect(services.retrievalAnalyticsService.getRetrievalTraceStats()).resolves.toEqual(traceStats);
|
|
144
|
+
expect(initializeCalls).toBe(3);
|
|
145
|
+
});
|
|
146
|
+
|
|
147
|
+
it('rejects default retriever stores that lack retriever read methods', () => {
|
|
148
|
+
const incompleteStore = {
|
|
149
|
+
recordRetrievalTrace: async (_input: Record<string, unknown>) => {},
|
|
150
|
+
incrementAccessCount: async (_eventIds: string[]) => {},
|
|
151
|
+
recordRetrieval: async (_eventId: string, _sessionId: string, _score: number, _query: string) => {},
|
|
152
|
+
getEvent: async (id: string) => event(id),
|
|
153
|
+
getSessionEvents: async (_sessionId: string) => [event('e1')],
|
|
154
|
+
getRetrievalTraceStats: async () => ({
|
|
155
|
+
totalQueries: 0,
|
|
156
|
+
avgCandidateCount: 0,
|
|
157
|
+
avgSelectedCount: 0,
|
|
158
|
+
selectionRate: 0
|
|
159
|
+
}),
|
|
160
|
+
getRecentRetrievalTraces: async () => [],
|
|
161
|
+
getMostAccessed: async () => [],
|
|
162
|
+
evaluateSessionHelpfulness: async (_sessionId: string) => {},
|
|
163
|
+
getUnevaluatedSessions: async () => [],
|
|
164
|
+
getHelpfulMemories: async () => [],
|
|
165
|
+
getHelpfulnessStats: async () => ({
|
|
166
|
+
avgScore: 0,
|
|
167
|
+
totalEvaluated: 0,
|
|
168
|
+
totalRetrievals: 0,
|
|
169
|
+
helpful: 0,
|
|
170
|
+
neutral: 0,
|
|
171
|
+
unhelpful: 0
|
|
172
|
+
})
|
|
173
|
+
} as unknown as RetrievalEventStore;
|
|
174
|
+
|
|
175
|
+
expect(() => createRetrievalServices({
|
|
176
|
+
initialize: async () => {},
|
|
177
|
+
eventStore: incompleteStore,
|
|
178
|
+
vectorStore: { marker: 'vector' } as unknown as VectorStore,
|
|
179
|
+
embedder: { marker: 'embedder' } as unknown as Embedder,
|
|
180
|
+
matcher: { marker: 'matcher' } as unknown as Matcher,
|
|
181
|
+
getProjectHash: () => null,
|
|
182
|
+
hasSharedStore: () => false
|
|
183
|
+
})).toThrow(/getRecentEvents/);
|
|
184
|
+
});
|
|
185
|
+
});
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
import { describe, it, expect } from 'vitest';
|
|
2
|
-
import { Retriever } from '
|
|
3
|
-
import { Matcher } from '
|
|
4
|
-
import type { MemoryEvent } from '
|
|
2
|
+
import { Retriever } from '../../src/core/retriever.js';
|
|
3
|
+
import { Matcher } from '../../src/core/matcher.js';
|
|
4
|
+
import type { MemoryEvent } from '../../src/core/types.js';
|
|
5
5
|
|
|
6
6
|
function ev(id: string, content: string): MemoryEvent {
|
|
7
7
|
return {
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
import { describe, it, expect } from 'vitest';
|
|
2
|
-
import { Retriever } from '
|
|
3
|
-
import { Matcher } from '
|
|
4
|
-
import type { MemoryEvent } from '
|
|
2
|
+
import { Retriever } from '../../src/core/retriever.js';
|
|
3
|
+
import { Matcher } from '../../src/core/matcher.js';
|
|
4
|
+
import type { MemoryEvent } from '../../src/core/types.js';
|
|
5
5
|
|
|
6
6
|
