fix(eda holes): canonical sample classifier + style normalization

- models.py: add classify_sample_family() SSOT (token-aware, DM-suffix gated,
  strong-contains fallback) + extend SAMPLE_FAMILIES cues → unclassified 89→34
  uses (~93% classified); honest unknowns (armora, 90s_matrix) stay None.
- models.py: STYLE_ALIASES + norm_style() merges nu-jazz/nujazz, breaks/breakbeat,
  chip/chiptune, hip-hop/hip so the style chart is honest.
- tide_eda + tests use the canonical classifier (DRY); +3 mechanical tests
  (token cases, no-overreach incl. the shil-'oh' FP, style norm). 53 pass.
- regen tokens (cues feed match[] arrays).

armada/tide-table/eda_report.json

@@ -1148,27 +1148,22 @@
   "2026": 3
  },
  "styles": {
-  "dnb": 43,
-  "nujazz": 34,
+  "dnb": 45,
+  "breaks": 41,
+  "nujazz": 40,
   "techno": 31,
-  "breaks": 31,
   "lounge": 24,
   "ambient": 11,
-  "breakbeat": 10,
   "lofi": 7,
-  "nu-jazz": 6,
+  "acid": 6,
   "drill": 5,
-  "acid": 4,
   "punk": 3,
+  "chiptune": 2,
   "collab": 2,
-  "hip-hop": 2,
-  "punk-dnb": 2,
-  "acid-techno": 2,
+  "hiphop": 2,
   "hybrid": 2,
-  "chiptune": 1,
   "funk": 1,
   "downtempo": 1,
-  "chip": 1,
   "dance": 1
  },
  "palette_top": {
@@ -1199,34 +1194,35 @@
   "sn": 5
  },
  "families": {
-  "break": 106,
-  "synth": 57,
-  "snare": 54,
-  "bass": 51,
-  "kick": 23,
-  "fx": 19,
-  "keys": 18,
-  "hat": 13,
-  "vox": 5,
-  "pad": 1,
-  "lead": 1
+  "break": 112,
+  "synth": 85,
+  "snare": 72,
+  "bass": 64,
+  "hat": 39,
+  "keys": 37,
+  "kick": 27,
+  "vox": 20,
+  "fx": 18,
+  "perc": 9,
+  "pad": 5,
+  "lead": 4
  },
  "unclassified_top": {
-  "h2ogmhh": 20,
-  "90s_synatm": 9,
-  "giorgio_syn": 7,
-  "vec1_snare": 7,
-  "cbow": 6,
-  "fbass": 6,
-  "drum": 5,
-  "vec1_acid": 5,
-  "fguitar": 5,
-  "h2ogmsn": 4,
-  "vec2_synth_acid": 3,
   "armora": 3,
-  "rhadamanthe_vocal": 3,
-  "h2ogmcp": 3,
-  "vec1_claps": 3
+  "dr": 3,
+  "90s_matrix": 3,
+  "praise": 2,
+  "fsynth": 2,
+  "superfork": 2,
+  "90s_megafx": 2,
+  "shiloh": 2,
+  "supersiren": 2,
+  "nujazz_beats120": 2,
+  "ifdrums": 1,
+  "rhadamanthe_melo": 1,
+  "ccc": 1,
+  "ghost": 1,
+  "jane_wang": 1
  },
  "idioms_top": [
   {

armada/tide-table/models.py

@@ -15,6 +15,7 @@ read. Enums serialize as their string value.
 """
 from __future__ import annotations
 
