@furongjun1999/dsh-memory 0.4.7 → 0.4.9
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/README.md +130 -47
- package/codebuddy/CODEBUDDY.md +21 -10
- package/codebuddy/README.md +13 -1
- package/codebuddy/mcp.json +9 -0
- package/docs/GBrain/345/217/257/345/200/237/351/211/264/347/202/271_/347/201/265/346/236/242/350/220/275/347/202/271/344/272/244/346/216/245_20260919.md +169 -0
- package/docs/Pi/345/217/257/345/255/246/344/271/240/344/274/230/347/202/271_/347/201/265/346/236/242/345/244/247/350/204/221/346/224/271/350/277/233/344/272/244/346/216/245_20260915.md +144 -0
- package/docs/README.md +111 -0
- package/docs/discipline/harnesses.yaml +226 -152
- package/docs/discipline/templates/full.md.tmpl +61 -58
- package/docs/eval//347/254/254/344/270/211/346/226/271/351/252/214/350/257/201/346/212/245/345/221/212_LoCoMo_/347/201/265/346/236/242_.md +127 -0
- package/docs/eval//347/254/254/344/270/211/346/226/271/351/252/214/350/257/201/346/212/245/345/221/212_LoCoMo_/347/201/265/346/236/242_.png +0 -0
- package/docs/eval//347/254/254/344/270/211/346/226/271/351/252/214/350/257/201/346/212/245/345/221/212_/347/201/265/346/236/242_vs_dejavu_/347/273/237/344/270/200/350/257/204/345/210/206_v7.md +202 -0
- package/docs/experiments/linkref_backfill/candidates_20260917.json +726 -0
- package/docs/experiments/linkref_backfill/candidates_internal_20260917.json +602 -0
- package/docs/experiments/linkref_backfill/candidates_internal_v2.json +603 -0
- package/docs/experiments/linkref_backfill/candidates_secret_20260917.json +884 -0
- package/docs/experiments/linkref_backfill/candidates_secret_v2.json +789 -0
- package/docs/experiments/m4-role-probe/census.py +86 -0
- package/docs/experiments/m4-role-probe/probe.py +89 -0
- package/docs/experiments/mapped_confidence/bench_mapped_conf.py +216 -0
- package/docs/experiments/mapped_confidence/post_check.py +75 -0
- package/docs/experiments/mapped_confidence/repro_thirdparty_tol.py +83 -0
- package/docs/experiments/mapped_confidence/result_mapped_conf.json +215 -0
- package/docs/hive//345/244/232/347/253/257harness/351/200/232/344/277/241/345/245/221/347/272/246_v0.1.md +42 -0
- package/docs/hive//346/243/200/347/264/242/346/224/266/346/225/233/345/256/236/346/265/213/344/270/216S1b/350/256/276/350/256/241_v0.1.md +43 -0
- package/docs/hive//346/243/200/347/264/242/350/267/257/345/276/204/344/270/216/350/256/244/347/237/245/347/273/223/346/236/204/345/245/221/347/272/246_v0.1.md +102 -0
- package/docs/hive//347/234/237/345/256/236/345/272/223/347/253/257/345/210/260/347/253/257/345/256/236/346/265/213_S1b/344/270/216/345/217/254/345/233/236/346/235/203/350/241/241_v0.1.md +57 -0
- package/docs/hive//347/234/237/345/256/236/345/272/223/347/253/257/345/210/260/347/253/257/345/256/236/346/265/213_S7/345/200/222/346/216/222/345/200/231/351/200/211/345/261/202_v0.1.md +126 -0
- package/docs/hive//350/234/202/345/267/242/345/217/214/345/256/236/344/276/213/344/272/222/351/252/214_/350/256/276/350/256/241/345/256/232/347/250/277.md +503 -0
- package/docs/{ → hive/}/350/234/202/345/267/242/345/267/245/344/275/234/350/256/260/345/277/206_/351/241/271/347/233/256/350/256/241/345/210/222.md +14 -2
- package/docs/mdcg/D_meta_/345/267/245/347/250/213/345/214/226/346/226/271/346/241/210_v0.2.md +216 -0
- package/docs/{README → mdcg/README}/350/257/246/347/273/206/347/211/210_v0.4.5.md +631 -625
- package/docs/{ → mdcg/}/344/270/273/344/273/243/347/220/206/345/255/220/344/273/243/347/220/206/350/256/260/345/277/206/346/236/266/346/236/204/350/256/276/350/256/241.md +1 -1
- package/docs/mdcg//344/273/243/347/240/201/350/257/204/345/256/241/344/270/216/346/235/241/344/273/266/345/214/226/346/263/250/351/207/212_/345/245/221/347/272/246_v0.1.md +82 -0
- package/docs/mdcg//345/205/250/345/272/223/344/273/243/347/240/201/350/257/204/345/256/241/344/270/216/346/235/241/344/273/266/345/214/226/346/263/250/351/207/212_/350/256/241/345/210/222_v0.1.md +600 -0
- package/docs/mdcg//345/212/237/350/203/275/350/260/203/347/224/250/346/230/240/345/260/204/350/241/250_v0.1.md +263 -0
- package/docs/{ → mdcg/}/345/215/225/345/205/203/350/207/252/346/210/221/351/224/232/347/202/271_/347/263/273/347/273/237/346/217/220/347/244/272/350/257/215/346/240/207/345/207/206_v0.3.md +1 -1
- package/docs/mdcg//345/255/220/344/273/243/347/220/206/351/205/215/347/275/256/346/240/207/345/207/206_v0.5.md +236 -0
- package/docs/{ → mdcg/}/346/272/220/347/240/201/347/272/247/346/236/266/346/236/204/345/256/241/350/256/241_GPT/346/211/271/350/257/204/345/257/271/347/205/247_v1.0.md +2 -2
- package/docs/{ → mdcg/}/347/201/265/346/236/242/344/270/211/345/261/202/346/213/206/345/210/206/350/247/204/345/210/222_v0.1.md +181 -181
- package/docs/mdcg//347/201/265/346/236/242/350/256/260/345/277/206/345/212/250/350/257/215/345/215/217/350/256/256_v1.0-draft.md +172 -0
- package/docs/mdcg//347/216/257/344/272/214_/347/231/275/347/256/261/345/241/253/345/205/205/346/265/201/346/260/264/347/272/277_v0.1.md +31 -0
- package/docs/{ → mdcg/}/347/274/272/345/217/243/345/215/225_P0/346/224/266/345/217/243_v0.1.md +2 -2
- package/docs/{ → mdcg/}/350/256/244/347/237/245/345/233/276_G4-G8/347/274/272/345/217/243/350/243/201/345/256/232/345/215/225_v0.1.md +1 -1
- package/docs/{ → mdcg/}/350/256/244/347/237/245/345/233/276_/347/264/242/345/274/225/344/270/216/345/267/245/347/250/213/350/247/204/350/214/203/345/214/226_/350/256/241/345/210/222_v0.1.md +4 -4
- package/docs/{ → mdcg/}/350/256/260/345/277/206/346/223/215/344/275/234/347/263/273/347/273/237_MdCGOS/344/270/216MCP/346/216/245/345/205/245_v0.1.md +2 -2
- package/docs/mdcg//350/267/250/347/253/257/351/252/214/350/257/201/344/270/216/345/220/214/346/255/245/345/215/217/350/256/256_v0.1.md +93 -0
- package/docs/plans/GridWorld/346/234/200/345/260/217/351/227/255/347/216/257/350/247/204/346/240/274_v0.1.md +176 -0
- package/docs/{ → swarm/}/350/234/202/347/276/244/344/272/222/350/201/224_v0.1.md +3 -3
- package/docs/swarm//350/234/202/347/276/244/345/220/214/351/224/231/346/243/200/346/265/213/345/256/236/351/252/214/345/215/217/350/256/256_v0.1.md +242 -0