function ev(id: string, sessionId: string, eventType: MemoryEvent['eventType'], content: string, canonicalKey: string): MemoryEvent {
|
|
7
7
|
return {
|
|
@@ -94,4 +94,71 @@ describe('Retriever strategy/scope', () => {
|
|
|
94
94
|
expect(out.memories.length).toBe(1);
|
|
95
95
|
expect(out.memories[0].event.id).toBe('e1');
|
|
96
96
|
});
|
|
97
|
+
|
|
98
|
+
it('returns no memories for command-artifact queries instead of semantic false positives', async () => {
|
|
99
|
+
const retriever = new Retriever(fakeEventStore as any, fakeVectorStore as any, fakeEmbedder as any, new Matcher());
|
|
100
|
+
|
|
101
|
+
const out = await retriever.retrieve('local-command-stdout command-name opus', {
|
|
102
|
+
strategy: 'deep',
|
|
103
|
+
topK: 5,
|
|
104
|
+
minScore: 0.1,
|
|
105
|
+
includeSessionContext: false
|
|
106
|
+
});
|
|
107
|
+
|
|
108
|
+
expect(out.memories).toEqual([]);
|
|
109
|
+
expect(out.fallbackTrace).toContain('guard:command-artifact-query');
|
|
110
|
+
});
|
|
111
|
+
|
|
112
|
+
it('filters technical identifier queries when candidates have no exact technical overlap', async () => {
|
|
113
|
+
const retriever = new Retriever(fakeEventStore as any, fakeVectorStore as any, fakeEmbedder as any, new Matcher());
|
|
114
|
+
|
|
115
|
+
const out = await retriever.retrieve('DuckDB legacy storage migrate', {
|
|
116
|
+
strategy: 'deep',
|
|
117
|
+
topK: 5,
|
|
118
|
+
minScore: 0.1,
|
|
119
|
+
includeSessionContext: false
|
|
120
|
+
});
|
|
121
|
+
|
|
122
|
+
expect(out.memories).toEqual([]);
|
|
123
|
+
});
|
|
124
|
+
|
|
125
|
+
it('keeps technical identifier matches when candidate content contains the identifier', async () => {
|
|
126
|
+
const technical = ev(
|
|
127
|
+
'e3',
|
|
128
|
+
'agent:main:alpha',
|
|
129
|
+
'agent_response',
|
|
130
|
+
'SQLite FTS failure no such column T.event_id is fixed with event_id UNINDEXED.',
|
|
131
|
+
'fix/sqlite/fts'
|
|
132
|
+
);
|
|
133
|
+
const eventStore = {
|
|
134
|
+
...fakeEventStore,
|
|
135
|
+
async getEvent(id: string) {
|
|
136
|
+
if (id === 'e3') return technical;
|
|
137
|
+
return fakeEventStore.getEvent(id);
|
|
138
|
+
}
|
|
139
|
+
};
|
|
140
|
+
const vectorStore = {
|
|
141
|
+
async search() {
|
|
142
|
+
return [{
|
|
143
|
+
id: 'v3',
|
|
144
|
+
eventId: 'e3',
|
|
145
|
+
content: technical.content,
|
|
146
|
+
score: 0.93,
|
|
147
|
+
sessionId: technical.sessionId,
|
|
148
|
+
eventType: technical.eventType,
|
|
149
|
+
timestamp: technical.timestamp.toISOString()
|
|
150
|
+
}];
|
|
151
|
+
}
|
|
152
|
+
};
|
|
153
|
+
const retriever = new Retriever(eventStore as any, vectorStore as any, fakeEmbedder as any, new Matcher());
|
|
154
|
+
|
|
155
|
+
const out = await retriever.retrieve('T.event_id FTS rebuild', {
|
|
156
|
+
strategy: 'deep',
|
|
157
|
+
topK: 5,
|
|
158
|
+
minScore: 0.1,
|
|
159
|
+
includeSessionContext: false
|
|
160
|
+
});
|
|
161
|
+
|
|
162
|
+
expect(out.memories.map((memory) => memory.event.id)).toEqual(['e3']);
|
|
163
|
+
});
|
