Add LLM OCR engine via OpenCode Go + mimo-v2-omni

Adds optional LLM-based OCR as an alternative to Tesseract for reading
machine IDs from photos.

Backend (server.py):
- New run_ocr_llm() function calls OpenCode Go API (mimo-v2-omni model)
- Auto-falls back to Tesseract if API key missing or call fails
- Endpoints /api/analyze and /api/bulk-process accept ?ocr_engine=llm
  query param (default: tesseract) and ?ocr_model for model override
- Configurable via env vars: OPENCODE_GO_API_KEY, LLM_OCR_MODEL
- Requires User-Agent: Hermes-Agent/1.0 header for OpenCode Go API

Frontend (static/index.html):
- Toggle checkbox 'Use LLM OCR' in the UI
- OCR engine badge shown in results (llm vs tesseract + model name)
- getOcrParams() helper appends ?ocr_engine=llm to API calls

Infrastructure:
- .gitignore for uploads/ directory

Closes: #2

.gitignore

@@ -0,0 +1,3 @@
+uploads/
+__pycache__/
+*.py[cod]

server.py

@@ -1,8 +1,8 @@
 """EXIF + OCR test backend — validate that GPS survives upload pipeline."""
-import io, json, re, uuid, sqlite3
+import io, json, os, re, uuid, sqlite3, urllib.request
 from pathlib import Path
 
-from fastapi import FastAPI, File, Form, HTTPException, UploadFile
+from fastapi import FastAPI, File, Form, HTTPException, Query, UploadFile
 from fastapi.staticfiles import StaticFiles
 from PIL import Image as PILImage
 
@@ -20,6 +20,11 @@ try:
 except ImportError:
     HAS_TESSERACT = False
 
+# === LLM OCR via OpenCode Go ===
+OPENCODE_GO_KEY = os.environ.get("OPENCODE_GO_API_KEY", "")
+OPENCODE_GO_BASE = os.environ.get("OPENCODE_GO_BASE_URL", "https://opencode.ai/zen/go/v1")
+LLM_OCR_MODEL = os.environ.get("LLM_OCR_MODEL", "mimo-v2-omni")
+
 UPLOADS = Path(__file__).parent / "uploads"
 UPLOADS.mkdir(exist_ok=True)
 CANTEEN_DB = Path(__file__).parent.parent / "canteen-asset-tracker" / "assets.db"
@@ -114,6 +119,70 @@ def run_ocr(image_bytes: bytes) -> dict:
         tmp_path.unlink(missing_ok=True)
 
 
+def run_ocr_llm(image_bytes: bytes, model: str | None = None) -> dict:
+    """Run OCR via an LLM vision model on OpenCode Go.
+
+    Falls back to Tesseract if the API key is missing or the call fails.
+    Returns the same shape as run_ocr() with an additional ''engine'' field.
+    """
+    if not OPENCODE_GO_KEY:
+        result = run_ocr(image_bytes)
+        result["engine"] = "tesseract"
+        result["llm_fallback_reason"] = "no_api_key"
+        return result
+
+    import base64
+
+    model = model or LLM_OCR_MODEL
+    b64 = base64.b64encode(image_bytes).decode()
+
+    body = json.dumps({
+        "model": model,
+        "messages": [{
+            "role": "user",
+            "content": [
+                {"type": "text", "text": "Read ALL text and numbers visible in this photo. "
+                 "Return the exact text shown, nothing else."},
+                {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{b64}"}}
+            ]
+        }],
+        "max_tokens": 200,
+    }).encode()
+
+    req = urllib.request.Request(
+        f"{OPENCODE_GO_BASE}/chat/completions",
+        data=body,
+        headers={
+            "Authorization": f"Bearer {OPENCODE_GO_KEY}",
+            "Content-Type": "application/json",
+            "User-Agent": "Hermes-Agent/1.0",
+        },
+    )
+
+    try:
+        resp = urllib.request.urlopen(req, timeout=60)
+        result = json.loads(resp.read())
+        text = result["choices"][0]["message"]["content"].strip()
+    except Exception as exc:
+        # Fallback to Tesseract
+        result = run_ocr(image_bytes)
+        result["engine"] = "tesseract"
+        result["llm_fallback_reason"] = str(exc)[:200]
+        return result
+
+    match_5dash = re.search(r"(\d{5})[-\s]*(\d{6,})", text)
+    match_5plus = re.search(r"(\d{5,})", text)
+
+    return {
+        "available": True,
+        "engine": "llm",
+        "llm_model": model,
+        "raw_text": text[:500],
+        "match_5dash6": match_5dash.group(0) if match_5dash else None,
+        "match_5plus": match_5plus.group(0) if match_5plus else None,
+    }
+
+
 def lookup_machine_id(machine_id: str) -> dict | None:
     """Look up an asset by machine_id. Returns asset dict or None."""
     conn = _get_canteen_db()
@@ -150,8 +219,17 @@ def lookup_machine_id(machine_id: str) -> dict | None:
 