+import re
 from datetime import date
 from enum import Enum
 from typing import Optional
@@ -519,19 +520,67 @@ class ColorFamily(BaseModel):
 # conceptual identity, NOT a measured register claim (that's ROLE_FAMILIES, by audio).
 SAMPLE_FAMILIES = [
     ColorFamily(key="kick",  label="Kick",   glyph="●", hue=25,  match=["kick", "kik", "bd", "808bd", "909", "bassdrum"]),
-    ColorFamily(key="snare", label="Snare",  glyph="◆", hue=50,  match=["snare", "sn", "sd", "clap", "cp", "rim", "rs"]),
-    ColorFamily(key="perc",  label="Perc",   glyph="▴", hue=80,  match=["perc", "conga", "bongo", "tom", "clave", "shaker", "tabla", "cowbell"]),
-    ColorFamily(key="hat",   label="Hat",    glyph="✦", hue=110, match=["hat", "hh", "oh", "ch", "hihat", "cymbal", "cym", "ride", "crash"]),
-    ColorFamily(key="break", label="Break",  glyph="≈", hue=150, match=["break", "amen", "loop", "jungle", "dnb", "jazz"]),
-    ColorFamily(key="pad",   label="Pad",    glyph="◌", hue=180, match=["pad", "drone", "choir", "string", "ambient", "atmos"]),
-    ColorFamily(key="keys",  label="Keys",   glyph="♬", hue=205, match=["key", "keys", "piano", "rhodes", "epiano", "organ", "fpiano", "qstab", "nujazz"]),
-    ColorFamily(key="lead",  label="Lead",   glyph="♪", hue=230, match=["lead", "arp", "pluck", "stab", "blip", "saw", "square"]),
-    ColorFamily(key="synth", label="Synth",  glyph="◈", hue=260, match=["synth", "commodore", "chip", "fm", "moog", "reese"]),
-    ColorFamily(key="bass",  label="Bass",   glyph="▂", hue=290, match=["bass", "sub", "808", "acid", "wobble", "meth_bass", "ramplem", "bassline"]),
-    ColorFamily(key="fx",    label="FX",     glyph="✺", hue=320, match=["fx", "riser", "sweep", "noise", "impact", "downlifter", "uplifter", "glitch", "vinyl", "foley"]),
-    ColorFamily(key="vox",   label="Vox",    glyph="◍", hue=355, match=["vox", "vocal", "voc", "acap", "speech", "voice"]),
+    ColorFamily(key="snare", label="Snare",  glyph="◆", hue=50,  match=["snare", "sn", "sd", "clap", "claps", "cp", "rim", "rs"]),
+    ColorFamily(key="perc",  label="Perc",   glyph="▴", hue=80,  match=["perc", "conga", "bongo", "tom", "clave", "shaker", "tabla", "cowbell", "drum"]),
+    ColorFamily(key="hat",   label="Hat",    glyph="✦", hue=110, match=["hat", "hh", "ho", "oh", "ch", "hihat", "cymbal", "cym", "ride", "crash"]),
+    ColorFamily(key="break", label="Break",  glyph="≈", hue=150, match=["break", "amen", "loop", "jungle", "dnb", "jazz", "breaks165", "fbreak"]),
+    ColorFamily(key="pad",   label="Pad",    glyph="◌", hue=180, match=["pad", "drone", "choir", "string", "ambient", "atmos", "atm", "airport", "trance"]),
+    ColorFamily(key="keys",  label="Keys",   glyph="♬", hue=205, match=["key", "keys", "piano", "rhodes", "epiano", "organ", "fpiano", "qstab", "cbow", "cpluck", "clav", "marimba", "mandolin", "guitar", "organ", "forgan"]),
+    ColorFamily(key="lead",  label="Lead",   glyph="♪", hue=230, match=["lead", "arp", "pluck", "stab", "blip", "saw", "square", "brass", "sax", "horn", "trump", "tromb"]),
+    ColorFamily(key="synth", label="Synth",  glyph="◈", hue=260, match=["synth", "syn", "commodore", "chip", "fm", "moog", "reese", "giorgio", "electro", "cs80", "vec"]),
+    ColorFamily(key="bass",  label="Bass",   glyph="▂", hue=290, match=["bass", "sub", "808", "acid", "wobble", "meth_bass", "ramplem", "bassline", "fbass"]),
+    ColorFamily(key="fx",    label="FX",     glyph="✺", hue=320, match=["fx", "riser", "risers", "sweep", "noise", "impact", "downlifter", "uplifter", "glitch", "vinyl", "foley"]),
+    ColorFamily(key="vox",   label="Vox",    glyph="◍", hue=355, match=["vox", "vocal", "voc", "acap", "speech", "voice", "movie", "dialog", "dialogs"]),
 ]
 