|
97
164
|
});
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
import { describe, it, expect } from 'vitest';
|
|
2
|
-
import { Retriever } from '
|
|
3
|
-
import type { MemoryEvent, MatchResult } from '
|
|
4
|
-
import type { SearchResult } from '
|
|
2
|
+
import { Retriever } from '../../src/core/retriever.js';
|
|
3
|
+
import type { MemoryEvent, MatchResult } from '../../src/core/types.js';
|
|
4
|
+
import type { SearchResult } from '../../src/core/vector-store.js';
|
|
5
5
|
|
|
6
6
|
class FakeEventStore {
|
|
7
7
|
constructor(private readonly events: Record<string, MemoryEvent>) {}
|
|
@@ -0,0 +1,78 @@
|
|
|
1
|
+
import { mkdtempSync, rmSync, utimesSync, writeFileSync } from 'node:fs';
|
|
2
|
+
import { tmpdir } from 'node:os';
|
|
3
|
+
import { join } from 'node:path';
|
|
4
|
+
import { afterEach, describe, expect, it, vi } from 'vitest';
|
|
5
|
+
|
|
6
|
+
import {
|
|
7
|
+
isClaudeLocalCommandArtifact,
|
|
8
|
+
isWorthStoringPrompt,
|
|
9
|
+
SessionHistoryImporter,
|
|
10
|
+
type ImportResult
|
|
11
|
+
} from '../../src/services/session-history-importer.js';
|
|
12
|
+
|
|
13
|
+
const tempDirs: string[] = [];
|
|
14
|
+
|
|
15
|
+
function tempDir() {
|
|
16
|
+
const dir = mkdtempSync(join(tmpdir(), 'cml-session-importer-'));
|
|
17
|
+
tempDirs.push(dir);
|
|
18
|
+
return dir;
|
|
19
|
+
}
|
|
20
|
+
|
|
21
|
+
function makeImportResult(overrides: Partial<ImportResult> = {}): ImportResult {
|
|
22
|
+
return {
|
|
23
|
+
totalSessions: 1,
|
|
24
|
+
totalMessages: 1,
|
|
25
|
+
importedPrompts: 1,
|
|
26
|
+
importedResponses: 0,
|
|
27
|
+
skippedDuplicates: 0,
|
|
28
|
+
errors: [],
|
|
29
|
+
...overrides
|
|
30
|
+
};
|
|
31
|
+
}
|
|
32
|
+
|
|
33
|
+
afterEach(() => {
|
|
34
|
+
for (const dir of tempDirs.splice(0)) {
|
|
35
|
+
rmSync(dir, { recursive: true, force: true });
|
|
36
|
+
}
|
|
37
|
+
});
|
|
38
|
+
|
|
39
|
+
describe('session history importer prompt filtering', () => {
|
|
40
|
+
it('drops Claude local-command artifacts that dilute retrieval quality', () => {
|
|
41
|
+
const artifact = `<command-name>/model</command-name>\n<local-command-stdout>Using model opus</local-command-stdout>`;
|
|
42
|
+
|
|
43
|
+
expect(isClaudeLocalCommandArtifact(artifact)).toBe(true);
|
|
44
|
+
expect(isWorthStoringPrompt(artifact)).toBe(false);
|
|
45
|
+
});
|
|
46
|
+
|
|
47
|
+
it('keeps substantive imported user prompts', () => {
|
|
48
|
+
expect(isWorthStoringPrompt('이 프로젝트에서 memory retrieval 구조를 더 가볍게 개선해줘')).toBe(true);
|
|
49
|
+
});
|
|
50
|
+
|
|
51
|
+
it('imports only the most recently modified matching Claude session when sessionLimit is set', async () => {
|
|
52
|
+
const dir = tempDir();
|
|
53
|
+
const older = join(dir, 'older.jsonl');
|
|
54
|
+
const newest = join(dir, 'newest.jsonl');
|
|
55
|
+
const oldest = join(dir, 'oldest.jsonl');
|
|
56
|
+
for (const filePath of [older, newest, oldest]) {
|
|
57
|
+
writeFileSync(filePath, '{}\n', 'utf8');