- package/docs/theory//345/215/225/347/272/277/347/250/213/344/270/216/346/263/250/346/204/217/345/212/233/351/233/206/344/270/255_/346/227/240/344/272/211/350/256/256/347/220/206/350/256/272/346/226/207/346/241/243_v1.0.md +129 -0
- package/docs/theory//345/271/266/345/217/221/345/277/205/347/204/266/346/200/247/347/220/206/350/256/272_v0.2.md +134 -0
- package/docs/theory//346/246/202/345/277/265/345/210/206/345/261/202/345/257/271/351/275/220/350/241/250_v0.1.md +145 -0
- package/docs/{ → theory/}/347/220/206/350/256/272_/346/234/272/345/210/266_/344/273/243/347/240/201_/345/256/236/351/252/214_/347/274/272/345/217/243/347/237/251/351/230/265_v0.1.md +6 -6
- package/docs/{ → theory/}/347/220/206/350/256/272/344/273/223/346/213/206/345/210/206/344/270/216/347/231/275/347/256/261/347/237/245/350/257/206/345/272/223/345/206/205/350/277/201_v0.1.md +1 -1
- package/docs/theory//350/256/244/347/237/245/344/273/243/347/220/206/344/270/216/346/224/266/346/225/233/347/273/223/346/236/204_/346/235/241/344/273/266/350/256/272/351/207/215/346/236/204_v0.1.md +221 -0
- package/docs/world_model//344/270/226/347/225/214/346/250/241/345/236/213_/347/245/236/347/273/217/347/275/221/347/273/234/345/272/225/345/261/202/346/236/266/346/236/204_v1.0.md +237 -0
- package/docs//345/255/230/347/256/227/344/270/200/344/275/223/346/236/266/346/236/204/345/255/246/344/271/240_/345/244/226/351/203/250/347/220/206/350/256/272/345/257/271/347/205/247/344/270/216/350/267/257/347/272/277/344/272/244/346/216/245_20260916.md +98 -0
- package/docs//345/255/230/347/256/227/344/270/200/344/275/223/347/245/236/347/273/217/347/275/221/347/273/234/346/236/266/346/236/204_/346/200/273/347/272/262/344/270/216/344/272/244/346/216/245_20260916.md +109 -0
- package/docs//345/267/245/344/275/234/347/272/252/345/276/213_/350/256/244/347/237/245/345/233/276/346/235/241/347/233/256_v1.1.json +434 -405
- package/docs//347/201/265/346/236/242/350/207/252/346/210/221/346/224/271/350/277/233/345/267/245/344/275/234/350/256/241/345/210/222_/345/244/226/351/203/250/347/240/224/347/251/266/347/263/273/345/210/227/345/220/270/346/224/266_v1_20260919.md +286 -0
- package/docs//347/231/275/347/256/261/345/214/226/347/272/262/351/242/206_/344/270/211/351/241/271/347/233/256/345/255/246/344/271/240/346/200/273/346/236/266/346/236/204/344/270/216Ghidra/344/272/244/346/216/245_20260916.md +104 -0
- package/docs//350/256/260/345/277/206/350/257/204/345/256/241/347/263/273/347/273/237_/347/253/213/351/241/271/350/256/276/350/256/241/344/270/216/346/226/275/345/267/245/344/272/244/346/216/245_20260915.md +183 -0
- package/dsh/README.md +34 -1
- package/dsh/cordis.yml.example +13 -7
- package/dsh/hive-mcp-probe.mjs +94 -0
- package/dsh/hive-mcp.example.yml +62 -0
- package/dsh/update-lingshu.bat +11 -0
- package/dsh/update-lingshu.ps1 +337 -0
- package/lib/hooks.d.ts +8 -0
- package/lib/hooks.js +27 -24
- package/lib/index.d.ts +4 -2
- package/lib/index.js +29 -6
- package/lib/lib/datapath.d.ts +76 -1
- package/lib/lib/datapath.js +199 -13
- package/lib/lib/mdcg_client.d.ts +6 -3
- package/lib/lib/mdcg_client.js +6 -5
- package/lib/lib/mutual.js +4 -4
- package/lib/lib/prompt_safety.d.ts +48 -0
- package/lib/lib/prompt_safety.js +62 -0
- package/lib/lib/token_store.js +4 -5
- package/md_cg/audit.py +91 -2
- package/md_cg/autonomy.py +86 -15
- package/md_cg/backfill.py +36 -1
- package/md_cg/backfill_bigdomain.py +34 -0
- package/md_cg/bench6_arms.py +28 -1
- package/md_cg/bench6_common.py +10 -1
- package/md_cg/bench6_competitors.py +6 -1
- package/md_cg/bench_axis_domain.py +9 -1
- package/md_cg/bench_blind_comp.py +7 -1
- package/md_cg/bench_en_atoms_public.py +9 -0
- package/md_cg/bench_governance.py +348 -0
- package/md_cg/bench_lme_zh.py +16 -1
- package/md_cg/bench_locomo.py +2 -1
- package/md_cg/bench_locomo_zh.py +16 -1
- package/md_cg/bench_locomo_zh_public.py +4 -1
- package/md_cg/bench_longmem.py +2 -1
- package/md_cg/bench_membench.py +37 -6
- package/md_cg/bench_p0.py +4 -1
- package/md_cg/bench_progressive.py +61 -10
- package/md_cg/bench_role_views.py +238 -0
- package/md_cg/bench_task_ab.py +8 -1
- package/md_cg/bench_task_ab_llm.py +13 -1
- package/md_cg/bench_unified_en.py +6 -1
- package/md_cg/bench_zh_mad.py +20 -1
- package/md_cg/blindspot_tickets.py +123 -0
- package/md_cg/branches.py +286 -0
- package/md_cg/build_postings.py +73 -0
- package/md_cg/ccgc.py +949 -0
- package/md_cg/census.py +5 -1
- package/md_cg/chain.py +24 -3
- package/md_cg/codeindex.py +134 -17
- package/md_cg/coldverify.py +265 -0
- package/md_cg/comment_gate.py +338 -0
- package/md_cg/cond_compose.py +190 -0
- package/md_cg/cond_facts.py +155 -0
- package/md_cg/cond_template.json +107 -0
- package/md_cg/condition_anchor.py +143 -0
- package/md_cg/conformance.py +727 -0
- package/md_cg/consistency.py +25 -2
- package/md_cg/consolidate.py +53 -2
- package/md_cg/corpus.py +5 -1
- package/md_cg/crosscheck.py +42 -2
- package/md_cg/crypto.py +35 -1
- package/md_cg/d_meta.py +310 -0
- package/md_cg/datapath.py +201 -26
- package/md_cg/docindex.py +122 -1
- package/md_cg/eval_common.py +35 -6
- package/md_cg/evidence.py +27 -1
- package/md_cg/evolution.py +21 -1
- package/md_cg/export.py +11 -1
- package/md_cg/forgetting.py +34 -4
- package/md_cg/fsutil.py +81 -2
- package/md_cg/hotcache.py +214 -0
- package/md_cg/hyperedge.py +251 -0
- package/md_cg/identity.py +20 -3
- package/md_cg/insight.py +19 -3
- package/md_cg/lexicon/build_cedict_en_zh.py +9 -0
- package/md_cg/lexicon/build_standard_en.py +171 -168
- package/md_cg/lexicon/expand_en_zh.py +6 -0
- package/md_cg/lifecycle.py +273 -0
- package/md_cg/linkref.py +281 -0
- package/md_cg/links.py +29 -1
- package/md_cg/mcp_server.py +894 -168
- package/md_cg/md_whitebox.py +53 -1
- package/md_cg/mdcg.py +1205 -41
- package/md_cg/mdcos.py +1060 -76
- package/md_cg/metacognition.py +39 -4
- package/md_cg/migrate.py +4 -0
- package/md_cg/migrate_aeis.py +221 -213
- package/md_cg/migrate_roleplay.py +8 -0
