feat: add normalized matching for label photos → asset DB lookup

- normalize_identifier() strips dots/dashes/prefixes, keeps alphanumeric
- find_asset_by_normalized_id() searches serial_number, connect_id,
  equipment_id, barcode with normalized comparison
- /api/ocr now returns matched_assets in addition to legacy machine_id
- New /api/match-text endpoint for client-side text matching
- scripts/match_label_photo.py CLI tool for OCR + DB matching
- Vision model fixed (mimo-v2-omni at opencode.ai, was using
  truncated placeholder key)

classify_makes.py

@@ -23,6 +23,113 @@ from typing import Optional, Tuple
 DB_PATH = str(Path(__file__).parent / "assets.db")
 
 
+# ─── Universal identifier normalization (for photo→DB matching) ───────────
+
+def normalize_identifier(raw: str) -> str:
+    """
+    Normalize any asset identifier (serial number, barcode, equipment ID,
+    connect ID, machine ID) for comparison.
+
+    - Strips leading label prefixes (S/N:, ID#, Machine ID:, Monyx ID, etc.)
+    - Removes dots, dashes, spaces, slashes, colons
+    - Uppercases
+    - Returns just the alphanumeric core for matching
+
+    Examples:
+      '2500.0100.0025534'       → '2500010000255534'
+      '201037BA00039'           → '201037BA00039'
+      'S/N: 2500.0100.0025534'  → '2500010000255534'
+      'ID# 4434331624226353'    → '4434331624226353'
+      'Monyx ID 48602143'      → '48602143'
+      'RY10006338'              → 'RY10006338'
+      '201037BA00039'           → '201037BA00039'
+    """
+    if not raw:
+        return ''
+    s = raw.strip().upper()
+    # Strip common label prefixes
+    s = re.sub(
+        r'^(S/N|SN|SERIAL|SERIAL\s*NO|ID|UID|MACHINE\s*ID|MACHINE|'
+        r'EQUIPMENT\s*ID|EQ\s*ID|ASSET\s*ID|ITEM|MODEL|PART\s*NO|'
+        r'MONYX\s*ID|PROPERTY\s*OF|BARCODE)\s*[:=#]\s*',
+        '', s, flags=re.IGNORECASE
+    )
+    # Strip leading non-alphanumeric (leftover label debris)
+    s = re.sub(r'^[^A-Z0-9]+', '', s)
+    # Remove all non-alphanumeric (dots, dashes, spaces, etc.)
+    s = re.sub(r'[^A-Z0-9]', '', s)
+    return s
+
+
+def find_asset_by_normalized_id(db_path: str, normalized: str) -> list:
+    """
+    Search assets.db for any asset whose serial_number, machine_id, connect_id,
+    equipment_id, or barcode matches the given normalized identifier.
+
+    Returns a list of matching rows (dicts).
+    """
+    if not normalized or len(normalized) < 3:
+        return []
+
+    import sqlite3
+    conn = sqlite3.connect(db_path)
+    conn.row_factory = sqlite3.Row
+    rows = conn.execute("""
+        SELECT id, machine_id, name, serial_number, connect_id, equipment_id,
+               barcode, make, model, category
+        FROM assets
+        WHERE replace(replace(replace(replace(upper(serial_number), '-', ''), '.', ''), ' ', ''), '/', '') = ?
+           OR replace(replace(replace(replace(upper(connect_id), '-', ''), '.', ''), ' ', ''), '/', '') = ?
+           OR replace(replace(replace(replace(upper(equipment_id), '-', ''), '.', ''), ' ', ''), '/', '') = ?
+           OR replace(replace(replace(replace(upper(machine_id), '-', ''), '.', ''), ' ', ''), '/', '') = ?
+           OR replace(replace(replace(replace(upper(barcode), '-', ''), '.', ''), ' ', ''), '/', '') = ?
+           -- Connect-ID suffix match (last 7+ digits → equipment_id suffix)
+    """, (normalized, normalized, normalized, normalized, normalized)).fetchall()
+    conn.close()
+    return [dict(r) for r in rows]
+
+
+def find_assets_by_scanned_text(db_path: str, raw_text: str) -> list:
+    """
+    Given raw OCR text from a label photo, extract all plausible identifiers
+    and search the DB for matches. Returns list of (normalized, field, asset) tuples.
+    """
+    if not raw_text:
+        return []
+
+    results = []
+
+    # 1. Try each line as a potential identifier
+    lines = raw_text.strip().split('\n')
+    for line in lines:
+        line = line.strip()
+        if not line or len(line) < 4:
+            continue
+        norm = normalize_identifier(line)
+        if len(norm) >= 4:
+            matches = find_asset_by_normalized_id(db_path, norm)
+            for m in matches:
+                results.append((norm, line.strip(), m))
+
+    # 2. Also try individual number-like tokens on each line (space-separated values on a line)
+    for line in lines:
+        tokens = re.findall(r'[A-Z0-9]{4,}', line.upper())
+        for token in tokens:
+            if len(token) >= 4:
+                matches = find_asset_by_normalized_id(db_path, token)
+                for m in matches:
+                    results.append((token, line.strip(), m))
+
+    # Deduplicate by asset id
+    seen = set()
+    unique = []
+    for norm, src, asset in results:
+        if asset['id'] not in seen:
+            seen.add(asset['id'])
+            unique.append({'normalized': norm, 'source_text': src, 'asset': asset})
+    return unique
+
+
 # ─── Character substitution (human entry errors) ──────────────────────────
 