|
|
58
|
+
}
|
|
59
|
+
utimesSync(oldest, new Date('2026-05-01T00:00:00.000Z'), new Date('2026-05-01T00:00:00.000Z'));
|
|
60
|
+
utimesSync(older, new Date('2026-05-02T00:00:00.000Z'), new Date('2026-05-02T00:00:00.000Z'));
|
|
61
|
+
utimesSync(newest, new Date('2026-05-03T00:00:00.000Z'), new Date('2026-05-03T00:00:00.000Z'));
|
|
62
|
+
|
|
63
|
+
const importer = new SessionHistoryImporter({} as never) as SessionHistoryImporter & {
|
|
64
|
+
findProjectDirs: ReturnType<typeof vi.fn>;
|
|
65
|
+
findSessionFiles: ReturnType<typeof vi.fn>;
|
|
66
|
+
importSessionFile: ReturnType<typeof vi.fn>;
|
|
67
|
+
};
|
|
68
|
+
importer.findProjectDirs = vi.fn(async () => [dir]);
|
|
69
|
+
importer.findSessionFiles = vi.fn(async () => [older, newest, oldest]);
|
|
70
|
+
importer.importSessionFile = vi.fn(async () => makeImportResult());
|
|
71
|
+
|
|
72
|
+
const result = await importer.importProject('/repo/current', { sessionLimit: 1 });
|
|
73
|
+
|
|
74
|
+
expect(result.totalSessions).toBe(1);
|
|
75
|
+
expect(importer.importSessionFile).toHaveBeenCalledTimes(1);
|
|
76
|
+
expect(importer.importSessionFile).toHaveBeenCalledWith(newest, expect.objectContaining({ sessionLimit: 1 }));
|
|
77
|
+
});
|
|
78
|
+
});
|
|
@@ -0,0 +1,250 @@
|
|
|
1
|
+
import { mkdir, writeFile } from 'node:fs/promises';
|
|
2
|
+
import * as path from 'node:path';
|
|
3
|
+
import { tmpdir } from 'node:os';
|
|
4
|
+
|
|
5
|
+
import { describe, expect, it } from 'vitest';
|
|
6
|
+
|
|
7
|
+
import {
|
|
8
|
+
buildSessionQrelsFixtureFromJsonl,
|
|
9
|
+
collectClaudeSessionJsonlFiles,
|
|
10
|
+
summarizeSessionQrelsFixture
|
|
11
|
+
} from '../../src/core/session-qrels.js';
|
|
12
|
+
|
|
13
|
+
describe('session qrels fixture generation', () => {
|
|
14
|
+
it('turns Claude-style user/assistant session turns into replay qrels', () => {
|
|
15
|
+
const jsonl = [
|
|
16
|
+
JSON.stringify({
|
|
17
|
+
type: 'user',
|
|
18
|
+
sessionId: 's1',
|
|
19
|
+
message: { role: 'user', content: 'retrieval benchmark should report nDCG and precision together' },
|
|
20
|
+
timestamp: '2026-05-05T00:00:00.000Z'
|
|
21
|
+
}),
|
|
22
|
+
JSON.stringify({
|
|
23
|
+
type: 'assistant',
|
|
24
|
+
sessionId: 's1',
|
|
25
|
+
message: { role: 'assistant', content: [{ type: 'text', text: 'Use graded relevance qrels and report nDCG@k beside Precision@k and Recall@k.' }] },
|
|
26
|
+
timestamp: '2026-05-05T00:01:00.000Z'
|
|
27
|
+
}),
|
|
28
|
+
JSON.stringify({
|
|
29
|
+
type: 'user',
|
|
30
|
+
sessionId: 's1',
|
|
31
|
+
message: { role: 'user', content: '<command-name>/model</command-name>\n<local-command-stdout>opus</local-command-stdout>' }
|
|
32
|
+
}),
|
|
33
|
+
JSON.stringify({
|
|
34
|
+
type: 'assistant',
|
|
35
|
+
sessionId: 's1',
|
|
36
|
+
message: { role: 'assistant', content: 'This local command result must not become a benchmark qrel.' }
|
|
37
|
+
}),
|
|
38
|
+
JSON.stringify({
|
|
39
|
+
