fix: recover table-style nutrition evidence

Codex-authored nutrition-table parsing, evidence diagnostics, and conservative OCR near-matching.

frontend/app.js

@@ -115,6 +115,7 @@ function render(data) {
     ["Whole-packet sodium", packet.sodium_mg == null ? "Not calculable" : `${packet.sodium_mg}mg`],
   ];
   $("#nutrition-grid").innerHTML = facts.map(([key, value]) => `<div><span>${key}</span><b>${value}</b></div>`).join("");
+  $("#nutrition-explanation").textContent = packet.explanation;
   $("#expiry-status").textContent = data.expiry.status;
   $("#opening-status").textContent = data.expiry.after_opening_instruction
     ? `After opening: ${data.expiry.after_opening_instruction}`

frontend/index.html

@@ -113,7 +113,7 @@
       </section>
       <div class="claim-grid" id="claim-grid"></div>
       <div class="evidence-summary">
-        <article><p class="kicker">NUTRITION EVIDENCE</p><div id="nutrition-grid"></div></article>
+        <article><p class="kicker">NUTRITION EVIDENCE</p><div id="nutrition-grid"></div><p class="nutrition-explanation" id="nutrition-explanation"></p></article>
         <article class="date-card"><p class="kicker">DATE EVIDENCE</p><h3 id="expiry-status"></h3><p id="opening-status"></p><p>Expiry interpretation is evidence, not a food-safety guarantee.</p></article>
       </div>
       <details><summary>View machine-readable evidence case</summary><pre id="raw-json"></pre></details>

frontend/styles.css

@@ -19,6 +19,7 @@ main{max-width:1320px;margin:auto;padding:0 4vw}.hero{min-height:670px;display:g
 .gap-section{margin-bottom:28px;padding:26px;border:1px solid var(--ink);background:#1b1b17;color:var(--cream);border-radius:18px}.gap-heading{display:flex;justify-content:space-between;gap:20px;align-items:end;margin-bottom:18px}.gap-heading .kicker{color:#bdb4a6}.gap-heading h3{font:700 clamp(27px,4vw,46px)/1 Georgia,serif;max-width:700px;margin:0}.gap-grid{display:grid;grid-template-columns:repeat(2,1fr);gap:12px}.gap-card,.gap-empty{padding:19px;border:1px solid #4a483f;border-radius:13px;background:#26251f}.gap-card.high{border-color:var(--red)}.gap-card.medium{border-color:var(--amber)}.gap-severity{font:500 8px ui-monospace,SFMono-Regular,Menlo,monospace;letter-spacing:.13em;text-transform:uppercase;color:#d9a65e}.gap-card h4{font-size:20px;margin:10px 0 16px}.gap-compare{display:grid;grid-template-columns:1fr 1fr;gap:10px}.gap-compare p{margin:0;padding:11px;background:#313029;border-radius:8px;font-size:12px;line-height:1.45}.gap-compare b{display:block;font:500 8px ui-monospace,SFMono-Regular,Menlo,monospace;color:#bdb4a6;margin-bottom:5px}.gap-card .evidence{border-top-color:#4a483f}.gap-empty{display:grid;gap:5px;color:#bdb4a6}