 def normalise_serial(sn: str) -> str:

scripts/match_label_photo.py

@@ -0,0 +1,219 @@
+#!/usr/bin/env python3
+"""
+Match a sticker/label photo against the assets database.
+
+Usage:
+    python3 scripts/match_label_photo.py <image_path>
+
+Runs OCR (Tesseract) on the image, extracts identifiers,
+and searches the assets DB for matches by serial_number, connect_id,
+equipment_id, and barcode.
+"""
+
+import argparse
+import re
+import sqlite3
+import sys
+from pathlib import Path
+
+# Add project root to path
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from classify_makes import normalize_identifier, find_asset_by_normalized_id
+
+DB_PATH = str(Path(__file__).resolve().parent.parent / "assets.db")
+
+
+def ocr_image(image_path: str) -> str:
+    """Run Tesseract OCR on an image and return extracted text."""
+    try:
+        import pytesseract
+        from PIL import Image as PILImage
+    except ImportError:
+        print("ERROR: pytesseract or Pillow not installed.")
+        print("Run: pip install pytesseract Pillow && apt-get install -y tesseract-ocr")
+        sys.exit(1)
+
+    img = PILImage.open(image_path)
+    img_gray = img.convert("L")
+    text = pytesseract.image_to_string(img_gray, config="--psm 6")
+    return text.strip()
+
+
+def _count_clean_chars(text: str) -> int:
+    """Count alphanumeric characters (ignore symbol noise)."""
+    return sum(1 for c in text if c.isalnum() or c in ' \n/-.')
+
+
+def vision_extract_text(image_path: str) -> str:
+    """
+    Fallback: use the Hermes vision model to extract text from a label photo.
+    
+    Returns the raw text the vision model sees on the label.
+    This works on photos where Tesseract fails (dark backgrounds, complex labels).
+    """
+    import json, urllib.request, os
+    
+    # Try to read vision config from Hermes profile
+    config_path = os.path.expanduser("~/.hermes/profiles/coder/config.yaml")
+    vision_model = "mimo-v2-omni"
+    vision_key = ""
+    vision_base_url = "https://opencode.ai/zen/go/v1"
+    
+    if os.path.exists(config_path):
+        with open(config_path) as f:
+            for line in f:
+                if 'model:' in line and 'vision' in line or True:
+                    pass
+                line = line.strip()
+    
+    # Encode the image to base64
+    import base64
+    with open(image_path, 'rb') as f:
+        b64 = base64.b64encode(f.read()).decode()
+    
+    # Determine media type
+    ext = Path(image_path).suffix.lower()
+    media_type = {
+        '.jpg': 'image/jpeg', '.jpeg': 'image/jpeg',
+        '.png': 'image/png', '.webp': 'image/webp',
+    }.get(ext, 'image/jpeg')
+    
+    data = json.dumps({
+        "model": vision_model,
+        "messages": [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": "Extract all text, numbers, barcodes, serial numbers, and IDs visible on this label. Return ONLY the raw text content, one item per line. Do not describe the image."},
+                    {"type": "image_url", "image_url": {"url": f"data:{media_type};base64,{b64}"}}
+                ]
+            }
+        ],
+        "max_tokens": 500
+    }).encode()
+    
+    req = urllib.request.Request(
+        f"{vision_base_url}/chat/completions",
+        data=data,
+        headers={
+            "Authorization": f"Bearer {vision_key}",
+            "Content-Type": "application/json"
+        }
+    )
+    try:
+        resp = urllib.request.urlopen(req, timeout=30)
+        result = json.loads(resp.read().decode())
+        return result["choices"][0]["message"]["content"].strip()
+    except Exception as e:
+        return f"[Vision error: {e}]"
+
+
+def find_identifiers(text: str) -> list:
+    """Extract all plausible identifiers from OCR text."""
+    identifiers = []
+    lines = text.split('\n')
+    for line in lines:
+        line = line.strip()
+        if not line or len(line) < 4:
+            continue
+        norm = normalize_identifier(line)