type: 'user',
|
|
40
|
+
sessionId: 's1',
|
|
41
|
+
message: { role: 'user', content: [{ type: 'text', text: 'fast search CLI should avoid embedding startup for qrels smoke tests' }] }
|
|
42
|
+
}),
|
|
43
|
+
JSON.stringify({
|
|
44
|
+
type: 'assistant',
|
|
45
|
+
sessionId: 's1',
|
|
46
|
+
message: { role: 'assistant', content: 'Fast search benchmark fixtures should stay lightweight and deterministic.' }
|
|
47
|
+
})
|
|
48
|
+
].join('\n');
|
|
49
|
+
|
|
50
|
+
const fixture = buildSessionQrelsFixtureFromJsonl(jsonl, {
|
|
51
|
+
name: 'session-qrels-test',
|
|
52
|
+
ks: [1, 3]
|
|
53
|
+
});
|
|
54
|
+
|
|
55
|
+
expect(fixture.name).toBe('session-qrels-test');
|
|
56
|
+
expect(fixture.ks).toEqual([1, 3]);
|
|
57
|
+
expect(fixture.queries).toHaveLength(2);
|
|
58
|
+
expect(fixture.memories).toHaveLength(2);
|
|
59
|
+
expect(fixture.queries[0]).toMatchObject({
|
|
60
|
+
queryId: 'q-s1-1',
|
|
61
|
+
query: 'retrieval benchmark should report nDCG and precision together',
|
|
62
|
+
expectedIds: ['m-s1-1'],
|
|
63
|
+
expectedRelevance: { 'm-s1-1': 2 }
|
|
64
|
+
});
|
|
65
|
+
expect(fixture.memories[0]).toMatchObject({
|
|
66
|
+
id: 'm-s1-1',
|
|
67
|
+
content: 'Use graded relevance qrels and report nDCG@k beside Precision@k and Recall@k.',
|
|
68
|
+
sourceSessionId: 's1'
|
|
69
|
+
});
|
|
70
|
+
expect(fixture.queries.map((query) => query.query)).not.toContain('opus');
|
|
71
|
+
});
|
|
72
|
+
|
|
73
|
+
it('pairs pending prompts independently per session id', () => {
|
|
74
|
+
const jsonl = [
|
|
75
|
+
JSON.stringify({
|
|
76
|
+
type: 'user',
|
|
77
|
+
sessionId: 's1',
|
|
78
|
+
message: { role: 'user', content: 'session one asks about retrieval replay metrics' }
|
|
79
|
+
}),
|
|
80
|
+
JSON.stringify({
|
|
81
|
+
type: 'user',
|
|
82
|
+
sessionId: 's2',
|
|
83
|
+
message: { role: 'user', content: 'session two asks about qrels generation safety' }
|
|
84
|
+
}),
|
|
85
|
+
JSON.stringify({
|
|
86
|
+
type: 'assistant',
|
|
87
|
+
sessionId: 's1',
|
|
88
|
+
message: { role: 'assistant', content: 'Session one answer explains nDCG replay metrics.' }
|
|
89
|
+
}),
|
|
90
|
+
JSON.stringify({
|
|
91
|
+
type: 'assistant',
|
|
92
|
+
sessionId: 's2',
|
|
93
|
+
message: { role: 'assistant', content: 'Session two answer explains qrels generation safety.' }
|
|
94
|
+
})
|
|
95
|
+
].join('\n');
|
|
96
|
+
|
|
97
|
+
const fixture = buildSessionQrelsFixtureFromJsonl(jsonl);
|
|
98
|
+
|
|
99
|
+
expect(fixture.queries.map((query) => [query.queryId, query.query])).toEqual([
|
|
100
|
+
['q-s1-1', 'session one asks about retrieval replay metrics'],
|
|
101
|
+
['q-s2-1', 'session two asks about qrels generation safety']
|
|
102
|
+
]);
|
|
103
|
+
expect(fixture.memories.map((memory) => [memory.id, memory.content])).toEqual([
|
|
104
|
+
['m-s1-1', 'Session one answer explains nDCG replay metrics.'],
|
|
105
|
+
['m-s2-1', 'Session two answer explains qrels generation safety.']