 
 @app.post("/api/analyze")
-async def analyze_photo(file: UploadFile = File(...)):
+async def analyze_photo(
-    """Upload a photo, get back EXIF + OCR results."""
+    file: UploadFile = File(...),
+    ocr_engine: str = Query(default="tesseract"),
+    ocr_model: str = Query(default=""),
+):
+    """Upload a photo, get back EXIF + OCR results.
+
+    Query params:
+      - ocr_engine: ''tesseract'' (default) or ''llm''
+      - ocr_model:  model name override (e.g. ''mimo-v2-omni'', ''glm-5.1'')
+    """
     contents = await file.read()
     file_size = len(contents)
 
@@ -162,7 +240,14 @@ async def analyze_photo(file: UploadFile = File(...)):
     (UPLOADS / fname).write_bytes(contents)
 
     exif_result = extract_exif(contents)
-    ocr_result = run_ocr(contents) if HAS_TESSERACT else {"available": False, "text": ""}
+
+    if ocr_engine == "llm":
+        ocr_result = run_ocr_llm(contents, ocr_model or None)
+    elif HAS_TESSERACT:
+        ocr_result = run_ocr(contents)
+        ocr_result["engine"] = "tesseract"
+    else:
+        ocr_result = {"available": False, "text": "", "engine": "none"}
 
     machine_id = None
     asset = None
@@ -183,9 +268,17 @@ async def analyze_photo(file: UploadFile = File(...)):
 
 
 @app.post("/api/bulk-process")
-async def bulk_process(files: list[UploadFile] = File(...)):
+async def bulk_process(
+    files: list[UploadFile] = File(...),
+    ocr_engine: str = Query(default="tesseract"),
+    ocr_model: str = Query(default=""),
+):
     """Process multiple photos: OCR each, extract EXIF GPS, look up matching assets.
 
+    Query params:
+      - ocr_engine: ''tesseract'' (default) or ''llm''
+      - ocr_model:  model name override
+
     Returns a list of results, each with:
       - filename, exif (gps), ocr match, matched asset (if found)
       - needs_gps: true if asset exists AND has no coordinates AND photo has GPS
@@ -208,7 +301,14 @@ async def bulk_process(files: list[UploadFile] = File(...)):
         (UPLOADS / fname).write_bytes(contents)
 
         exif_result = extract_exif(contents)
-        ocr_result = run_ocr(contents) if HAS_TESSERACT else {"available": False, "text": ""}
+
+        if ocr_engine == "llm":
+            ocr_result = run_ocr_llm(contents, ocr_model or None)
+        elif HAS_TESSERACT:
+            ocr_result = run_ocr(contents)
+            ocr_result["engine"] = "tesseract"
+        else:
+            ocr_result = {"available": False, "text": "", "engine": "none"}
 
         has_gps = exif_result.get("gps") is not None
         if has_gps:

static/index.html

@@ -187,6 +187,20 @@
     margin-top: 6px;
   }
 
+  /* OCR engine toggle */
+  .ocr-toggle {
+    display: flex; align-items: center; gap: 8px; margin-bottom: 8px;
+    font-size: 12px; color: var(--text2);
+  }
+  .ocr-toggle label { display: flex; align-items: center; gap: 4px; cursor: pointer; }
+  .ocr-toggle input[type="checkbox"] { accent-color: var(--accent); }
+  .engine-badge {
+    font-size: 10px; padding: 2px 8px; border-radius: 10px; font-weight: 600;
+    background: var(--card2); color: var(--text2); display: inline-block; margin-left: 4px;
+  }
+  .engine-badge.llm { background: var(--accent); color: #fff; }
+  .engine-badge.tesseract { background: var(--card2); color: var(--text2); }
+
   /* Bulk results */
   #mapContainer { height: 250px; border-radius: var(--radius-sm); margin-bottom: 10px; display: none; }
   .bulk-card {
@@ -263,6 +277,14 @@
   <div id="serverResults"></div>
 </div>
 