- package/md_cg/migrate_wisdom_graph.py +14 -1
- package/md_cg/mreview/__init__.py +25 -0
- package/md_cg/mreview/__main__.py +110 -0
- package/md_cg/mreview/bundle.py +178 -0
- package/md_cg/mreview/candidates.py +262 -0
- package/md_cg/mreview/govern.py +694 -0
- package/md_cg/mreview/locate.py +939 -0
- package/md_cg/mreview/pipeline.py +729 -0
- package/md_cg/mreview/rules/duplication.json +21 -0
- package/md_cg/mreview/rules/field_coverage.json +54 -0
- package/md_cg/mreview/rules/source_license.json +21 -0
- package/md_cg/mreview/rules/template_flow.json +21 -0
- package/md_cg/mreview/ruleset.py +253 -0
- package/md_cg/nodefile.py +269 -1
- package/md_cg/pooling.py +23 -1
- package/md_cg/postings.py +298 -0
- package/md_cg/predict.py +113 -15
- package/md_cg/progressive.py +3 -0
- package/md_cg/protect.py +18 -5
- package/md_cg/protocol.py +372 -0
- package/md_cg/provenance.py +262 -0
- package/md_cg/reach.py +453 -0
- package/md_cg/refindex.py +88 -4
- package/md_cg/refine.py +20 -1
- package/md_cg/roleviews.py +89 -0
- package/md_cg/routing.py +76 -0
- package/md_cg/scrub.py +68 -3
- package/md_cg/security.py +26 -1
- package/md_cg/self_state.py +65 -2
- package/md_cg/selfreport.py +151 -0
- package/md_cg/semantic/canonical.py +5 -0
- package/md_cg/semantic/en_normalizer.py +364 -295
- package/md_cg/semantic/zh_en_atoms.py +139 -136
- package/md_cg/signer.py +41 -1
- package/md_cg/sources.py +583 -547
- package/md_cg/statushdr.py +179 -0
- package/md_cg/stg.py +59 -18
- package/md_cg/subgraph.py +23 -0
- package/md_cg/sustain.py +74 -1
- package/md_cg/tasks.py +471 -0
- package/md_cg/test_action_derive.py +203 -0
- package/md_cg/test_audit_rotate.py +270 -0
- package/md_cg/test_autonomy.py +26 -0
- package/md_cg/test_bench_governance.py +102 -0
- package/md_cg/test_blindspot_tickets.py +166 -0
- package/md_cg/test_branches.py +249 -0
- package/md_cg/test_ccg_perturb.py +1 -1
- package/md_cg/test_ccgc.py +433 -0
- package/md_cg/test_codeindex.py +338 -0
- package/md_cg/test_comment_gate.py +187 -0
- package/md_cg/test_cond_compose_anchors.py +76 -0
- package/md_cg/test_condition_anchor.py +82 -0
- package/md_cg/test_conformance.py +343 -0
- package/md_cg/test_d_meta.py +412 -0
- package/md_cg/test_datapath_root.py +188 -0
- package/md_cg/test_en_pipeline.py +24 -0
- package/md_cg/test_gain_gate.py +47 -1
- package/md_cg/test_health_scale.py +173 -0
- package/md_cg/test_hot_cold.py +187 -0
- package/md_cg/test_hyperedge.py +245 -0
- package/md_cg/test_identity_attribution.py +6 -0
- package/md_cg/test_index_durability.py +224 -0
- package/md_cg/test_lifecycle.py +309 -0
- package/md_cg/test_linkref.py +306 -0
- package/md_cg/test_md_access_parity.py +15 -3
- package/md_cg/test_mr_m1.py +769 -0
- package/md_cg/test_mr_m2.py +587 -0
- package/md_cg/test_mr_m3.py +710 -0
- package/md_cg/test_mr_m4.py +485 -0
- package/md_cg/test_p1.py +1 -1
- package/md_cg/test_p11_consistency.py +1 -1
- package/md_cg/test_p12_metacognition.py +2 -2
- package/md_cg/test_p16_self_state.py +1 -1
- package/md_cg/test_p26_refindex.py +50 -21
- package/md_cg/test_p27_docindex.py +258 -4
- package/md_cg/test_p28_refcheck.py +1 -1
- package/md_cg/test_p29_session_ingest_export.py +1 -1
- package/md_cg/test_p2_mcp.py +7 -1
- package/md_cg/test_p31_insight.py +24 -0
- package/md_cg/test_p38_contextualize.py +1 -1
- package/md_cg/test_p39_vision_evidence.py +1 -1
- package/md_cg/test_p40_refine_worklist.py +1 -1
- package/md_cg/test_p41_evolve_patrol.py +1 -1
- package/md_cg/test_p42_provenance.py +1 -1
- package/md_cg/test_p43_pooling.py +41 -13
- package/md_cg/test_p44_md_whitebox.py +14 -1
- package/md_cg/test_protocol.py +243 -0
- package/md_cg/test_reach.py +378 -0
- package/md_cg/test_reach_keys.py +201 -0
- package/md_cg/test_reach_meta_exits.py +145 -0
- package/md_cg/test_read_clip.py +141 -0
- package/md_cg/test_retr_s1.py +340 -0
- package/md_cg/test_retr_s1b.py +209 -0
- package/md_cg/test_retr_s3.py +194 -0
- package/md_cg/test_retr_s4.py +163 -0
- package/md_cg/test_retr_s5.py +200 -0
- package/md_cg/test_retr_s6.py +157 -0
- package/md_cg/test_retr_s7.py +385 -0
- package/md_cg/test_retr_s8_time.py +316 -0
- package/md_cg/test_retr_s9_edges.py +286 -0
- package/md_cg/test_retr_s9_entity_ctx.py +175 -0
- package/md_cg/test_review_conformance.py +367 -0
- package/md_cg/test_role_views.py +354 -0
- package/md_cg/test_tasks.py +409 -0
- package/md_cg/test_tool_face.py +189 -0
- package/md_cg/test_trust.py +361 -0
- package/md_cg/test_twophase.py +286 -0
- package/md_cg/test_units_poll.py +71 -0
- package/md_cg/test_v14_fixes.py +397 -0
- package/md_cg/test_validity_filter.py +280 -0
- package/md_cg/test_wisdom_md_store.py +7 -3
- package/md_cg/test_writepipe.py +214 -0
- package/md_cg/theory.py +17 -2
- package/md_cg/tokens.py +134 -12
- package/md_cg/tool_face.py +261 -0
- package/md_cg/trust.py +943 -0
- package/md_cg/twophase.py +232 -0
- package/md_cg/units.py +665 -0
- package/md_cg/vision_evidence.py +25 -2
- package/md_cg/weights.py +25 -2
- package/md_cg/whitebox.py +33 -2
- package/md_cg/whitebox_kb/data/verify_cache.json +21233 -360
- package/md_cg/whitebox_kb/data/verify_savings.jsonl +5124 -0
- package/md_cg/whitebox_kb/wisdom/audit_log/chain_heat.json +10 -10
- package/md_cg/whitebox_kb/wisdom/code_compose.py +113 -6
- package/md_cg/whitebox_kb/wisdom/code_solidified.json +1 -1
- package/md_cg/whitebox_kb/wisdom/verifier.py +340 -55
- package/md_cg/whitebox_kb/wisdom/wisdom-book-cloud.db +0 -0
- package/md_cg/writelimit.py +37 -7
- package/md_cg/writepipe.py +543 -0
- package/package.json +2 -2
- package/skills/skills/designer-perspective/scripts/__pycache__/designer.cpython-310.pyc +0 -0
- package/skills/skills/designer-perspective/scripts/designer.py +17 -1
- package/skills/skills/designer-perspective/tests/selftest.py +3 -1
- package/src/hooks.ts +27 -22
- package/src/index.ts +33 -6
- package/src/lib/datapath.ts +211 -13
- package/src/lib/mdcg_client.ts +12 -8
- package/src/lib/mutual.ts +411 -411
- package/src/lib/prompt_safety.ts +62 -0