|
|
106
|
+
]);
|
|
107
|
+
});
|
|
108
|
+
|
|
109
|
+
it('adds known-answer labels and explicit no-match qrels', () => {
|
|
110
|
+
const jsonl = [
|
|
111
|
+
JSON.stringify({
|
|
112
|
+
type: 'user',
|
|
113
|
+
sessionId: 's-known',
|
|
114
|
+
message: { role: 'user', content: 'how should replay qrels record known answer fixtures' }
|
|
115
|
+
}),
|
|
116
|
+
JSON.stringify({
|
|
117
|
+
type: 'assistant',
|
|
118
|
+
sessionId: 's-known',
|
|
119
|
+
message: { role: 'assistant', content: 'Known-answer qrels should keep the expected assistant answer tied to the expected memory id.' }
|
|
120
|
+
})
|
|
121
|
+
].join('\n');
|
|
122
|
+
|
|
123
|
+
const fixture = buildSessionQrelsFixtureFromJsonl(jsonl, {
|
|
124
|
+
noMatchQueries: [
|
|
125
|
+
{
|
|
126
|
+
queryId: 'q-negative-empty-result',
|
|
127
|
+
query: 'unanswered moon cheese deployment question',
|
|
128
|
+
forbiddenIds: ['m-s-known-1'],
|
|
129
|
+
sourceSessionId: 's-known'
|
|
130
|
+
}
|
|
131
|
+
]
|
|
132
|
+
});
|
|
133
|
+
|
|
134
|
+
expect(fixture.queries).toHaveLength(2);
|
|
135
|
+
expect(fixture.queries[0]).toMatchObject({
|
|
136
|
+
queryId: 'q-s-known-1',
|
|
137
|
+
expectation: 'match',
|
|
138
|
+
expectedIds: ['m-s-known-1'],
|
|
139
|
+
knownAnswer: 'Known-answer qrels should keep the expected assistant answer tied to the expected memory id.'
|
|
140
|
+
});
|
|
141
|
+
expect(fixture.queries[1]).toMatchObject({
|
|
142
|
+
queryId: 'q-negative-empty-result',
|
|
143
|
+
query: 'unanswered moon cheese deployment question',
|
|
144
|
+
expectation: 'no_match',
|
|
145
|
+
expectedIds: [],
|
|
146
|
+
expectedRelevance: {},
|
|
147
|
+
forbiddenIds: ['m-s-known-1'],
|
|
148
|
+
sourceSessionId: 's-known'
|
|
149
|
+
});
|
|
150
|
+
|
|
151
|
+
const summary = summarizeSessionQrelsFixture(fixture);
|
|
152
|
+
expect(summary).toMatchObject({
|
|
153
|
+
queryCount: 2,
|
|
154
|
+
positiveQueryCount: 1,
|
|
155
|
+
noMatchQueryCount: 1,
|
|
156
|
+
knownAnswerCount: 1
|
|
157
|
+
});
|
|
158
|
+
});
|
|
159
|
+
|
|
160
|
+
it('can redact raw session text when generating shareable qrels metadata', () => {
|
|
161
|
+
const jsonl = [
|
|
162
|
+
JSON.stringify({
|
|
163
|
+
type: 'user',
|
|
164
|
+
sessionId: 'sensitive',
|
|
165
|
+
message: { role: 'user', content: 'SECRET customer project prompt should not leak' }
|
|
166
|
+
}),
|
|
167
|
+
JSON.stringify({
|
|
168
|
+
type: 'assistant',
|
|
169
|
+
sessionId: 'sensitive',
|
|
170
|
+
message: { role: 'assistant', content: 'SECRET implementation answer should not leak' }
|
|
171
|
+
})
|
|
172
|
+
].join('\n');
|
|
173
|
+
|
|
174
|
+
const fixture = buildSessionQrelsFixtureFromJsonl(jsonl, { redactContent: true });
|
|
175
|
+
const serialized = JSON.stringify(fixture);
|
|
176
|
+
|
|
177
|
+
expect(serialized).not.toContain('SECRET');