+# Drum-machine name SUFFIXES (h2ogmhh, …) and strong embedded tokens, for names that
+# don't lead with their family cue. Order matters: these run after the cue loop.
+_DM_SUFFIX = {"hh": "hat", "oh": "hat", "ch": "hat", "sn": "snare", "sd": "snare",
+              "cp": "snare", "bd": "kick", "cy": "hat"}
+_STRONG_CONTAINS = [("kick", "kick"), ("snare", "snare"), ("clap", "snare"),
+                    ("bass", "bass"), ("hat", "hat"), ("guitar", "keys"),
+                    ("piano", "keys"), ("rhodes", "keys"), ("vocal", "vox"),
+                    ("voice", "vox"), ("dialog", "vox")]
+
+
+def classify_sample_family(name: str):
+    """Canonical sample-name → family key (or None). The SSOT classifier every surface
+    should use (DRY). Token-aware: a name like `vec1_snare` or `nujazz_bass125` resolves
+    by its embedded instrument token, not just its prefix. Sample-world model, not a
+    measured-register claim. Returns the family `key` or None when genuinely unknown."""
+    s = name.lower()
+    toks = [t for t in re.split(r"[_\-0-9]+", s) if t]
+    for f in SAMPLE_FAMILIES:
+        for m in f.match:
+            if s == m or s.startswith(m) or m in toks or \
+               (len(m) >= 3 and any(t.startswith(m) for t in toks)):
+                return f.key
+    # drum-machine suffixes only on DM-style names (h2ogmhh, tr909sn) — not "shiloh"
+    if re.search(r"gm|dm|tr\d|rytm|\dmt", s) or re.search(r"\d(hh|oh|ch|sn|sd|cp|bd|cy)$", s):
+        for suf, fam in _DM_SUFFIX.items():
+            if s.endswith(suf):
+                return fam
+    for sub, fam in _STRONG_CONTAINS:              # gtkick, fbass, weird_dialogs, …
+        if sub in s:
+            return fam
+    return None
+
+
+# ── style label normalization (corner C metadata is hand-typed & fragmented) ──
+STYLE_ALIASES = {
+    "nu-jazz": "nujazz", "nujazz": "nujazz", "breakbeat": "breaks", "breaks": "breaks",
+    "chip": "chiptune", "chiptune": "chiptune", "hip-hop": "hiphop", "hip": "hiphop",
+    "acid-techno": "acid", "punk-dnb": "dnb", "lofi": "lofi", "lo-fi": "lofi",
+}
+
+
+def norm_style(s: str) -> str:
+    """Canonicalize a hand-typed style label. Unknown labels pass through lowercased."""
+    if not s:
+        return s
+    k = s.strip().lower()
+    return STYLE_ALIASES.get(k, k)
+
 # The whole vocabulary, for gen_tokens (and anyone who needs the fleet paint).
 ONTOLOGY = {
     "role": ROLE_FAMILIES,

armada/tide-table/tests/test_tokens.py

@@ -82,17 +82,32 @@ def test_tokens_validate_against_model():
 
 
 def test_sample_classifier_examples():
-    """The lexical sample-family rules should land obvious cases (coherent model)."""
-    fam = {f.key: f for f in M.SAMPLE_FAMILIES}
-    def classify(name):
-        s = name.lower()
-        for f in M.SAMPLE_FAMILIES:
-            if any(s == m or s.startswith(m) for m in f.match):
-                return f.key
-        return None
-    assert classify("kick") == "kick"
-    assert classify("808bd") == "kick"
-    assert classify("hats") == "hat"
-    assert classify("meth_bass") == "bass"
-    assert classify("jazz") == "break"
-    assert classify("superpiano") is None or classify("piano") == "keys"
+    """The canonical SSOT classifier lands obvious cases AND the token-aware ones."""
+    c = M.classify_sample_family
+    assert c("kick") == "kick"
+    assert c("808bd") == "kick"
+    assert c("hats") == "hat"
+    assert c("meth_bass") == "bass"
+    assert c("jazz") == "break"
+    assert c("piano") == "keys"
+    # token-aware / drum-machine / embedded (the coverage-improving cases)
+    assert c("h2ogmhh") == "hat"          # DM suffix, gated by 'gm'
+    assert c("vec1_snare") == "snare"     # embedded instrument token
+    assert c("nujazz_bass125") == "bass"  # token wins over 'nujazz'
+    assert c("nujazz_keys120") == "keys"
+    assert c("fbass") == "bass"
+    assert c("weird_dialogs") == "vox"
+    assert c("ho") == "hat"
+
+
+def test_classifier_does_not_overreach():
+    """Honest unknowns stay unknown — the 'shil-OH' false-positive must not return."""
+    assert M.classify_sample_family("shiloh") != "hat"
+    for unknown in ("armora", "90s_matrix"):
+        assert M.classify_sample_family(unknown) is None
+
+
+def test_style_normalization():
+    assert M.norm_style("nu-jazz") == M.norm_style("nujazz") == "nujazz"
+    assert M.norm_style("breakbeat") == M.norm_style("breaks") == "breaks"
+    assert M.norm_style("techno") == "techno"          # unknown passes through

armada/tide-table/tide_eda.py

@@ -80,13 +80,8 @@ def parse_tempo(src):
     return (fixed or cands)[0]
 
 
-# ── sample-family classifier (DRY: same cues the fleet color language uses) ────
-def sample_family(name):
-    s = name.lower()
-    for f in M.SAMPLE_FAMILIES:
-        if any(s == m or s.startswith(m) for m in f.match):
-            return f.key
-    return None
+# ── sample-family classifier (DRY: the canonical SSOT classifier in models.py) ─
+sample_family = M.classify_sample_family
 
 
 def gig_year_index():
@@ -110,7 +105,7 @@ def metadata_bpm():
         for t in (d.get("tracks") or []):
             rows.append({"gig": slug, "date": (d.get("date") or "")[:10],
                          "file": t.get("file"), "name": t.get("name"),
-                         "bpm": t.get("bpm"), "style": t.get("style")})
+                         "bpm": t.get("bpm"), "style": M.norm_style(t.get("style"))})
     return rows
 