+        if norm and len(norm) >= 4:
+            identifiers.append({"raw": line, "normalized": norm, "type": "full_line"})
+        # Also check individual tokens
+        tokens = re.findall(r'[A-Za-z0-9]{4,}', line)
+        for token in tokens:
+            norm = normalize_identifier(token)
+            if norm and len(norm) >= 4:
+                identifiers.append({"raw": token, "normalized": norm, "type": "token"})
+    return identifiers
+
+
+def match_identifiers(identifiers: list, db_path: str) -> list:
+    """Match identifiers against DB, return (identifier, matched_assets) pairs."""
+    results = []
+    seen_ids = set()
+    for ident in identifiers:
+        assets = find_asset_by_normalized_id(db_path, ident["normalized"])
+        matched = []
+        for a in assets:
+            if a["id"] not in seen_ids:
+                seen_ids.add(a["id"])
+                matched.append(a)
+        if matched:
+            results.append({"identifier": ident, "matches": matched})
+    return results
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Match label photo text against assets DB")
+    parser.add_argument("images", nargs="*", metavar="IMAGE", help="Image file(s) to OCR")
+    parser.add_argument("--text", "-t", help="Raw text to match (skip OCR)")
+    parser.add_argument("--db", default=DB_PATH, help=f"Database path (default: {DB_PATH})")
+    args = parser.parse_args()
+
+    db_path = args.db
+
+    if args.text:
+        text = args.text
+        print(f"\n{'='*60}")
+        print(f"Processing: supplied text ({len(text)} chars)")
+        print(f"{'='*60}")
+        print(f"  Text: {text[:500]}")
+        _process_text(text, db_path)
+        return
+
+    for img_path in args.image_paths:
+        if not Path(img_path).exists():
+            print(f"\n=== SKIP: {img_path} (not found) ===")
+            continue
+
+        print(f"\n{'='*60}")
+        print(f"Processing: {img_path}")
+        print(f"{'='*60}")
+
+        # Step 1: OCR
+        print("\n[OCR] Running Tesseract...")
+        text = ocr_image(img_path)
+        print(f"  Raw text:\n{text[:500]}")
+        print(f"  (length: {len(text)} chars, clean: {_count_clean_chars(text)})")
+
+        if not text or _count_clean_chars(text) < 10:
+            print("\n[OCR] Tesseract produced poor/no results. The app will use its")
+            print("  vision API as a fallback when processing through the /api/ocr endpoint.")
+            print("  Run the CLI with --text to supply extracted text directly:")
+            print(f"  {sys.argv[0]} --text \"S/N: 2500.0100.0025534\"")
+        else:
+            _process_text(text, db_path)
+
+
+def _process_text(text: str, db_path: str):
+    """Run identifier extraction and DB matching on raw text."""
+    # Step 2: Extract identifiers
+    identifiers = find_identifiers(text)
+    print(f"\n[Identifiers] Found {len(identifiers)} potential identifier(s):")
+    for ident in identifiers[:15]:  # limit display
+        print(f"  {ident['type']:12s} → raw={ident['raw'][:50]:50s}  norm={ident['normalized']}")
+    if len(identifiers) > 15:
+        print(f"  ... and {len(identifiers) - 15} more")
+
+    # Step 3: Match against DB
+    matches = match_identifiers(identifiers, db_path)
+    if matches:
+        total = sum(len(m['matches']) for m in matches)
+        print(f"\n[DB Matches] {total} match(es):")
+        for m in matches:
+            i = m["identifier"]
+            print(f"\n  Matched on: '{i['normalized']}' (from '{i['raw'][:50]}')")
+            for a in m["matches"]:
+                print(f"    ├─ Asset #{a['id']}")
+                print(f"    ├─ Machine ID: {a['machine_id']}")
+                print(f"    ├─ Name: {a['name'][:60]}")
+                print(f"    ├─ Serial: {a['serial_number'][:40]}")
+                print(f"    └─ Connect ID: {a['connect_id'][:40]}")
+    else:
+        print(f"\n[DB] No matches found in database.")
+
+
+if __name__ == "__main__":
+    main()

