Fix ChromaDB tenant error; add real CRPF tenders; update Tab 2

- core/vectorstore.py: switch to EphemeralClient (in-memory) — eliminates
ChromaDB 0.5.x tenant/SQLite compatibility error entirely; index rebuilds
on each evaluation run so persistence is not needed
- data/tender/real_tenders/: two real CRPF tender PDFs downloaded from
crpf.gov.in (Bhopal water tanks NIT-71 Aug-2025, J&K office repair Jan-2026)
- ui/tab_tender.py: pre-loaded tender selector with tabs (pre-loaded vs upload);
real tenders shown with green badge; active tender name always displayed

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

core/vectorstore.py

@@ -7,14 +7,11 @@ from core.config import CHROMA_DIR
 
 @st.cache_resource
 def get_client():
-    import shutil
     import chromadb
-    try:
-        return chromadb.PersistentClient(path=CHROMA_DIR)
-    except ValueError:
-        # Stale or version-incompatible .chroma directory — wipe and recreate
-        shutil.rmtree(CHROMA_DIR, ignore_errors=True)
-        return chromadb.PersistentClient(path=CHROMA_DIR)
+    # EphemeralClient (in-memory) avoids all ChromaDB 0.5.x tenant/SQLite
+    # compatibility issues. The index is rebuilt each evaluation run, so
+    # persistence is not needed.
+    return chromadb.EphemeralClient()
 
 
 def get_collection(name: str):

data/tender/real_tenders/crpf_bhopal_water_tanks_2025.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:718a035f8911294cd1777805dd388b1781edbc5527af7ba5275bb1e8343d888b
+size 1267097

data/tender/real_tenders/crpf_jammu_office_repair_2026.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdc49ca6e2bc7a733c11b8672bf4adeceb487537f344d57ab0da67f18788e566
+size 484042

ui/tab_overview.py

@@ -27,7 +27,7 @@ def render() -> None:
     )
 
     # KPIs
-    criteria_count = len(st.session_state.get("criteria", load_criteria()))
+    criteria_count = len(st.session_state.get("criteria") or [])
     verdicts       = st.session_state.get("verdicts", {})
     checked        = sum(1 for bv in verdicts.values() for _ in bv)
     audit_count    = len(audit.query())

ui/tab_tender.py

@@ -7,63 +7,105 @@ from core import criteria_extractor
 from core.config import DATA_DIR
 
 _MOCK_TENDER = DATA_DIR / "tender" / "crpf_construction_tender.pdf"
+_REAL_DIR    = DATA_DIR / "tender" / "real_tenders"
 
-_CATEGORY_COLORS = {
-    "financial": "🔵",
-    "technical": "🟢",
-    "compliance": "🟠",
+_REAL_LABELS = {
+    "crpf_bhopal_water_tanks_2025.pdf":
+        "CRPF GC Bhopal — Water Storage Tanks (NIT-71, Aug 2025, Est. ₹62.9L)",
+    "crpf_jammu_office_repair_2026.pdf":
+        "CRPF J&K Zone HQ Jammu — Office Building Repair (Jan 2026, Est. ₹24.3L)",
 }
 
+_CAT_ICONS = {"financial": "🔵", "technical": "🟢", "compliance": "🟠"}
+
 
 def render() -> None:
-    st.header("Tender Analysis")
+    st.markdown(
+        '<h2 style="font-family:Inter,sans-serif;font-weight:800;font-size:1.5rem;'
+        'color:#0D1B2A;margin-bottom:4px;">Tender Analysis</h2>'
+        '<p style="color:#64748B;font-size:0.875rem;margin-bottom:1rem;">'
+        'Extract eligibility criteria from any tender PDF using DeepSeek.</p>',
+        unsafe_allow_html=True,
+    )
+
+    # ── Tender source selector ────────────────────────────────────────────────
+    real_files = sorted(_REAL_DIR.glob("*.pdf")) if _REAL_DIR.exists() else []
+
+    preset_options = {"Mock tender (CRPF Construction, 5 criteria)": _MOCK_TENDER}
+    for f in real_files:
+        preset_options[_REAL_LABELS.get(f.name, f.name)] = f
+
+    tab_preset, tab_upload = st.tabs(["📂 Pre-loaded Tenders", "⬆️ Upload Your Own"])
+
+    with tab_preset:
+        chosen_label = st.selectbox("Select tender", options=list(preset_options.keys()))
+        tender_path  = preset_options[chosen_label]
+        tender_name  = tender_path.name
 
-    uploaded = st.file_uploader("Upload tender PDF (leave blank to use pre-loaded mock)", type=["pdf"])
+        if real_files and chosen_label != "Mock tender (CRPF Construction, 5 criteria)":
+            st.markdown(
+                '<div style="background:#F0FDF4;border:1px solid #BBF7D0;border-radius:8px;'
+                'padding:10px 14px;font-size:0.83rem;color:#166534;margin-top:6px;">'
+                '✅ <strong>Real government tender</strong> — downloaded from crpf.gov.in</div>',
+                unsafe_allow_html=True,
+            )
 