 

armada/tide-table/tokens.json

@@ -199,6 +199,7 @@
     "sn",
     "sd",
     "clap",
+    "claps",
     "cp",
     "rim",
     "rs"
@@ -227,7 +228,8 @@
     "clave",
     "shaker",
     "tabla",
-    "cowbell"
+    "cowbell",
+    "drum"
    ]
   },
   {
@@ -248,6 +250,7 @@
    "match": [
     "hat",
     "hh",
+    "ho",
     "oh",
     "ch",
     "hihat",
@@ -278,7 +281,9 @@
     "loop",
     "jungle",
     "dnb",
-    "jazz"
+    "jazz",
+    "breaks165",
+    "fbreak"
    ]
   },
   {
@@ -302,7 +307,10 @@
     "choir",
     "string",
     "ambient",
-    "atmos"
+    "atmos",
+    "atm",
+    "airport",
+    "trance"
    ]
   },
   {
@@ -329,7 +337,14 @@
     "organ",
     "fpiano",
     "qstab",
-    "nujazz"
+    "cbow",
+    "cpluck",
+    "clav",
+    "marimba",
+    "mandolin",
+    "guitar",
+    "organ",
+    "forgan"
    ]
   },
   {
@@ -354,7 +369,12 @@
     "stab",
     "blip",
     "saw",
-    "square"
+    "square",
+    "brass",
+    "sax",
+    "horn",
+    "trump",
+    "tromb"
    ]
   },
   {
@@ -374,11 +394,16 @@
    ],
    "match": [
     "synth",
+    "syn",
     "commodore",
     "chip",
     "fm",
     "moog",
-    "reese"
+    "reese",
+    "giorgio",
+    "electro",
+    "cs80",
+    "vec"
    ]
   },
   {
@@ -404,7 +429,8 @@
     "wobble",
     "meth_bass",
     "ramplem",
-    "bassline"
+    "bassline",
+    "fbass"
    ]
   },
   {
@@ -425,6 +451,7 @@
    "match": [
     "fx",
     "riser",
+    "risers",
     "sweep",
     "noise",
     "impact",
@@ -456,7 +483,10 @@
     "voc",
     "acap",
     "speech",
-    "voice"
+    "voice",
+    "movie",
+    "dialog",
+    "dialogs"
    ]
   }
  ]

armada/ui/src/tokens.json

@@ -199,6 +199,7 @@
     "sn",
     "sd",
     "clap",
+    "claps",
     "cp",
     "rim",
     "rs"
@@ -227,7 +228,8 @@
     "clave",
     "shaker",
     "tabla",
-    "cowbell"
+    "cowbell",
+    "drum"
    ]
   },
   {
@@ -248,6 +250,7 @@
    "match": [
     "hat",
     "hh",
+    "ho",
     "oh",
     "ch",
     "hihat",
@@ -278,7 +281,9 @@
     "loop",
     "jungle",
     "dnb",
-    "jazz"
+    "jazz",
+    "breaks165",
+    "fbreak"
    ]
   },
   {
@@ -302,7 +307,10 @@
     "choir",
     "string",
     "ambient",
-    "atmos"
+    "atmos",
+    "atm",
+    "airport",
+    "trance"
    ]
   },
   {
@@ -329,7 +337,14 @@
     "organ",
     "fpiano",
     "qstab",
-    "nujazz"
+    "cbow",
+    "cpluck",
+    "clav",
+    "marimba",
+    "mandolin",
+    "guitar",
+    "organ",
+    "forgan"
    ]
   },
   {
@@ -354,7 +369,12 @@
     "stab",
     "blip",
     "saw",
-    "square"
+    "square",
+    "brass",
+    "sax",
+    "horn",
+    "trump",
+    "tromb"
    ]
   },
   {
@@ -374,11 +394,16 @@
    ],
    "match": [
     "synth",
+    "syn",
     "commodore",
     "chip",
     "fm",
     "moog",
-    "reese"
+    "reese",
+    "giorgio",
+    "electro",
+    "cs80",
+    "vec"
    ]
   },
   {
@@ -404,7 +429,8 @@
     "wobble",
     "meth_bass",
     "ramplem",
-    "bassline"
+    "bassline",
+    "fbass"
    ]
   },
   {
@@ -425,6 +451,7 @@
    "match": [
     "fx",
     "riser",
+    "risers",
     "sweep",
     "noise",
     "impact",
@@ -456,7 +483,10 @@
     "voc",
     "acap",
     "speech",
-    "voice"
+    "voice",
+    "movie",
+    "dialog",
+    "dialogs"
    ]
   }
  ]