 .claim-grid{grid-template-columns:repeat(2,1fr)}.claim-card{background:var(--cream);border:1px solid var(--line);border-top:6px solid var(--muted);border-radius:16px;padding:23px}.claim-card.supported{border-top-color:var(--green)}.claim-card.contradicted{border-top-color:var(--red)}.claim-card.context{border-top-color:var(--amber)}.claim-top{display:flex;justify-content:space-between;gap:10px;align-items:start}.claim-name{font-size:21px;font-weight:800}.verdict{font:500 8px/1.3 ui-monospace,SFMono-Regular,Menlo,monospace;letter-spacing:.08em;border:1px solid var(--line);border-radius:99px;padding:7px 9px;text-align:right}.confidence{display:block;margin-top:8px;font:500 8px ui-monospace,SFMono-Regular,Menlo,monospace;letter-spacing:.1em;text-transform:uppercase;color:var(--muted)}.summary{min-height:45px;color:#534d43;line-height:1.5}.evidence{padding:11px 0;border-top:1px solid var(--line)}.evidence b,.evidence span{display:block}.evidence b{font:500 8px ui-monospace,SFMono-Regular,Menlo,monospace;letter-spacing:.12em;color:var(--muted);text-transform:uppercase}.evidence span{font-size:13px;margin-top:4px}.caveat{font-size:11px;color:var(--muted);margin-top:15px}
 .evidence-summary{margin-top:16px}.evidence-summary article{padding:25px;border:1px solid var(--line);border-radius:16px;background:var(--cream)}#nutrition-grid div{display:flex;justify-content:space-between;padding:10px 0;border-bottom:1px solid var(--line);font-size:13px}.date-card{background:var(--ink)!important;color:var(--cream)}.date-card .kicker{color:#cfc5b6}.date-card h3{font:700 28px/1.15 "Playfair Display"}details{margin-top:16px;border:1px solid var(--line);border-radius:14px;padding:17px;background:var(--cream)}summary{cursor:pointer;font-weight:700}pre{white-space:pre-wrap;font:11px/1.5 "DM Mono";overflow:auto}
+.nutrition-explanation{font-size:11px;line-height:1.5;color:var(--muted);margin:14px 0 0}
 .method{border-top:1px solid var(--line)}.method-grid{grid-template-columns:repeat(4,1fr);margin-top:40px}.method-grid div{padding:20px;border-top:2px solid var(--ink)}.method-grid span{font:500 10px "DM Mono";color:var(--red)}.method-grid p{font-size:13px;line-height:1.5;color:var(--muted)}
 .feedback-agent{display:grid;grid-template-columns:.8fr 1.2fr;gap:28px;margin-top:18px;padding:26px;border:1px solid var(--line);border-radius:18px;background:var(--cream)}.feedback-agent h3{font:700 clamp(25px,4vw,42px)/1 Georgia,serif;margin:0}.feedback-agent p{font-size:12px;line-height:1.55;color:var(--muted)}.feedback-controls{display:grid;gap:10px}.feedback-choice{display:flex;gap:8px;flex-wrap:wrap}.feedback-choice button.active{background:var(--red);border-color:var(--red);color:white}.feedback-controls textarea{min-height:95px;padding:13px;border:1px solid var(--line);border-radius:11px;background:#f8f3e9;resize:vertical}.feedback-controls>span{font:500 9px/1.5 ui-monospace,SFMono-Regular,Menlo,monospace;color:var(--green)}
 footer{display:flex;justify-content:space-between;gap:20px;padding:25px 5vw;border-top:1px solid var(--line);font:500 10px "DM Mono";color:var(--muted)}