- package/src/lib/token_store.ts +4 -5
- package/zcode/AGENTS.md +21 -10
- package/zcode/README.md +4 -0
- package/docs//345/212/237/350/203/275/350/260/203/347/224/250/346/230/240/345/260/204/350/241/250_v0.1.md +0 -221
- /package/docs/{AGI → eval/AGI}/344/270/203/347/273/264/350/257/204/345/210/206/346/212/245/345/221/212_md_cg_v1.0.md" +0 -0
- /package/docs/{AGI → eval/AGI}/344/270/203/347/273/264/350/257/204/345/210/206/346/212/245/345/221/212_md_cg_v2.0.md" +0 -0
- /package/docs/{ → eval/}/345/256/236/351/252/214/346/226/271/346/241/210_/345/255/246/344/271/240/351/227/255/347/216/257AB/344/270/216/346/250/252/350/257/204_v1.0.md" +0 -0
- /package/docs/{ → eval/}/346/250/252/350/257/204_/345/205/255/345/256/266100/351/242/230/344/270/255/350/213/261/345/217/214/346/237/245_v1.0.md" +0 -0
- /package/docs/{guardrail-charter.md → mdcg/guardrail-charter.md} +0 -0
- /package/docs/{lingshu_tutorial.html → mdcg/lingshu_tutorial.html} +0 -0
- /package/docs/{memory-assessment.html → mdcg/memory-assessment.html} +0 -0
- /package/docs/{memory_score.html → mdcg/memory_score.html} +0 -0
- /package/docs/{memory_score.png → mdcg/memory_score.png} +0 -0
- /package/docs/{release_v0.3.0.md → mdcg/release_v0.3.0.md} +0 -0
- /package/docs/{release_v0.4.5.md → mdcg/release_v0.4.5.md} +0 -0
- /package/docs/{tool_table_v0.3.0.md → mdcg/tool_table_v0.3.0.md} +0 -0
- /package/docs/{ → mdcg/}/344/273/244/347/211/214/344/270/216/350/247/222/350/211/262/346/235/203/350/201/214/345/210/206/347/246/273_v0.1.md" +0 -0
- /package/docs/{ → mdcg/}/347/201/265/346/236/24282/345/267/245/345/205/267_/345/212/237/350/203/275/346/225/264/347/220/206/344/270/216/350/277/201/347/247/273/346/230/240/345/260/204_v0.1.md" +0 -0
- /package/docs/{ → mdcg/}/347/201/265/346/236/242MCP/345/267/245/345/205/267/346/200/273/350/241/250_v3.4.md" +0 -0
- /package/docs/{ → mdcg/}/347/201/265/346/236/242_/350/207/252/346/210/221/345/261/202/345/256/232/344/271/211.md" +0 -0
- /package/docs/{ → mdcg/}/350/256/244/347/237/245/345/233/276_MD/347/233/256/345/275/225/346/226/271/346/241/210_v0.1.md" +0 -0
- /package/docs/{ → mdcg/}/350/256/244/347/237/245/345/233/276_/346/235/241/344/273/266/347/251/272/351/227/264/345/220/210/346/210/220/344/270/216/347/224/237/346/225/210/346/235/241/344/273/266/345/217/243/345/276/204_v0.1.md" +0 -0
- /package/docs/{ → mdcg/}/350/256/244/347/237/245/345/233/276/344/275/234/344/270/272/350/256/260/345/277/206/346/223/215/344/275/234/347/263/273/347/273/237_/350/257/204/344/274/260/344/270/216/350/267/257/347/272/277_v0.1.md" +0 -0
- /package/docs/{ → plans/}/345/221/275/345/220/215/346/262/273/347/220/206_/351/241/271/347/233/256/350/256/241/345/210/222.md" +0 -0
- /package/docs/{ → plans/}/347/237/245/350/257/206/345/233/276/350/260/261/351/241/271/347/233/256_/344/272/244/346/216/245/346/226/207/346/241/243_20260914.md" +0 -0
- /package/docs/{ → swarm/}/350/234/202/347/276/244/345/244/232/346/231/272/350/203/275/344/275/223_/344/275/277/347/224/250/346/212/245/345/221/212_20260913.md" +0 -0
- /package/docs/{ → swarm/}/350/234/202/347/276/244/345/244/232/346/231/272/350/203/275/344/275/223_/345/212/237/350/203/275/350/257/264/346/230/216_v0.6.md" +0 -0
- /package/docs/{ → swarm/}/350/234/202/347/276/244/347/233/262/345/214/272/346/240/207/350/256/260_v1.0.md" +0 -0
- /package/docs/{ → theory/}/344/277/241/346/201/257/345/267/256/344/270/272/344/273/200/344/271/210/345/277/205/347/204/266/345/255/230/345/234/250/344/270/224/350/207/252/347/204/266/346/211/251/345/244/247.md" +0 -0
- /package/docs/{ → theory/}/346/231/272/350/203/275/347/232/204/345/205/254/347/220/206/345/214/226/345/237/272/347/237/263.md" +0 -0
- /package/docs/{ → theory/}/346/231/272/350/203/275/347/232/204/350/256/244/347/237/245/350/277/207/347/250/213.md" +0 -0
- /package/docs/{ → theory/}/346/231/272/350/203/275/350/256/2723.4.md" +0 -0
- /package/docs/{ → theory/}/347/231/275/347/256/261/346/231/272/350/203/275/346/230/257/344/273/200/344/271/210/357/274/237.md" +0 -0
- /package/docs/{ → theory/}/347/231/275/347/256/261/346/231/272/350/203/275/347/263/273/345/210/227/302/267/347/254/254/344/272/224/347/257/207/357/274/232/350/256/251AI/347/234/237/346/255/243/350/243/205/344/270/212/350/256/260/345/277/206.md" +0 -0
|
@@ -0,0 +1,340 @@
|
|
|
1
|
+
# -*- coding: utf-8 -*-
|
|
2
|
+
"""S1/S2 检索前门控探针(脚本式,exit code 定成败)。
|
|
3
|
+
|
|
4
|
+
用法:python -X utf8 -m md_cg.test_retr_s1
|
|
5
|
+
覆盖(含独立复核指出的缺口):
|
|
6
|
+
- routing.domain_terms / classify_text(取词、去重、limit、空值、无信号)
|
|
7
|
+
- 写入侧:默认关**不写** big_domain(口径不变);开总开关才写;索引快照同口径
|
|
8
|
+
- 默认关:meta 不含 gates 键(逐字节等价)、scanned=全量
|
|
9
|
+
- S1:按域收敛;域内不足→回退全量;query 无域信号→不收敛
|
|
10
|
+
- S2:observation_position 不匹配→丢;信息不足→放行;time_window 不相交→丢(平铺字段);全滤→回退
|
|
11
|
+
- backfill_big_domain:dry-run 不改盘、真跑落盘、二次跑幂等
|
|
12
|
+
"""
|
|
13
|
+
import io
|
|
14
|
+
import json
|
|
15
|
+
import os
|
|
16
|
+
import sys
|
|
17
|
+
import tempfile
|
|
18
|
+
|
|
19
|
+
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
|
20
|
+
|
|
21
|
+
from md_cg import nodefile, routing # noqa: E402
|
|
22
|
+
from md_cg.fsutil import ShardedLog # noqa: E402
|
|
23
|
+
from md_cg.mdcg import MdCG # noqa: E402
|
|
24
|
+
|
|
25
|
+
passed = 0
|
|
26
|
+
failed = 0
|
|
27
|
+
GONG = "工程 结构 应力 梁 截面 桥梁 材料 机械"
|
|
28
|
+
YI = "医学 疾病 诊断 药物 处方 症状 治疗"
|
|
29
|
+
FAR_TW = [10 ** 13, 10 ** 13 + 1]
|
|
30
|
+
|
|
31
|
+
|
|
32
|
+
def check(name, cond, detail=""):
|
|
33
|
+
global passed, failed
|
|
34
|
+
if cond:
|
|
35
|
+
passed += 1
|
|
36
|
+
print(" [PASS] " + name)
|
|
37
|
+
else:
|
|
38
|
+
failed += 1
|
|
39
|
+
print(" [FAIL] " + name + " " + detail)
|
|
40
|
+
|
|
41
|
+
|
|
42
|
+
def _setenv(**kv):
|
|
43
|
+
old = {}
|
|
44
|
+
for k, v in kv.items():
|
|
45
|
+
old[k] = os.environ.get(k)
|
|
46
|
+
if v is None:
|
|
47
|
+
os.environ.pop(k, None)
|
|
48
|
+
else:
|
|
49
|
+
os.environ[k] = v
|
|
50
|
+
return old
|
|
51
|
+
|
|
52
|
+
|
|
53
|
+
def _restore(old):
|
|
54
|
+
for k, v in old.items():
|
|
55
|
+
if v is None:
|
|
56
|
+