|
|
178
|
+
expect(fixture.queries[0]).toMatchObject({
|
|
179
|
+
queryId: 'q-sensitive-1',
|
|
180
|
+
query: '[redacted query q-sensitive-1]',
|
|
181
|
+
expectedIds: ['m-sensitive-1']
|
|
182
|
+
});
|
|
183
|
+
expect(fixture.memories[0]).toMatchObject({
|
|
184
|
+
id: 'm-sensitive-1',
|
|
185
|
+
content: '[redacted memory m-sensitive-1]'
|
|
186
|
+
});
|
|
187
|
+
});
|
|
188
|
+
|
|
189
|
+
it('summarizes generated qrels without leaking raw prompt or memory content', () => {
|
|
190
|
+
const jsonl = [
|
|
191
|
+
JSON.stringify({
|
|
192
|
+
type: 'user',
|
|
193
|
+
sessionId: 'sensitive',
|
|
194
|
+
message: { role: 'user', content: 'SECRET customer retrieval question with enough words' }
|
|
195
|
+
}),
|
|
196
|
+
JSON.stringify({
|
|
197
|
+
type: 'assistant',
|
|
198
|
+
sessionId: 'sensitive',
|
|
199
|
+
message: { role: 'assistant', content: 'SECRET assistant answer that should stay out of reports' }
|
|
200
|
+
})
|
|
201
|
+
].join('\n');
|
|
202
|
+
const fixture = buildSessionQrelsFixtureFromJsonl(jsonl, {
|
|
203
|
+
name: 'private-real-session-qrels',
|
|
204
|
+
ks: [1, 3, 10],
|
|
205
|
+
sourceFileCount: 2,
|
|
206
|
+
rawContentIncluded: true
|
|
207
|
+
});
|
|
208
|
+
|
|
209
|
+
const summary = summarizeSessionQrelsFixture(fixture);
|
|
210
|
+
const serialized = JSON.stringify(summary);
|
|
211
|
+
|
|
212
|
+
expect(summary).toMatchObject({
|
|
213
|
+
name: 'private-real-session-qrels',
|
|
214
|
+
queryCount: 1,
|
|
215
|
+
memoryCount: 1,
|
|
216
|
+
sourceSessionCount: 1,
|
|
217
|
+
sourceFileCount: 2,
|
|
218
|
+
rawContentIncluded: true,
|
|
219
|
+
ks: [1, 3, 10]
|
|
220
|
+
});
|
|
221
|
+
expect(summary.perSession).toEqual([
|
|
222
|
+
{
|
|
223
|
+
sourceSessionId: 'sensitive',
|
|
224
|
+
queryCount: 1,
|
|
225
|
+
memoryCount: 1,
|
|
226
|
+
firstTurnIndex: 1,
|
|
227
|
+
lastTurnIndex: 1
|
|
228
|
+
}
|
|
229
|
+
]);
|
|
230
|
+
expect(serialized).not.toContain('SECRET');
|
|
231
|
+
});
|
|
232
|
+
|
|
233
|
+
it('collects Claude JSONL sessions recursively while excluding subagents by default', async () => {
|
|
234
|
+
const root = path.join(tmpdir(), `session-qrels-${process.pid}-${Date.now()}`);
|
|
235
|
+
const projectDir = path.join(root, 'project-a');
|
|
236
|
+
const subagentDir = path.join(projectDir, 'subagents');
|
|
237
|
+
await mkdir(subagentDir, { recursive: true });
|
|
238
|
+
await writeFile(path.join(projectDir, 'top.jsonl'), '{}\n', 'utf8');
|
|
239
|
+
await writeFile(path.join(subagentDir, 'agent.jsonl'), '{}\n', 'utf8');
|
|
240
|
+
await writeFile(path.join(projectDir, 'notes.txt'), 'ignore', 'utf8');
|
|
241
|
+
|
|
242
|
+
await expect(collectClaudeSessionJsonlFiles(root)).resolves.toEqual([
|
|
243
|
+
path.join(projectDir, 'top.jsonl')
|
|
244
|
+
]);
|
|
245
|
+