nemotron_space/app.py

@@ -19,6 +19,8 @@ compact JSON with these keys:
 - priority: one short sentence naming the most important next action
 - evidence_request: one short sentence, or an empty string
 - rationale: one short sentence grounded only in the supplied investigation
+Prioritize missing evidence required to resolve front-of-pack claims. Treat
+expiry evidence as secondary unless expiry or shelf life is itself a front claim.
 """
 
 

src/packetcourt/audit.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import re
+from difflib import SequenceMatcher
 
 from .models import ClaimAudit, Evidence, PacketAudit, PersuasionFinding, Verdict
 from .investigator import build_investigation
@@ -221,6 +222,26 @@ def _audit_claim(claim: str, back_text: str, ingredients: list[str], nutrition)
                 confidence="high",
             )
 
+    enrichment_terms = [
+        term
+        for term in ("calcium", "dha", "iron", "vitamin", "mineral", "zinc")
+        if term in claim.lower()
+    ]
+    enrichment_matches = [
+        match.group(0)
+        for term in enrichment_terms
+        for match in re.finditer(rf"\b{re.escape(term)}\b[^,.;\n]{{0,45}}", back_text, re.IGNORECASE)
+    ]
+    if enrichment_matches:
+        return ClaimAudit(
+            claim=claim,
+            verdict=Verdict.CONTEXT_MISSING,
+            summary="Related enrichment evidence is visible, but the supplied label does not establish the full front-of-pack impression.",
+            evidence=[Evidence(source="back label", text=text.strip()) for text in dict.fromkeys(enrichment_matches)],
+            caveat="A quantified nutrient declaration is required to assess the strength of an enrichment claim.",
+            confidence="medium",
+        )
+
     meaningful_terms = [
         token.lower()
         for token in re.findall(r"[A-Za-z]{3,}", claim)
@@ -229,6 +250,16 @@ def _audit_claim(claim: str, back_text: str, ingredients: list[str], nutrition)
     evidence_matches = [
         item for item in ingredients if any(term in item.lower() for term in meaningful_terms)
     ]
+    fuzzy_matches = [
+        item
+        for item in ingredients
+        if item not in evidence_matches
+        and any(
+            SequenceMatcher(None, term, token).ratio() >= 0.8
+            for term in meaningful_terms
+            for token in re.findall(r"[a-z]{4,}", item.lower())
+        )
+    ]
     if evidence_matches:
         return ClaimAudit(
             claim=claim,
@@ -238,6 +269,15 @@ def _audit_claim(claim: str, back_text: str, ingredients: list[str], nutrition)
             caveat="PacketCourt will not infer quantity, quality, or nutritional significance from an ingredient name alone.",
             confidence="medium",
         )
+    if fuzzy_matches:
+        return ClaimAudit(
+            claim=claim,
+            verdict=Verdict.CANNOT_VERIFY,
+            summary="A possible one-character OCR match appears in the ingredient list, but the physical packet should be checked.",
+            evidence=_ingredient_evidence(ingredients, fuzzy_matches),
+            caveat="Near-matches are surfaced as possible OCR corrections, never treated as exact evidence.",
+            confidence="low",
+        )
 
     return ClaimAudit(
         claim=claim,

src/packetcourt/investigator.py

@@ -65,6 +65,17 @@ def build_investigation(
         missing.append("A readable ingredient list")
     if claim_names and nutrition.basis == "unknown" and any(policy_tool_for(name) == "inspect_nutrition" for name in claim_names):
         missing.append("A readable nutrition panel with its measurement basis")
+    elif claim_names and any(policy_tool_for(name) == "inspect_nutrition" for name in claim_names) and not any(
+        value is not None
+        for value in (
+            nutrition.protein_g,
+            nutrition.total_sugar_g,
+            nutrition.added_sugar_g,
+            nutrition.sodium_mg,
+            nutrition.saturated_fat_g,
+        )
+    ):
+        missing.append("Readable nutrient quantities from the nutrition table")
     if expiry.instruction and not expiry.packed_on:
         missing.append("The packing or manufacturing date needed to resolve relative shelf life")
 

src/packetcourt/parser.py

@@ -49,9 +49,18 @@ def extract_claims(front_text: str) -> list[str]:
 
 
 def _number_after(label: str, text: str, unit: str) -> float | None:
-    pattern = rf"\b{label}\b[^0-9]{{0,24}}(\d+(?:\.\d+)?)\s*{unit}\b"
-    match = re.search(pattern, text, re.IGNORECASE)
-    return float(match.group(1)) if match else None
+    # Labels transcribed from nutrition tables commonly place the unit before
+    # the value: "Protein (g) 12" or "Sodium | mg | 410".
+    patterns = [
+        rf"\b{label}\b[^0-9]{{0,28}}(\d+(?:\.\d+)?)\s*{unit}\b",
+        rf"\b{label}\b[\s|:()\[\]\-]*{unit}[\s|:()\[\]\-]*(\d+(?:\.\d+)?)\b",
+        rf"\b{label}\b[\s|:()\[\]\-]+(\d+(?:\.\d+)?)\b",
+    ]
+    for pattern in patterns:
+        match = re.search(pattern, text, re.IGNORECASE)
+        if match:
+            return float(match.group(1))
+    return None
 
 
 def parse_nutrition(back_text: str) -> NutritionFacts:
@@ -88,6 +97,21 @@ def calculate_whole_packet(nutrition: NutritionFacts) -> WholePacketNutrition:
         )
     if multiplier is None:
         return WholePacketNutrition(explanation=explanation)
+    values = (
+        nutrition.protein_g,
+        nutrition.total_sugar_g,
+        nutrition.added_sugar_g,
+        nutrition.sodium_mg,
+        nutrition.saturated_fat_g,
+    )
+    if not any(value is not None for value in values):
+        return WholePacketNutrition(
+            multiplier=round(multiplier, 2),
+            explanation=(
+                f"Found {nutrition.basis} and a {nutrition.package_size_g:g}g packet, "
+                "but no readable nutrient quantities were extracted."
+            ),
+        )
 