os.environ.pop(k, None)
|
|
57
|
+
else:
|
|
58
|
+
os.environ[k] = v
|
|
59
|
+
|
|
60
|
+
|
|
61
|
+
def _write_raw(root, nid, layer, content, pos=None, tw=None, bd=None):
|
|
62
|
+
fm = {"id": nid, "layer": layer, "modality": "text", "importance": 0.5,
|
|
63
|
+
"confidence": 0.6, "condition_space": {"time_window": tw or [0, 10 ** 12]},
|
|
64
|
+
"tags": [], "created_at": 0, "access_count": 0, "last_access": 0,
|
|
65
|
+
"edges": []}
|
|
66
|
+
if pos:
|
|
67
|
+
fm["condition_space"]["observation_position"] = pos
|
|
68
|
+
if bd:
|
|
69
|
+
fm["big_domain"] = bd
|
|
70
|
+
p = os.path.join(root, layer, nid + ".md")
|
|
71
|
+
os.makedirs(os.path.dirname(p), exist_ok=True)
|
|
72
|
+
io.open(p, "w", encoding="utf-8").write(nodefile.dumps(fm, content))
|
|
73
|
+
return p
|
|
74
|
+
|
|
75
|
+
|
|
76
|
+
def _ids(results):
|
|
77
|
+
return sorted(r[0]["id"] for r in results)
|
|
78
|
+
|
|
79
|
+
|
|
80
|
+
def main():
|
|
81
|
+
# ---- 1) routing 侧 ----
|
|
82
|
+
check("domain_terms 空值 -> []", routing.domain_terms("") == [])
|
|
83
|
+
ts = routing.domain_terms("工程 应力 桥梁 " * 3)
|
|
84
|
+
check("domain_terms 去重保序", ts == ["工程", "应力", "桥梁"], str(ts[:4]))
|
|
85
|
+
check("domain_terms limit 生效",
|
|
86
|
+
len(routing.domain_terms("工程 应力 桥梁 材料", limit=2)) == 2)
|
|
87
|
+
check("classify_text 有信号 -> 工程",
|
|
88
|
+
routing.classify_text(GONG) == "工程", str(routing.classify_text(GONG)))
|
|
89
|
+
check("classify_text 无信号 -> None", routing.classify_text("xyzzy zzz") is None)
|
|
90
|
+
|
|
91
|
+
# ---- 2) 写入侧:默认关不写,开总开关才写 ----
|
|
92
|
+
root = tempfile.mkdtemp(prefix="retr_s1_")
|
|
93
|
+
old = _setenv(MDCG_RETRIEVAL_PIPELINE=None, MDCG_GATE_S1_DOMAIN=None,
|
|
94
|
+
MDCG_GATE_S2_COND=None)
|
|
95
|
+
cg = MdCG(root)
|
|
96
|
+
cg.add("n_off", GONG, "knowledge", importance=0.9)
|
|
97
|
+
cg.add("n_yi", YI, "knowledge", importance=0.5)
|
|
98
|
+
cg.flush()
|
|
99
|
+
check("默认关:frontmatter 不写 big_domain",
|
|
100
|
+
"big_domain" not in (cg.get("n_off")["frontmatter"] or {}))
|
|
101
|
+
check("默认关:索引快照不含该键",
|
|
102
|
+
"big_domain" not in (cg.index["nodes"]["n_off"] or {}))
|
|
103
|
+
check("默认关:索引快照不含 observation_position 键",
|
|
104
|
+
"observation_position" not in (cg.index["nodes"]["n_off"] or {}))
|
|
105
|
+
|
|
106
|
+
# ---- 3) 默认关:meta 逐字节等价 ----
|
|
107
|
+
_r0, meta = cg.search("工程 应力", k=10, judge=False, record=False)
|
|
108
|
+
check("默认关:meta 不含 gates 键", "gates" not in meta, str(sorted(meta.keys())))
|
|
109
|
+
check("默认关:scanned=全量", meta.get("scanned") == 2, str(meta.get("scanned")))
|
|
110
|
+
ids_off = _ids(_r0)
|
|
111
|
+
|
|
112
|
+
# 开总开关后写入 → 落域标签(索引同口径)
|
|
113
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE="1", MDCG_GATE_S1_DOMAIN="1",
|
|
114
|
+
MDCG_GATE_S2_COND="0")
|
|
115
|
+
cg.add("n_gong", GONG + " 坝体受力", "knowledge", importance=0.8)
|
|
116
|
+
cg.add("n_none", "xyzzy zzz", "knowledge")
|
|
117
|
+
cg.flush()
|
|
118
|
+
check("开关开:写入落 big_domain",
|
|
119
|
+
cg.get("n_gong")["frontmatter"].get("big_domain") == "工程")
|
|
120
|
+
check("开关开:索引快照同口径",
|
|
121
|
+
cg.index["nodes"]["n_gong"].get("big_domain") == "工程")
|
|
122
|
+
check("开关开:无信号仍不写该字段",
|
|
123
|
+
"big_domain" not in (cg.get("n_none")["frontmatter"] or {}))
|
|
124
|
+
|
|
125
|
+
# ---- 4) S1:收敛 / 回退 / 无域信号 ----
|
|
126
|
+
# 收敛要以「已标域」为前提(未标域节点按兜底池保留)→ 先回填 cg 的存量节点
|
|
127
|
+
st_bf = cg.backfill_big_domain()
|
|
128
|
+
check("回填后:存量节点获得域标签",
|
|
129
|
+
cg.index["nodes"]["n_off"].get("big_domain") == "工程"
|
|
130
|
+
and cg.index["nodes"]["n_yi"].get("big_domain") == "医学", str(st_bf))
|
|
131
|
+
_r1, meta = cg.search("工程 应力", k=10, judge=False, record=False)
|
|
132
|
+
s1 = (meta.get("gates") or {}).get("s1") or {}
|
|
133
|
+
check("S1 收敛到工程域", s1.get("domain") == "工程", str(s1))
|
|
134
|
+
check("S1 丢弃异域候选、保留未标域",
|
|
135
|
+
s1.get("dropped") == 1 and s1.get("in") == 3, str(s1))
|
|
136
|
+
check("S1 scanned 下降(4 -> 3)", meta.get("scanned") == 3,
|
|
137
|
+
str(meta.get("scanned")))
|
|
138
|
+
|
|
139
|
+
root_fb = tempfile.mkdtemp(prefix="retr_s1fb_")
|
|
140
|
+
_write_raw(root_fb, "g1", "knowledge", GONG)
|
|
141
|
+
_write_raw(root_fb, "y1", "knowledge", YI)
|
|
142
|
+
_write_raw(root_fb, "u1", "knowledge", "xyzzy zzz") # 无域信号 → 兜底池
|
|
143
|
+
cgfb = MdCG(root_fb)
|
|
144
|
+
cgfb.rebuild_index()
|
|
145
|
+
cgfb.backfill_big_domain()
|
|
146
|
+
_r2, meta = cgfb.search("工程 应力", k=5, min_results=3, judge=False, record=False)
|
|
147
|
+
s1 = (meta.get("gates") or {}).get("s1") or {}
|
|
148
|
+
check("S1 域内不足 -> 回退全量", s1.get("fallback") == "insufficient", str(s1))
|
|
149
|
+
check("S1 回退后 scanned=全量", meta.get("scanned") == 3, str(meta.get("scanned")))
|
|
150
|
+
|
|
151
|
+
# 兜底召回:未标域节点必须留在域内集合里(否则历史节点被域收敛整片丢弃)
|
|
152
|
+
_r3b, meta = cgfb.search("工程 应力", k=1, judge=False, record=False)
|
|
153
|
+
s1 = (meta.get("gates") or {}).get("s1") or {}
|
|
154
|
+
check("S1 保留未标域节点(兜底池)",
|
|
155
|
+
s1.get("in") == 2 and s1.get("dropped") == 1 and meta.get("scanned") == 2,
|
|
156
|
+
str(s1) + " scanned=" + str(meta.get("scanned")))
|
|
157
|
+
|
|
158
|
+
_r3, meta = cgfb.search("xyzzy zzz", k=10, judge=False, record=False)
|
|
159
|
+
s1 = (meta.get("gates") or {}).get("s1") or {}
|
|
160
|
+
check("S1 无域信号 -> 不收敛",
|
|
161
|
+
s1.get("reason") == "no_domain_signal" and meta.get("scanned") == 3, str(s1))
|
|
162
|
+
|
|
163
|
+
# ---- 5) S2:位置 / 时间窗 / 信息不足 / 全滤回退 ----
|
|
164
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE="1", MDCG_GATE_S1_DOMAIN="0",
|
|
165
|
+
MDCG_GATE_S2_COND="1")
|
|
166
|
+
root2 = tempfile.mkdtemp(prefix="retr_s2_")
|
|
167
|
+
_write_raw(root2, "a", "knowledge", GONG, pos="工程 结构", tw=[0, 10 ** 12])
|
|
168
|
+
_write_raw(root2, "b", "knowledge", YI, pos="医学 疾病", tw=[0, 10 ** 12])
|
|
169
|
+
cg2 = MdCG(root2)
|
|
170
|
+
cg2.rebuild_index()
|
|
171
|
+
|
|
172
|
+
_r, meta = cg2.search("xyzzy", k=10, judge=False, record=False,
|
|
173
|
+
context={"observation_position": "工程 结构"})
|
|
174
|
+
s2 = (meta.get("gates") or {}).get("s2") or {}
|
|
175
|
+
