Update app.py

app.py

@@ -85,50 +85,55 @@ def clean_json_response(text):
     if not text:
         return None
     orig = text
-    # strip fences
+    # strip ``` fences
     text = re.sub(r'```(?:json)?', '', text).strip()
-    # grab braces
-    start = text.find('{')
-    end = text.rfind('}') + 1
+    # find outer braces
+    start, end = text.find('{'), text.rfind('}') + 1
     if start < 0 or end < 1:
-        st.error("Couldn't locate JSON")
+        st.error("Couldn't locate JSON in response.")
         st.code(orig)
         return None
-    fragment = text[start:end]
-    # remove stray trailing commas before } or ]
-    fragment = re.sub(r',\s*([}\]])', r'\1', fragment)
+    frag = text[start:end]
+    # remove stray trailing commas
+    frag = re.sub(r',\s*([}\]])', r'\1', frag)
     try:
-        return json.loads(fragment)
+        return json.loads(frag)
     except json.JSONDecodeError as e:
-        st.error(f"JSON parse error: {e}")
-        st.code(fragment)
-        return None
+        # attempt to insert missing commas between adjacent fields
+        repaired = re.sub(r'"\s*"\s*(?="[^"]+"\s*:)', '","', frag)
+        try:
+            return json.loads(repaired)
+        except json.JSONDecodeError:
+            st.error(f"JSON parse error: {e}")
+            st.code(frag)
+            return None
 
 def fallback_supplier(text):
-    # first non-empty line heuristic
-    lines = [l.strip() for l in text.splitlines() if l.strip()]
-    return lines[0] if lines else None
+    for line in text.splitlines():
+        line = line.strip()
+        if line:
+            return line
+    return None
 
 def get_extraction_prompt(model_choice, txt):
-    # every prompt now demands "json" and COMPACT JSON output
     if model_choice.startswith("DeepSeek"):
         return (
-            "Extract full invoice info below and RETURN ONLY a valid json object (compact, single line) with these fields:\n"
-            '{"invoice_number":"string","invoice_date":"YYYY-MM-DD","po_number":"string|null",'
-            '"invoice_value":"string with currency","line_items":[{"description":"string","quantity":"number",'
-            '"unit_price":"string with currency","total_price":"string with currency"}]}\n'
-            "Use null for missing fields. NO extra text.\n\n"
+            "Extract full invoice info and RETURN ONLY a single-line json object with fields:\n"
+            '{"invoice_number":"string","invoice_date":"YYYY-MM-DD",'
+            '"po_number":"string|null","invoice_value":"string with currency",'
+            '"line_items":[{"description":"string","quantity":"number","unit_price":"string with currency","total_price":"string with currency"}]}\n'
+            "Use null for missing. NO extra text.\n\n"
             f"Invoice Text:\n{txt}"
         )
     else:
         return (
-            "You are given invoice text. Extract data and RETURN ONLY a compact json object (one line) exactly like this:\n"
+            "Extract invoice data and RETURN ONLY a compact, one-line json object exactly:\n"
             '{"invoice_header":{"invoice_number":"string","invoice_date":"YYYY-MM-DD",'
             '"po_number":"string|null","invoice_value":"string with currency",'
             '"supplier_name":"string|null","customer_name":"string|null"},'
             '"line_items":[{"item_number":"string|null","description":"string","quantity":number,'
             '"unit_price":"string with currency","total_price":"string with currency"}]}\n'
-            "Use null for missing. NO explanations or extra keys.\n\n"
+            "Use null for missing. NO extras.\n\n"
             f"Invoice Text:\n{txt}"
         )
 
@@ -141,62 +146,62 @@ def extract_invoice_info(model_choice, text):
     if not data:
         return None
 