     def scale(value: float | None) -> float | None:
         return round(value * multiplier, 1) if value is not None else None
@@ -116,7 +140,7 @@ def extract_ingredients(back_text: str) -> list[str]:
     )
     if not match:
         return []
-    return [item.strip(" .") for item in re.split(r"[,;]", match.group(1)) if item.strip()]
+    return [item.strip(" .*_") for item in re.split(r"[,;]", match.group(1)) if item.strip(" .*_")]
 
 
 def _parse_date(value: str) -> date | None:

src/packetcourt/vlm.py

@@ -17,6 +17,9 @@ PROMPTS = {
         "Transcribe the visible food-label evidence from this package image. Focus on the "
         "ingredient list, nutrition values with their basis, net weight, FSSAI license, "
         "manufacturing or packing date, best-before or use-by date, and after-opening instructions. "
+        "For nutrition tables, preserve every visible row as 'nutrient name | unit | value' and "
+        "include the declared basis such as per 100g or per serving. Do not summarize or omit rows "
+        "whose value is zero. "
         "Return only visibly printed evidence. Do not explain or infer anything."
     ),
 }

tests/test_audit.py

@@ -130,3 +130,45 @@ def test_multiple_photo_extractions_are_labeled_and_exact_duplicates_are_skipped
     assert text.count("HIGH PROTEIN") == 1
     assert "2 unique front photos" in status
     assert "Exact duplicate skipped" in images[1]["status"]
+
+
+def test_table_style_nutrition_values_are_calculated():
+    result = audit_packet(
+        "SUGAR FREE",
+        (
+            "Nutrition Information Per 100g | Protein (g) 12 | Total Sugars | g | 0 | "
+            "Added Sugar (g) 0 | Sodium (mg) 410 | Saturated Fat g 2.5 | Net Weight: 200g."
+        ),
+    )
+    assert result.nutrition.protein_g == 12
+    assert result.nutrition.total_sugar_g == 0
+    assert result.nutrition.sodium_mg == 410
+    assert result.whole_packet.protein_g == 24
+    assert result.whole_packet.total_sugar_g == 0
+    assert result.whole_packet.sodium_mg == 820
+
+
+def test_basis_and_packet_without_nutrient_rows_explain_the_real_missing_evidence():
+    result = audit_packet("SUGAR FREE", "Nutrition Information Per 100g. Net Weight: 200g.")
+    assert result.whole_packet.calculable is False
+    assert "no readable nutrient quantities" in result.whole_packet.explanation.lower()
+    assert any("nutrient quantities" in item.lower() for item in result.investigation.missing_evidence)
+
+
+def test_enrichment_claim_cites_visible_back_label_evidence():
+    result = audit_packet(
+        "Extra Calcium with DHA",
+        "Nutrition per 100g: Calcium 400mg, DHA 25mg. Net Weight: 200g.",
+    )
+    claim = by_claim(result, "Extra Calcium with DHA")
+    assert claim.verdict == Verdict.CONTEXT_MISSING
+    assert any("Calcium" in evidence.text for evidence in claim.evidence)
+
+
+def test_one_character_ocr_claim_mismatch_is_surfaced_conservatively():
+    result = audit_packet("Real Badar", "Ingredients: ** Maltodextrin (65%), Badam, Sucralose.")
+    claim = by_claim(result, "Real Badar")
+    assert claim.verdict == Verdict.CANNOT_VERIFY
+    assert claim.confidence == "low"
+    assert any(evidence.text == "Badam" for evidence in claim.evidence)
+    assert result.ingredients[0] == "Maltodextrin (65%)"

vision_space/app.py

@@ -17,7 +17,9 @@ PROMPTS = {
     ),
     "back": (
         "Transcribe only visibly printed food-label evidence. Focus on ingredients, nutrition values "
-        "and basis, net weight, FSSAI license, dates, and after-opening instructions. Do not infer."
+        "and basis, net weight, FSSAI license, dates, and after-opening instructions. For nutrition "
+        "tables, preserve every visible row as 'nutrient name | unit | value', include the declared "
+        "basis, and do not omit zero values. Do not summarize or infer."
     ),
 }