check("S2 位置不匹配被门控", s2.get("dropped") == 1, str(s2))
|
|
176
|
+
|
|
177
|
+
_r, meta = cg2.search("xyzzy", k=10, judge=False, record=False, context={})
|
|
178
|
+
s2 = (meta.get("gates") or {}).get("s2") or {}
|
|
179
|
+
check("S2 信息不足一律放行", s2.get("dropped") == 0, str(s2))
|
|
180
|
+
|
|
181
|
+
_r, meta = cg2.search("xyzzy", k=10, judge=False, record=False,
|
|
182
|
+
context={"time_window": FAR_TW})
|
|
183
|
+
s2 = (meta.get("gates") or {}).get("s2") or {}
|
|
184
|
+
check("S2 时间窗不相交被门控", s2.get("dropped") == 2, str(s2))
|
|
185
|
+
|
|
186
|
+
_r, meta = cg2.search("xyzzy", k=10, judge=False, record=False,
|
|
187
|
+
context={"time_window": [0, 100]})
|
|
188
|
+
s2 = (meta.get("gates") or {}).get("s2") or {}
|
|
189
|
+
check("S2 时间窗相交放行", s2.get("dropped") == 0, str(s2))
|
|
190
|
+
|
|
191
|
+
_r, meta = cg2.search("xyzzy", k=10, judge=False, record=False,
|
|
192
|
+
context={"observation_position": "艺术 绘画"})
|
|
193
|
+
s2 = (meta.get("gates") or {}).get("s2") or {}
|
|
194
|
+
check("S2 全滤 -> 回退(宁多勿漏)",
|
|
195
|
+
s2.get("fallback") == "empty" and meta.get("scanned") == 2, str(s2))
|
|
196
|
+
_restore(old)
|
|
197
|
+
|
|
198
|
+
# ---- 6) backfill(运维口径:先开总开关,再补齐存量)----
|
|
199
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE="1")
|
|
200
|
+
# 6a 默认关时回填必须是 no-op(不写 fm、不写索引)
|
|
201
|
+
root6 = tempfile.mkdtemp(prefix="retr_bf0_")
|
|
202
|
+
_write_raw(root6, "g9", "knowledge", GONG)
|
|
203
|
+
cg6 = MdCG(root6)
|
|
204
|
+
cg6.rebuild_index()
|
|
205
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE=None)
|
|
206
|
+
st6 = cg6.backfill_big_domain()
|
|
207
|
+
check("默认关:回填 no-op(skipped)",
|
|
208
|
+
st6.get("skipped") == "pipeline_disabled" and st6["written"] == 0,
|
|
209
|
+
str(st6))
|
|
210
|
+
check("默认关:回填不改 fm/索引",
|
|
211
|
+
"big_domain" not in (cg6.get("g9")["frontmatter"] or {})
|
|
212
|
+
and "big_domain" not in cg6.index["nodes"]["g9"])
|
|
213
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE="1")
|
|
214
|
+
|
|
215
|
+
root3 = tempfile.mkdtemp(prefix="retr_bf_")
|
|
216
|
+
_write_raw(root3, "g1", "knowledge", GONG)
|
|
217
|
+
_write_raw(root3, "y1", "knowledge", YI)
|
|
218
|
+
_write_raw(root3, "x1", "knowledge", "xyzzy")
|
|
219
|
+
cg3 = MdCG(root3)
|
|
220
|
+
cg3.rebuild_index()
|
|
221
|
+
st1 = cg3.backfill_big_domain(dry_run=True)
|
|
222
|
+
check("backfill dry-run 统计",
|
|
223
|
+
st1["written"] == 2 and st1["no_signal"] == 1, str(st1))
|
|
224
|
+
check("backfill dry-run 不改盘",
|
|
225
|
+
cg3.get("g1")["frontmatter"].get("big_domain") is None)
|
|
226
|
+
st2 = cg3.backfill_big_domain()
|
|
227
|
+
check("backfill 真跑落盘",
|
|
228
|
+
st2["written"] == 2
|
|
229
|
+
and cg3.get("g1")["frontmatter"].get("big_domain") == "工程", str(st2))
|
|
230
|
+
check("backfill 同步索引",
|
|
231
|
+
cg3.index["nodes"]["y1"].get("big_domain") == "医学")
|
|
232
|
+
st3 = cg3.backfill_big_domain()
|
|
233
|
+
check("backfill 幂等",
|
|
234
|
+
st3["written"] == 0 and st3["already"] == 2
|
|
235
|
+
and st3["no_signal"] == 1, str(st3))
|
|
236
|
+
# 索引持久化:重开库(走 _index_log 重放)后条目仍带域标签
|
|
237
|
+
cg3r = MdCG(root3)
|
|
238
|
+
check("backfill 索引持久化(重载可见)",
|
|
239
|
+
cg3r.index["nodes"]["g1"].get("big_domain") == "工程",
|
|
240
|
+
str(cg3r.index["nodes"]["g1"].get("big_domain")))
|
|
241
|
+
|
|
242
|
+
# 对账支路(真状态):**磁盘索引缺键**而 frontmatter 有 → 只修索引不重写文件,且修完重载可见。
|
|
243
|
+
# 造法:rebuild 写出的磁盘索引里删掉该键并清空索引日志(否则日志重放会把键带回来,测试就变成自证)。
|
|
244
|
+
root4 = tempfile.mkdtemp(prefix="retr_bf2_")
|
|
245
|
+
_write_raw(root4, "g1", "knowledge", GONG, bd="工程")
|
|
246
|
+
cg4 = MdCG(root4)
|
|
247
|
+
cg4.rebuild_index()
|
|
248
|
+
idx = json.load(io.open(cg4.index_path, encoding="utf-8"))
|
|
249
|
+
idx["nodes"]["g1"].pop("big_domain", None)
|
|
250
|
+
io.open(cg4.index_path, "w", encoding="utf-8").write(
|
|
251
|
+
json.dumps(idx, ensure_ascii=False))
|
|
252
|
+
ShardedLog.clear(cg4.index_log_dir)
|
|
253
|
+
cg4 = MdCG(root4) # 重载:内存索引确实缺键,而 fm 有
|
|
254
|
+
check("对账前状态成立(索引缺键 / fm 有键)",
|
|
255
|
+
cg4.index["nodes"]["g1"].get("big_domain") is None
|
|
256
|
+
and cg4.get("g1")["frontmatter"].get("big_domain") == "工程")
|
|
257
|
+
st4d = cg4.backfill_big_domain(dry_run=True)
|
|
258
|
+
check("对账 dry-run:只统计不改内存",
|
|
259
|
+
st4d["index_synced"] == 1
|
|
260
|
+
and "big_domain" not in cg4.index["nodes"]["g1"], str(st4d))
|
|
261
|
+
st4 = cg4.backfill_big_domain()
|
|
262
|
+
check("backfill 对账支路修索引",
|
|
263
|
+
st4["index_synced"] == 1 and st4["written"] == 0, str(st4))
|
|
264
|
+
cg4r = MdCG(root4)
|
|
265
|
+
check("对账后重载索引可见(真持久化)",
|
|
266
|
+
cg4r.index["nodes"]["g1"].get("big_domain") == "工程",
|
|
267
|
+
str(cg4r.index["nodes"]["g1"].get("big_domain")))
|
|
268
|
+
|
|
269
|
+
# ---- 7) 默认口径等价(真比较,不是自证)----
|
|
270
|
+
# 7a 无候选分支:开关关时 meta 不得多出 gates 键
|
|
271
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE=None, MDCG_GATE_S1_DOMAIN=None,
|
|
272
|
+
MDCG_GATE_S2_COND=None)
|
|
273
|
+
_r5, meta5 = cg.search("xyzzy", layer="goals", k=10, judge=False,
|
|
274
|
+
record=False)
|
|
275
|
+
check("默认关:无候选分支 meta 无 gates 键",
|
|
276
|
+
meta5.get("reason") == "no_candidates" and "gates" not in meta5,
|
|
277
|
+
str(meta5))
|
|
278
|
+
# 7b 同库同查询:关→开→关 三次,首末两次的 results/meta 必须完全一致
|
|
279
|
+
_ra, meta_a = cg.search("工程 应力", k=10, judge=False, record=False)
|
|
280
|
+
snap_a = ([(r[0]["id"], round(float(r[1]), 6), sorted(r[0].keys()))
|
|
281
|
+
for r in _ra],
|
|
282
|
+
{k: v for k, v in meta_a.items() if k != "gates"})
|
|
283
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE="1", MDCG_GATE_S1_DOMAIN="1",
|
|
284
|
+
MDCG_GATE_S2_COND="1")
|
|
285
|
+
cg.search("工程 应力", k=10, judge=False, record=False)
|
|
286
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE=None, MDCG_GATE_S1_DOMAIN=None,
|
|
287
|
+
MDCG_GATE_S2_COND=None)
|
|
288
|
+
_rb, meta_b = cg.search("工程 应力", k=10, judge=False, record=False)
|
|
289
|
+
snap_b = ([(r[0]["id"], round(float(r[1]), 6), sorted(r[0].keys()))
|
|
290
|
+
for r in _rb],
|
|
291
|
+
{k: v for k, v in meta_b.items() if k != "gates"})