await expect(collectClaudeSessionJsonlFiles(root, { includeSubagents: true })).resolves.toEqual([
|
|
246
|
+
path.join(subagentDir, 'agent.jsonl'),
|
|
247
|
+
path.join(projectDir, 'top.jsonl')
|
|
248
|
+
]);
|
|
249
|
+
});
|
|
250
|
+
});
|
package/tests/{sqlite-event-store-replication.test.ts → core/sqlite-event-store-replication.test.ts}
RENAMED
|
@@ -7,7 +7,7 @@ import * as fs from 'fs';
|
|
|
7
7
|
import * as path from 'path';
|
|
8
8
|
import * as os from 'os';
|
|
9
9
|
|
|
10
|
-
import { SQLiteEventStore } from '
|
|
10
|
+
import { SQLiteEventStore } from '../../src/core/sqlite-event-store.js';
|
|
11
11
|
|
|
12
12
|
describe('SQLiteEventStore replication helpers', () => {
|
|
13
13
|
let tempDir: string;
|
|
@@ -88,5 +88,40 @@ describe('SQLiteEventStore replication helpers', () => {
|
|
|
88
88
|
expect(dup.isDuplicate).toBe(true);
|
|
89
89
|
expect(dup.eventId).toBe(sourceEvent!.id);
|
|
90
90
|
});
|
|
91
|
+
|
|
92
|
+
it('keeps FTS event_id populated after deleteSessionEvents recreates triggers', async () => {
|
|
93
|
+
await storeA.append({ eventType: 'user_prompt', sessionId: 'delete-me', timestamp: new Date(), content: 'temporary kiwi memory' });
|
|
94
|
+
await storeA.append({ eventType: 'user_prompt', sessionId: 'keep-me', timestamp: new Date(), content: 'persistent pineapple memory' });
|
|
95
|
+
|
|
96
|
+
await expect(storeA.deleteSessionEvents('delete-me')).resolves.toBe(1);
|
|
97
|
+
|
|
98
|
+
const appendAfterTriggerRecreation = await storeA.append({
|
|
99
|
+
eventType: 'user_prompt',
|
|
100
|
+
sessionId: 'keep-me',
|
|
101
|
+
timestamp: new Date(),
|
|
102
|
+
content: 'fresh dragonfruit memory'
|
|
103
|
+
});
|
|
104
|
+
expect(appendAfterTriggerRecreation.success).toBe(true);
|
|
105
|
+
|
|
106
|
+
await expect(storeA.rebuildFtsIndex()).resolves.toBe(2);
|
|
107
|
+
const results = await storeA.keywordSearch('dragonfruit', 5);
|
|
108
|
+
|
|
109
|
+
expect(results.map((result) => result.event.content)).toEqual(['fresh dragonfruit memory']);
|
|
110
|
+
});
|
|
111
|
+
|
|
112
|
+
it('supports event updates and deletes on a fresh FTS table before manual rebuild', async () => {
|
|
113
|
+
const appendResult = await storeA.append({
|
|
114
|
+
eventType: 'user_prompt',
|
|
115
|
+
sessionId: 'mutable-session',
|
|
116
|
+
timestamp: new Date(),
|
|
117
|
+
content: 'mutable mango memory'
|
|
118
|
+
});
|
|
119
|
+
expect(appendResult.success).toBe(true);
|
|
120
|
+
if (!appendResult.success) return;
|
|
121
|
+
|
|
122
|
+
await expect(storeA.incrementAccessCount([appendResult.eventId])).resolves.toBeUndefined();
|
|
123
|
+
await expect(storeA.deleteSessionEvents('mutable-session')).resolves.toBe(1);
|
|
124
|
+
await expect(storeA.keywordSearch('mango', 5)).resolves.toEqual([]);
|
|
125
|
+
});
|
|
91
126
|
});
|
|
92
127
|
|