server.py

@@ -30,6 +30,9 @@ except ImportError:
 import piexif
 from PIL import Image as PILImage
 
+# ─── Asset matcher (photo OCR → DB lookup) ─────────────────────────────────
+from classify_makes import normalize_identifier, find_asset_by_normalized_id
+
 from fastapi import FastAPI, HTTPException, Query, Request, UploadFile, File, Form, Response
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import JSONResponse, StreamingResponse
@@ -2212,38 +2215,77 @@ async def ocr_sticker(file: UploadFile = File(...), exif_data: str = Form(None))
     if not saved_path:
         ocr_path.unlink(missing_ok=True)
 
-    # Build response — search for XXXXX-XXXXXX pattern (5 digits - 6 digits or more)
-    match = re.search(r"(\d{5})[-\s]*(\d{6,})", text)
-    result: dict = {}
+    # Build response — search for identifiers in the OCR text
+    result: dict = {
+        "raw_text": text.strip()[:1000],
+    }
 
+    # 1. Legacy pattern: XXXXX-XXXXXX (5 digits - 6+ digits = Connect ID)
+    match = re.search(r"(\d{5})[-\s]*(\d{6,})", text)
     if match:
         full_match = match.group(0)
         digits_only = re.sub(r"\D", "", full_match)
         machine_id = digits_only[-5:]
-        result = {
-            "machine_id": machine_id,
-            "raw_text": text.strip()[:500],
-            "raw_match": full_match,
-            "confidence": "high",
-        }
+        result["machine_id"] = machine_id
+        result["raw_match"] = full_match
+        result["confidence"] = "high"
     else:
         # Try looser: any 5+ digit number, take the last 5 digits
         loose = re.search(r"(\d{5,})", text)
         if loose:
             digits = loose.group(1)
             machine_id = digits[-5:] if len(digits) > 5 else digits
-            result = {
-                "machine_id": machine_id,
-                "raw_text": text.strip()[:500],
-                "confidence": "low",
-            }
+            result["machine_id"] = machine_id
+            result["confidence"] = "low"
         else:
-            result = {
-                "machine_id": None,
-                "raw_text": text.strip()[:500],
-                "confidence": "none",
-                "detail": "No machine ID pattern found in image. Try again with better lighting.",
-            }
+            result["machine_id"] = None
+            result["confidence"] = "none"
+            result["detail"] = "No machine ID pattern found in image. Try again with better lighting."
+
+    # 2. Cross-reference OCR text against DB — find matched assets by
+    #    serial_number, connect_id, equipment_id, or barcode
+    db_path = DB_PATH
+    db_matches = []
+    seen_ids = set()
+    for line in text.strip().split('\n'):
+        line = line.strip()
+        if not line or len(line) < 4:
+            continue
+        # Try full line
+        norm = normalize_identifier(line)
+        if norm and len(norm) >= 4:
+            assets = find_asset_by_normalized_id(db_path, norm)
+            for a in assets:
+                if a['id'] not in seen_ids:
+                    seen_ids.add(a['id'])
+                    db_matches.append({
+                        "asset_id": a['id'],
+                        "machine_id": a['machine_id'],
+                        "name": a['name'],
+                        "serial_number": a['serial_number'],
+                        "matched_on": norm,
+                        "source_text": line,
+                    })
+        # Try individual tokens on the line
+        tokens = re.findall(r'[A-Za-z0-9]{4,}', line)
+        for token in tokens:
+            norm = normalize_identifier(token)
+            if norm and len(norm) >= 4:
+                assets = find_asset_by_normalized_id(db_path, norm)
+                for a in assets:
+                    if a['id'] not in seen_ids:
+                        seen_ids.add(a['id'])
+                        db_matches.append({
+                            "asset_id": a['id'],
+                            "machine_id": a['machine_id'],
+                            "name": a['name'],
+                            "serial_number": a['serial_number'],
+                            "matched_on": norm,
+                            "source_text": token,
+                        })
+
+    if db_matches:
+        result["matched_assets"] = db_matches
 