|
|
292
|
+
check("默认口径等价:关→开→关 首末一致", snap_a == snap_b,
|
|
293
|
+
str(snap_a)[:160] + " vs " + str(snap_b)[:160])
|
|
294
|
+
check("默认关:meta 仍不含 gates 键", "gates" not in meta_b,
|
|
295
|
+
str(sorted(meta_b.keys())))
|
|
296
|
+
|
|
297
|
+
# 7c 候选 entry 形状:默认关时即便节点带 observation_position/big_domain,
|
|
298
|
+
# 返回的候选也不得平铺这两个键(独立复核指出的默认口径变化点)
|
|
299
|
+
root5 = tempfile.mkdtemp(prefix="retr_shape_")
|
|
300
|
+
_write_raw(root5, "p1", "knowledge", GONG, pos="工程 结构", bd="工程")
|
|
301
|
+
cg5 = MdCG(root5) # 默认关下建索引
|
|
302
|
+
cg5.rebuild_index()
|
|
303
|
+
check("默认关:索引条目不落门控字段",
|
|
304
|
+
"big_domain" not in cg5.index["nodes"]["p1"]
|
|
305
|
+
and "observation_position" not in cg5.index["nodes"]["p1"],
|
|
306
|
+
str(sorted(cg5.index["nodes"]["p1"].keys())))
|
|
307
|
+
_r5b, _m5b = cg5.search("工程 应力", k=10, judge=False, record=False)
|
|
308
|
+
check("默认关:候选 entry 无门控字段",
|
|
309
|
+
all(("big_domain" not in r[0] and "observation_position" not in r[0])
|
|
310
|
+
for r in _r5b),
|
|
311
|
+
str([sorted(r[0].keys()) for r in _r5b])[:160])
|
|
312
|
+
|
|
313
|
+
# 7d 索引曾在「开启态」构建(条目带门控键)→ 默认关时 search 必须返回剥离后的候选
|
|
314
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE="1")
|
|
315
|
+
root7 = tempfile.mkdtemp(prefix="retr_resid_")
|
|
316
|
+
_write_raw(root7, "q1", "knowledge", GONG, pos="工程 结构", bd="工程")
|
|
317
|
+
cg7 = MdCG(root7)
|
|
318
|
+
cg7.rebuild_index()
|
|
319
|
+
check("开启态:索引条目带门控键",
|
|
320
|
+
cg7.index["nodes"]["q1"].get("big_domain") == "工程"
|
|
321
|
+
and cg7.index["nodes"]["q1"].get("observation_position") == "工程 结构")
|
|
322
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE=None)
|
|
323
|
+
_r7, _m7 = cg7.search("工程 应力", k=10, judge=False, record=False)
|
|
324
|
+
check("默认关:残留门控键被剥离(候选)",
|
|
325
|
+
bool(_r7) and all(("big_domain" not in r[0]
|
|
326
|
+
and "observation_position" not in r[0])
|
|
327
|
+
for r in _r7),
|
|
328
|
+
str([sorted(r[0].keys()) for r in _r7])[:160])
|
|
329
|
+
check("默认关:索引条目本身未被就地改写",
|
|
330
|
+
cg7.index["nodes"]["q1"].get("big_domain") == "工程")
|
|
331
|
+
_restore(old)
|
|
332
|
+
|
|
333
|
+
print("\ntest_retr_s1: %d 通过 / %d 失败" % (passed, failed))
|
|
334
|
+
return 0 if failed == 0 else 1
|
|
335
|
+
|
|
336
|
+
|
|
337
|
+
if __name__ == "__main__":
|
|
338
|
+
if hasattr(sys.stdout, "reconfigure"):
|
|
339
|
+
sys.stdout.reconfigure(encoding="utf-8")
|
|
340
|
+
sys.exit(main())
|
|
@@ -0,0 +1,209 @@
|
|
|
1
|
+
# -*- coding: utf-8 -*-
|
|
2
|
+
"""S1b 细口径桶收敛探针(脚本式,exit code 定成败)。
|
|
3
|
+
|
|
4
|
+
用法:python -X utf8 -m md_cg.test_retr_s1b
|
|
5
|
+
覆盖:
|
|
6
|
+
- bucket_key_readable:'cond_感知系统_d94e90d2' → '感知系统';orphan/空 → '';无哈希段保持
|
|
7
|
+
- 默认关:无 gates 键;关→开→关 结果一致;子开关未设不生效
|
|
8
|
+
- S1b 开:query 侧推断出候选桶 → 候选收敛(out < in)、**orphan/无桶恒留兜底**、不引入任何新节点
|
|
9
|
+
- 命中不到任何桶 → no_key_match 且不收敛(回退)
|
|
10
|
+
- topk / min_sim 可配(topk=0 关闭;min_sim=1.0 只认完全同名键)
|
|
11
|
+
- 幂等:重复检索 / 索引重建后同口径
|
|
12
|
+
"""
|
|
13
|
+
import os
|
|
14
|
+
import sys
|
|
15
|
+
import tempfile
|
|
16
|
+
|
|
17
|
+
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
|
18
|
+
|
|
19
|
+
from md_cg import routing # noqa: E402
|
|
20
|
+
from md_cg.mdcg import MdCG # noqa: E402
|
|
21
|
+
|
|
22
|
+
passed = 0
|
|
23
|
+
failed = 0
|
|
24
|
+
QUERY = "感知系统 记忆 检索"
|
|
25
|
+
|
|
26
|
+
|
|
27
|
+
def check(name, cond, detail=""):
|
|
28
|
+
global passed, failed
|
|
29
|
+
if cond:
|
|
30
|
+
passed += 1
|
|
31
|
+
print(" [PASS] " + name)
|
|
32
|
+
else:
|
|
33
|
+
failed += 1
|
|
34
|
+
print(" [FAIL] " + name + " " + detail)
|
|
35
|
+
|
|
36
|
+
|
|
37
|
+
def _setenv(**kv):
|
|
38
|
+
old = {}
|
|
39
|
+
for k, v in kv.items():
|
|
40
|
+
old[k] = os.environ.get(k)
|
|
41
|
+
if v is None:
|
|
42
|
+
os.environ.pop(k, None)
|
|
43
|
+
else:
|
|
44
|
+
os.environ[k] = v
|
|
45
|
+
return old
|
|
46
|
+
|
|
47
|
+
|
|
48
|
+
def _restore(old):
|
|
49
|
+
for k, v in old.items():
|
|
50
|
+
if v is None:
|
|
51
|
+
os.environ.pop(k, None)
|
|
52
|
+
else:
|
|
53
|
+
os.environ[k] = v
|
|
54
|
+
|
|
55
|
+
|
|
56
|
+
def _ids(results):
|
|
57
|
+
return sorted(r[0]["id"] for r in results)
|
|
58
|
+
|
|
59
|
+
|
|
60
|
+
def _build(root):
|
|
61
|
+
cg = MdCG(root)
|
|
62
|
+
cg.add("s1", "感知系统 记忆 节点 检索", "knowledge", tags=["domain:感知系统"])
|
|
63
|
+
cg.add("s2", "感知系统 缓存 命中", "knowledge", tags=["domain:感知系统"])
|
|
64
|
+
cg.add("g1", "图数据库 遍历 邻接", "knowledge", tags=["domain:图数据库"])
|
|
65
|
+
cg.add("o1", "感知系统 但无域标签(落 orphan)", "knowledge")
|
|
66
|
+
cg.add("c1", "感知系统 情境层(不参与分桶)", "contextual")
|
|
67
|
+
cg.flush()
|
|
68
|
+
return cg
|
|
69
|
+
|
|
70
|
+
|
|
71
|
+
def main():
|
|
72
|
+
# ---- 1) key 解析 ----
|
|
73
|
+
check("bucket_key_readable 剥离哈希",
|
|
74
|
+
routing.bucket_key_readable("cond_感知系统_d94e90d2") == "感知系统")
|
|
75
|
+
check("bucket_key_readable orphan/空 → ''",
|
|
76
|
+
routing.bucket_key_readable("orphan") == ""
|
|
77
|
+
and routing.bucket_key_readable("") == ""
|
|
78
|
+
and routing.bucket_key_readable(None) == "")
|
|
79
|
+
check("bucket_key_readable 无哈希段保持",
|
|
80
|
+
routing.bucket_key_readable("cond_letta-code") == "letta-code")
|
|
81
|
+
|
|
82
|
+
root = tempfile.mkdtemp(prefix="retr_s1b_")
|
|
83
|
+
cg = _build(root)
|
|
84
|
+
b_sense = [e.get("bucket") for e in cg.index["nodes"].values()
|
|
85
|
+
if e.get("path", "").endswith("s1.md")][0]
|
|
86
|
+
check("写入侧确实落了桶", bool(b_sense) and b_sense != "orphan", str(b_sense))
|
|
87
|
+
|
|
88
|
+
# ---- 2) 默认关 ----
|
|
89
|
+
old = _setenv(MDCG_RETRIEVAL_PIPELINE=None, MDCG_GATE_S1B_BUCKET=None,
|
|
90
|
+
MDCG_GATE_S1_DOMAIN=None, MDCG_GATE_S2_COND=None,
|
|
91
|
+
MDCG_BUCKET_TOPK=None, MDCG_BUCKET_MIN_SIM=None)
|
|
92
|
+
r_off, m_off = cg.search(QUERY, k=10, judge=False, record=False)
|
|
93
|
+
base = _ids(r_off)
|
|
94
|
+
check("默认关:meta 不含 gates 键", "gates" not in m_off,
|
|
95
|
+
str(sorted(m_off.keys())))
|
|
96
|
+
|
|
97
|
+
# ---- 3) 子开关未设 → 不生效 ----
|
|
98
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE="1", MDCG_GATE_S1_DOMAIN="0",
|
|
99
|
+