+<!-- OCR engine toggle -->
+<div class="ocr-toggle" style="margin-bottom:4px;">
+  <label>
+    <input type="checkbox" id="ocrLlmToggle" onchange="onOcrToggle()">
+    🧠 Use LLM OCR (<code id="ocrModelLabel">mimo-v2-omni</code>)
+  </label>
+</div>
+
 <!-- Bulk process button -->
 <button class="btn btn-primary" id="bulkBtn" style="display:none;margin-top:12px;" onclick="startBulkProcess()">
   🔍 Bulk Process GPS Photos
@@ -440,7 +462,7 @@ async function uploadSelected() {
   fd.append('file', file, file.name || 'photo.jpg');
 
   try {
-    const resp = await fetch('/api/analyze', { method: 'POST', body: fd });
+    const resp = await fetch('/api/analyze' + getOcrParams(), { method: 'POST', body: fd });
     const data = await resp.json();
 
     let html = '';
@@ -465,21 +487,20 @@ async function uploadSelected() {
 
     // OCR
     const ocr = data.ocr;
-    if (ocr.available) {
+    const engineCls = ocr.engine === 'llm' ? 'llm' : 'tesseract';
-      html += '<div style="margin-top:10px;font-weight:600;">🔤 OCR:</div>';
+    html += '<div style="margin-top:10px;font-weight:600;">🔤 OCR <span class="engine-badge ' + engineCls + '">' + esc(ocr.engine || 'tesseract') + (ocr.llm_model ? ' ' + ocr.llm_model : '') + '</span></div>';
-      if (ocr.raw_text) {
+    if (ocr.raw_text) {
-        html += '<div class="ocr-text">' + esc(ocr.raw_text) + '</div>';
+      html += '<div class="ocr-text">' + esc(ocr.raw_text) + '</div>';
-      }
+    }
-      if (ocr.match_5dash6) {
+    if (ocr.match_5dash6) {
-        const mid = ocr.match_5dash6.replace(/[^0-9]/g,'').slice(-5);
+      const mid = ocr.match_5dash6.replace(/[^0-9]/g,'').slice(-5);
-        html += '<div style="margin-top:4px;color:var(--green);">✅ Matched: <strong>' + esc(ocr.match_5dash6) + '</strong> → machine ID: ' + mid + '</div>';
+      html += '<div style="margin-top:4px;color:var(--green);">✅ Matched: <strong>' + esc(ocr.match_5dash6) + '</strong> → machine ID: ' + mid + '</div>';
-        // Auto-lookup
+      // Auto-lookup
-        lookupAsset(mid);
+      lookupAsset(mid);
-      } else if (ocr.match_5plus) {
+    } else if (ocr.match_5plus) {
-        html += '<div style="margin-top:4px;color:var(--amber);">⚠️ Digits: <strong>' + esc(ocr.match_5plus) + '</strong> (no 5-6 pattern)</div>';
+      html += '<div style="margin-top:4px;color:var(--amber);">⚠️ Digits: <strong>' + esc(ocr.match_5plus) + '</strong> (no 5-6 pattern)</div>';
-      } else {
+    } else {
-        html += '<div style="margin-top:4px;color:var(--text3);">No machine ID found in image</div>';
+      html += '<div style="margin-top:4px;color:var(--text3);">No machine ID found in image</div>';
-      }
     }
 
     div.innerHTML = html;
@@ -540,7 +561,7 @@ async function startBulkProcess() {
   gpsPhotos.forEach(p => fd.append('files', p.file, p.file.name || 'photo.jpg'));
 
   try {
-    const resp = await fetch('/api/bulk-process', { method: 'POST', body: fd });
+    const resp = await fetch('/api/bulk-process' + getOcrParams(), { method: 'POST', body: fd });
     const data = await resp.json();
 
     // Merge thumbnails back into results
@@ -753,8 +774,13 @@ function esc(s) {
   return d.innerHTML;
 }
 
+function getOcrParams() {
+  const useLlm = document.getElementById('ocrLlmToggle').checked;
+  if (!useLlm) return '';
+  return '?ocr_engine=llm';
+}
+
 function formatSize(bytes) {
-  if (bytes < 1024) return bytes + ' B';
   if (bytes < 1048576) return (bytes / 1024).toFixed(1) + ' KB';
   return (bytes / 1048576).toFixed(1) + ' MB';
 }