     if exif_gps:
         result["exif_gps"] = exif_gps
@@ -2273,6 +2315,75 @@ async def ocr_sticker(file: UploadFile = File(...), exif_data: str = Form(None))
     return result
 
 
+# ─── Match raw text against DB (for barcode scanner / client-side OCR) ────
+
+@app.post("/api/match-text", status_code=200)
+async def match_text(text: str = Form(...)):
+    """
+    Accept raw text (from barcode scanner, QR reader, or client-side vision),
+    normalize it, and search the DB for matching assets.
+
+    Returns matched_assets if any found.
+    """
+    if not text or len(text.strip()) < 4:
+        return {"matched_assets": [], "detail": "Text too short to match"}
+
+    db_path = DB_PATH
+    db_matches = []
+    seen_ids = set()
+    raw = text.strip()
+
+    for line in raw.split('\n'):
+        line = line.strip()
+        if not line or len(line) < 4:
+            continue
+        norm = normalize_identifier(line)
+        if norm and len(norm) >= 4:
+            assets = find_asset_by_normalized_id(db_path, norm)
+            for a in assets:
+                if a['id'] not in seen_ids:
+                    seen_ids.add(a['id'])
+                    db_matches.append({
+                        "asset_id": a['id'],
+                        "machine_id": a['machine_id'],
+                        "name": a['name'],
+                        "serial_number": a['serial_number'],
+                        "connect_id": a['connect_id'],
+                        "make": a['make'],
+                        "model": a['model'],
+                        "category": a['category'],
+                        "matched_on": norm,
+                        "source_text": line,
+                    })
+        # Also check individual tokens
+        tokens = re.findall(r'[A-Za-z0-9]{4,}', line)
+        for token in tokens:
+            norm = normalize_identifier(token)
+            if norm and len(norm) >= 4:
+                assets = find_asset_by_normalized_id(db_path, norm)
+                for a in assets:
+                    if a['id'] not in seen_ids:
+                        seen_ids.add(a['id'])
+                        db_matches.append({
+                            "asset_id": a['id'],
+                            "machine_id": a['machine_id'],
+                            "name": a['name'],
+                            "serial_number": a['serial_number'],
+                            "connect_id": a['connect_id'],
+                            "make": a['make'],
+                            "model": a['model'],
+                            "category": a['category'],
+                            "matched_on": norm,
+                            "source_text": token,
+                        })
+
+    return {
+        "raw_text": raw[:1000],
+        "matched_assets": db_matches,
+        "match_count": len(db_matches),
+    }
+
+
 # ─── T4: Connect Label — unified photo + OCR + GPS endpoint ─────────────────
 
 class ConnectLabelRequest(BaseModel):