-    if uploaded:
-        tender_bytes = uploaded.read()
-        tender_name = uploaded.name
-        with tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) as tmp:
-            tmp.write(tender_bytes)
-            tender_path = Path(tmp.name)
-    else:
-        tender_path = _MOCK_TENDER
-        tender_name = _MOCK_TENDER.name
+    with tab_upload:
+        uploaded = st.file_uploader("Upload a tender PDF", type=["pdf"],
+                                    label_visibility="collapsed")
+        if uploaded:
+            tender_bytes = uploaded.read()
+            tender_name  = uploaded.name
+            with tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) as tmp:
+                tmp.write(tender_bytes)
+                tender_path = Path(tmp.name)
 
-    st.caption(f"Using: **{tender_name}**")
+    st.caption(f"Active: **{tender_name}**")
 
+    # ── Extract button ────────────────────────────────────────────────────────
     if st.button("Extract Criteria (Live LLM)", type="primary"):
         with st.spinner("Calling DeepSeek to extract eligibility criteria…"):
             criteria = criteria_extractor.extract_criteria(tender_path)
-        st.session_state["criteria"] = [c.model_dump() for c in criteria]
-        st.session_state["tender_path"] = str(tender_path)
+        st.session_state["criteria"]     = [c.model_dump() for c in criteria]
+        st.session_state["tender_path"]  = str(tender_path)
+        st.success(f"Extracted {len(criteria)} criteria.")
 
+    # ── Display results ───────────────────────────────────────────────────────
     criteria_data = st.session_state.get("criteria")
-    if criteria_data:
-        st.success(f"Extracted **{len(criteria_data)}** criteria")
-
-        if st.session_state.get("fallback_active"):
-            st.warning("⚠ Live API unavailable — showing pre-computed criteria.")
-
-        for c in criteria_data:
-            mandatory_badge = "🔴 Mandatory" if c["mandatory"] else "🟡 Optional"
-            cat_icon = _CATEGORY_COLORS.get(c["category"], "⚪")
-            label = f"{cat_icon} **{c['id']}** — {c['title']}  {mandatory_badge}"
-            with st.expander(label, expanded=False):
-                col1, col2 = st.columns([2, 1])
-                with col1:
-                    st.markdown(f"**Description:** {c['description']}")
-                    rule = c["rule"]
-                    rule_parts = [f"Type: `{rule['type']}`", f"Field: `{rule['field']}`",
-                                  f"Operator: `{rule['operator']}`"]
-                    if rule.get("value") is not None:
-                        rule_parts.append(f"Value: `{rule['value']}`")
-                    if rule.get("unit"):
-                        rule_parts.append(f"Unit: `{rule['unit']}`")
-                    st.markdown(" · ".join(rule_parts))
-                with col2:
-                    st.markdown(f"**Category:** {c['category'].capitalize()}")
-                    st.markdown(f"**Source:** Page {c['source_page']}, Clause {c['source_clause']}")
-                    if c.get("query_hints"):
-                        hints = ", ".join(f"`{h}`" for h in c["query_hints"])
-                        st.markdown(f"**Query hints:** {hints}")
+    if not criteria_data:
+        return
+
+    if st.session_state.get("fallback_active"):
+        st.warning("⚠ Live API unavailable — showing pre-computed criteria.")
+
+    st.markdown(
+        f'<div style="font-size:0.9rem;font-weight:700;color:#0D1B2A;margin:1rem 0 0.5rem;">'
+        f'Extracted {len(criteria_data)} criteria</div>',
+        unsafe_allow_html=True,
+    )
+
+    for c in criteria_data:
+        icon     = _CAT_ICONS.get(c["category"], "⚪")
+        mand_lbl = "🔴 Mandatory" if c["mandatory"] else "🟡 Optional"
+        rule     = c["rule"]
+        rule_str = f'`{rule["type"]}` · `{rule["field"]} {rule["operator"]}'
+        if rule.get("value") is not None:
+            rule_str += f' {rule["value"]}'
+        if rule.get("unit"):
+            rule_str += f' {rule["unit"]}'
+        rule_str += "`"
+
+        with st.expander(
+            f'{icon} **{c["id"]}** — {c["title"]}  ·  {mand_lbl}',
+            expanded=False,
+        ):
+            col1, col2 = st.columns([2, 1])
+            with col1:
+                st.markdown(f"**Description:** {c['description']}")
+                st.markdown(f"**Rule:** {rule_str}")
+                if c.get("query_hints"):
+                    hints = "  ·  ".join(f"`{h}`" for h in c["query_hints"])
+                    st.markdown(f"**Query hints:** {hints}")
+            with col2:
+                st.markdown(f"**Category:** {c['category'].capitalize()}")
+                st.markdown(f"**Source:** Page {c['source_page']}, Clause `{c['source_clause']}`")