MDCG_GATE_S2_COND="0", MDCG_GATE_S1B_BUCKET=None)
|
|
100
|
+
r_x, m_x = cg.search(QUERY, k=10, judge=False, record=False)
|
|
101
|
+
check("子开关未设:S1b 不生效",
|
|
102
|
+
"s1b" not in (m_x.get("gates") or {}) and _ids(r_x) == base,
|
|
103
|
+
str(m_x.get("gates")))
|
|
104
|
+
|
|
105
|
+
# ---- 4) S1b 开:桶收敛 + orphan 兜底 + 不引入新节点 ----
|
|
106
|
+
_setenv(MDCG_GATE_S1B_BUCKET="1", MDCG_BUCKET_TOPK="1", MDCG_BUCKET_MIN_SIM="0.34")
|
|
107
|
+
r1, m1 = cg.search(QUERY, k=10, judge=False, record=False)
|
|
108
|
+
g1 = (m1.get("gates") or {}).get("s1b") or {}
|
|
109
|
+
ids1 = _ids(r1)
|
|
110
|
+
check("S1b 推断出感知系统桶",
|
|
111
|
+
g1.get("keys") == [b_sense], str(g1))
|
|
112
|
+
check("S1b 收敛:out < in", g1.get("out", 99) < g1.get("in", 0), str(g1))
|
|
113
|
+
check("S1b 收敛后不引入任何新节点(⊆ 默认关结果集)",
|
|
114
|
+
set(ids1) <= set(base), str(ids1) + " vs " + str(base))
|
|
115
|
+
check("S1b 保留 orphan 兜底节点", "o1" in ids1, str(ids1))
|
|
116
|
+
check("S1b 丢弃异桶节点", "g1" not in ids1, str(ids1))
|
|
117
|
+
|
|
118
|
+
# ---- 5) 命中不到桶 → 回退不收敛(与「S1b 关」的同查询结果逐条比较,避免自证)----
|
|
119
|
+
_setenv(MDCG_GATE_S1B_BUCKET=None)
|
|
120
|
+
r2c, _m2c = cg.search("完全无关的词 词组", k=10, judge=False, record=False)
|
|
121
|
+
_setenv(MDCG_GATE_S1B_BUCKET="1")
|
|
122
|
+
r2, m2 = cg.search("完全无关的词 词组", k=10, judge=False, record=False)
|
|
123
|
+
g2 = (m2.get("gates") or {}).get("s1b") or {}
|
|
124
|
+
check("无键命中:no_key_match 且结果与关闭时一致",
|
|
125
|
+
g2.get("reason") == "no_key_match" and "out" not in g2
|
|
126
|
+
and _ids(r2) == _ids(r2c),
|
|
127
|
+
str(g2) + " " + str(_ids(r2)) + " vs " + str(_ids(r2c)))
|
|
128
|
+
# 5b) S4 显式关闭时 S1b 仍生效(证明门控未与 S4 耦合)
|
|
129
|
+
_setenv(MDCG_GATE_S4_LAYER="0")
|
|
130
|
+
r2d, m2d = cg.search(QUERY, k=10, judge=False, record=False)
|
|
131
|
+
g2d = (m2d.get("gates") or {}).get("s1b") or {}
|
|
132
|
+
check("S4 显式关闭时 S1b 仍生效",
|
|
133
|
+
g2d.get("keys") == [b_sense] and "s4" not in (m2d.get("gates") or {}),
|
|
134
|
+
str(g2d) + " gates=" + str(sorted((m2d.get("gates") or {}).keys())))
|
|
135
|
+
_setenv(MDCG_GATE_S4_LAYER=None)
|
|
136
|
+
|
|
137
|
+
# 5c) 候选不足 → 回退全量(专属小库:两个分桶各 1 节点、无 orphan;min_results=3)
|
|
138
|
+
root_ins = tempfile.mkdtemp(prefix="retr_s1bins_")
|
|
139
|
+
cgi = MdCG(root_ins)
|
|
140
|
+
cgi.add("a1", "感知系统 记忆", "knowledge", tags=["domain:感知系统"])
|
|
141
|
+
cgi.add("b1", "图数据库 邻接", "knowledge", tags=["domain:图数据库"])
|
|
142
|
+
cgi.flush()
|
|
143
|
+
bi = [e.get("bucket") for e in cgi.index["nodes"].values()
|
|
144
|
+
if e.get("path", "").endswith("a1.md")][0]
|
|
145
|
+
# 对照组必须用**同一个 min_results**(否则比的是两次不同检索,不是回退等价)
|
|
146
|
+
_setenv(MDCG_GATE_S1B_BUCKET=None)
|
|
147
|
+
r_ioff, _m = cgi.search(QUERY, k=10, judge=False, record=False, min_results=3)
|
|
148
|
+
_setenv(MDCG_GATE_S1B_BUCKET="1")
|
|
149
|
+
r_ins, m_ins = cgi.search(QUERY, k=10, judge=False, record=False, min_results=3)
|
|
150
|
+
gi = (m_ins.get("gates") or {}).get("s1b") or {}
|
|
151
|
+
check("候选不足 → fallback=insufficient",
|
|
152
|
+
gi.get("fallback") == "insufficient" and gi.get("keys") == [bi], str(gi))
|
|
153
|
+
check("回退时审计 out 记真实输出(= in,不回退口径误导)",
|
|
154
|
+
gi.get("out") == gi.get("in") == 2 and gi.get("would_keep") == 1, str(gi))
|
|
155
|
+
check("回退后结果与 S1b 关闭时一致(含异桶节点)",
|
|
156
|
+
_ids(r_ins) == _ids(r_ioff) and "b1" in _ids(r_ins),
|
|
157
|
+
str(_ids(r_ins)) + " vs " + str(_ids(r_ioff)))
|
|
158
|
+
|
|
159
|
+
# ---- 6) topk / min_sim 可配 ----
|
|
160
|
+
_setenv(MDCG_BUCKET_TOPK="0")
|
|
161
|
+
r3, m3 = cg.search(QUERY, k=10, judge=False, record=False)
|
|
162
|
+
g3 = (m3.get("gates") or {}).get("s1b") or {}
|
|
163
|
+
check("topk=0:关闭且候选不变",
|
|
164
|
+
g3.get("reason") == "disabled_by_topk" and _ids(r3) == base, str(g3))
|
|
165
|
+
_setenv(MDCG_BUCKET_TOPK="2", MDCG_BUCKET_MIN_SIM="0.0")
|
|
166
|
+
r4, m4 = cg.search(QUERY, k=10, judge=False, record=False)
|
|
167
|
+
g4 = (m4.get("gates") or {}).get("s1b") or {}
|
|
168
|
+
check("topk=2 + min_sim=0:取到 2 个桶",
|
|
169
|
+
len(g4.get("keys") or []) == 2, str(g4))
|
|
170
|
+
_setenv(MDCG_BUCKET_TOPK="1", MDCG_BUCKET_MIN_SIM="1.0")
|
|
171
|
+
r5, m5 = cg.search("感知", k=10, judge=False, record=False)
|
|
172
|
+
g5 = (m5.get("gates") or {}).get("s1b") or {}
|
|
173
|
+
check("min_sim=1.0:部分匹配不入选", g5.get("reason") == "no_key_match", str(g5))
|
|
174
|
+
_setenv(MDCG_BUCKET_MIN_SIM="0.34")
|
|
175
|
+
r6, m6 = cg.search("感知", k=10, judge=False, record=False)
|
|
176
|
+
g6 = (m6.get("gates") or {}).get("s1b") or {}
|
|
177
|
+
check("min_sim=0.34:部分匹配(感知→感知系统 0.5)入选",
|
|
178
|
+
g6.get("keys") == [b_sense], str(g6))
|
|
179
|
+
|
|
180
|
+
# ---- 7) 幂等 ----
|
|
181
|
+
_setenv(MDCG_BUCKET_TOPK="1", MDCG_BUCKET_MIN_SIM="0.34")
|
|
182
|
+
ra, ma = cg.search(QUERY, k=10, judge=False, record=False)
|
|
183
|
+
rb, mb = cg.search(QUERY, k=10, judge=False, record=False)
|
|
184
|
+
check("S1b 幂等:重复检索同口径",
|
|
185
|
+
_ids(ra) == _ids(rb)
|
|
186
|
+
and (ma.get("gates") or {}).get("s1b") == (mb.get("gates") or {}).get("s1b"))
|
|
187
|
+
cg.rebuild_index()
|
|
188
|
+
rc, mc = cg.search(QUERY, k=10, judge=False, record=False)
|
|
189
|
+
check("S1b 幂等:索引重建后同口径",
|
|
190
|
+
_ids(rc) == _ids(ra)
|
|
191
|
+
and (mc.get("gates") or {}).get("s1b") == (ma.get("gates") or {}).get("s1b"))
|
|
192
|
+
|
|
193
|
+
# ---- 8) 默认口径等价 ----
|
|
194
|
+
_setenv(MDCG_RETRIEVAL_PIPELINE=None, MDCG_GATE_S1B_BUCKET=None,
|
|
195
|
+
MDCG_GATE_S1_DOMAIN=None, MDCG_GATE_S2_COND=None,
|
|
196
|
+
MDCG_BUCKET_TOPK=None, MDCG_BUCKET_MIN_SIM=None)
|
|
197
|
+
r9, m9 = cg.search(QUERY, k=10, judge=False, record=False)
|
|
198
|
+
check("默认口径等价:关→开→关 首末一致",
|
|
199
|
+
_ids(r9) == base and "gates" not in m9, str(_ids(r9)))
|
|
200
|
+
_restore(old)
|
|
201
|
+
|
|
202
|
+
print("\ntest_retr_s1b: %d 通过 / %d 失败" % (passed, failed))
|
|
203
|
+
return 0 if failed == 0 else 1
|
|
204
|
+
|
|
205
|
+
|
|
206
|
+
if __name__ == "__main__":
|
|
207
|
+
if hasattr(sys.stdout, "reconfigure"):
|
|
208
|
+
sys.stdout.reconfigure(encoding="utf-8")
|
|
209
|
+
sys.exit(main())
|