-    # normalize header + fallback supplier
+    # normalize + supplier fallback
     if model_choice in ("Llama 4 Mavericks","Mistral Small"):
-        hdr = data.setdefault("invoice_header",{})
+        hdr = data.setdefault("invoice_header", {})
         for k in ("invoice_number","invoice_date","po_number","invoice_value","supplier_name","customer_name"):
-            hdr.setdefault(k,None)
+            hdr.setdefault(k, None)
         if not hdr.get("supplier_name"):
             hdr["supplier_name"] = fallback_supplier(text)
-        items = data.setdefault("line_items",[])
+        items = data.setdefault("line_items", [])
         for itm in items:
             for k in ("item_number","description","quantity","unit_price","total_price"):
-                itm.setdefault(k,None)
+                itm.setdefault(k, None)
     else:
         for k in ("invoice_number","invoice_date","po_number","invoice_value"):
-            data.setdefault(k,None)
-        items = data.setdefault("line_items",[])
+            data.setdefault(k, None)
+        items = data.setdefault("line_items", [])
         for itm in items:
             for k in ("description","quantity","unit_price","total_price"):
-                itm.setdefault(k,None)
+                itm.setdefault(k, None)
 
     return data
 
-# UI
+# ---- UI ----
 tab1, tab2 = st.tabs(["PDF Summarizer","Invoice Extractor"])
 
 with tab1:
-    st.title("PDF → Bullet-Point Summarizer")
-    pdf = st.file_uploader("Upload PDF",type="pdf")
-    pct = st.slider("Summarization %",1,100,20)
+    st.title("PDF → Bullet Points")
+    pdf = st.file_uploader("Upload PDF", type="pdf")
+    pct = st.slider("Summarization %", 1, 100, 20)
     if st.button("Summarize") and pdf:
         txt = read_pdf(io.BytesIO(pdf.getvalue()))
         keys = extract_key_phrases(txt)
-        scores = score_sentences(txt,keys)
+        scores = score_sentences(txt, keys)
         n = max(1, len(scores)*pct//100)
-        st.markdown(summarize_text(scores,num_points=n))
+        st.markdown(summarize_text(scores, num_points=n))
 
 with tab2:
     st.title("Invoice Extractor")
-    mdl = st.selectbox("Model",list(MODELS.keys()))
-    inv_pdf = st.file_uploader("Invoice PDF",type="pdf")
+    mdl = st.selectbox("Model", list(MODELS.keys()))
+    inv_pdf = st.file_uploader("Invoice PDF", type="pdf")
     if st.button("Extract") and inv_pdf:
         txt = read_pdf(io.BytesIO(inv_pdf.getvalue()))
-        info = extract_invoice_info(mdl,txt)
+        info = extract_invoice_info(mdl, txt)
         if info:
-            st.success("Done")
+            st.success("Extraction Complete")
             if mdl in ("Llama 4 Mavericks","Mistral Small"):
                 h=info["invoice_header"]
-                c1,c2,c3=st.columns(3)
-                c1.metric("Invoice #",h["invoice_number"]);c1.metric("Supplier",h["supplier_name"])
-                c2.metric("Date",h["invoice_date"]);c2.metric("Customer",h["customer_name"])
-                c3.metric("PO #",h["po_number"]);c3.metric("Total",h["invoice_value"])
-                st.table(info["line_items"])
+                c1,c2,c3 = st.columns(3)
+                c1.metric("Invoice #", h["invoice_number"]); c1.metric("Supplier", h["supplier_name"])
+                c2.metric("Date", h["invoice_date"]);    c2.metric("Customer", h["customer_name"])
+                c3.metric("PO #", h["po_number"]);        c3.metric("Total", h["invoice_value"])
+                st.subheader("Line Items"); st.table(info["line_items"])
             else:
-                c1,c2=st.columns(2)
-                c1.metric("Invoice #",info["invoice_number"]);c1.metric("PO #",info["po_number"])
-                c2.metric("Date",info["invoice_date"]);c2.metric("Value",info["invoice_value"])
-                st.table(info["line_items"])
+                c1,c2 = st.columns(2)
+                c1.metric("Invoice #", info["invoice_number"]); c1.metric("PO #", info["po_number"])
+                c2.metric("Date", info["invoice_date"]);        c2.metric("Value", info["invoice_value"])
+                st.subheader("Line Items"); st.table(info["line_items"])
 
     if "last_api" in st.session_state:
         with st.expander("Debug"):