Dashboard: live badge, provenance banner, and About & Skills page

README.md

@@ -1,7 +1,11 @@
 # DocuParse
 
+[![Open in Streamlit](https://static.streamlit.io/badges/streamlit_badge_black_white.svg)](https://docuparse.streamlit.app)
+
 **Intelligent document parser for financial filings** - Extracts text, tables, and validates data from SEC 10-K/10-Q documents using multiple AI models and cross-verification.
 
+🔗 **Live results dashboard:** [docuparse.streamlit.app](https://docuparse.streamlit.app)
+
 ## What This Does
 
 DocuParse is a complete pipeline that:
@@ -40,6 +44,10 @@ ls data/exports/
 
 ## 📊 Results Dashboard
 
+[![Open in Streamlit](https://static.streamlit.io/badges/streamlit_badge_black_white.svg)](https://docuparse.streamlit.app)
+
+**Live app:** [docuparse.streamlit.app](https://docuparse.streamlit.app)
+
 An interactive Streamlit dashboard visualizes the pipeline's recorded outputs —
 evaluation metrics, per-stage benchmarks (runtime & memory), build-vs-buy cost
 analysis, distribution drift, and the analysis reports. It reads the committed

dashboard/app.py

@@ -31,6 +31,16 @@ def _fmt(value, digits: int = 4) -> str:
     return "—" if value is None else f"{value:.{digits}f}"
 
 
+def render_provenance() -> None:
+    """A banner shown on every page so viewers know the data is a static snapshot."""
+    when = dl.data_as_of() or "a previous run"
+    st.info(
+        f"📌 **Static snapshot — figures are from the pipeline run on {when}.** "
+        "This dashboard reads committed result files; it does **not** run the "
+        "extraction pipeline live."
+    )
+
+
 # --------------------------------------------------------------------------- #
 # Sidebar
 # --------------------------------------------------------------------------- #
@@ -41,7 +51,8 @@ def _fmt(value, digits: int = 4) -> str:
 )
 page = st.sidebar.radio(
     "View",
-    ["Overview", "Benchmarks", "Cost (Build vs Buy)", "Distribution Drift", "Reports"],
+    ["About & Skills", "Overview", "Benchmarks", "Cost (Build vs Buy)",
+     "Distribution Drift", "Reports"],
 )
 st.sidebar.info(
     "This dashboard visualizes recorded runs. A live 'upload a PDF and parse' "
@@ -224,11 +235,61 @@ def render_reports() -> None:
     st.markdown(dl.read_report(reports[choice]))
 
 
+# --------------------------------------------------------------------------- #
+# About & Skills — the landing page
+# --------------------------------------------------------------------------- #
+# A curated shortlist of the most valuable concepts behind the project — not an
+# exhaustive catalogue of every library used.
+KEY_SKILLS = [
+    ("Document AI & layout understanding",
+     "Parsing complex financial PDFs with OCR fallback and layout/table models "
+     "(Docling, LayoutParser, Tesseract, Camelot)."),
+    ("Reproducible ML pipelines (MLOps)",
+     "A staged, parameterized DVC pipeline from download through export."),
+    ("Quantitative evaluation",
+     "Text WER/CER and table precision/recall/F1, with regression and "
+     "distribution-drift monitoring."),
+    ("Data validation",
+     "Cross-verifying extracted figures against authoritative SEC XBRL data."),
+    ("Performance & cost engineering",
+     "Per-stage runtime/memory benchmarking and a build-vs-buy cost analysis."),
+]
+
+
+def render_about() -> None:
+    st.title("📄 DocuParse — Financial Filing Parser")
+    st.markdown(
+        "An end-to-end pipeline that extracts text, tables, and structure from "
+        "**SEC financial filings (10-K / 10-Q)**, measures the extraction quality, "
+        "and cross-checks the numbers against authoritative **XBRL** data. "
+        "The tabs on the left present the pipeline's recorded results."
+    )
+
+    st.subheader("🧠 Key concepts & skills")
+    for title, desc in KEY_SKILLS:
+        st.markdown(f"- **{title}** — {desc}")
+
+    st.subheader("🛠️ Built with")
+    st.markdown("`Python` · `DVC` · `Docling` · `pandas` · `Streamlit`")
+
+    st.subheader("🔗 Links")
+    st.markdown(
+        "- **Source code:** https://github.com/Effyrt/Docuparse\n"
+        "- **Demo video:** "
+        "https://drive.google.com/file/d/1w8RPBch1nPV8BpZIw0tFLPD1BmK0rkfN/view\n"
+        "- **Interactive tutorial (CodeLabs):** "
+        "https://codelabs-preview.appspot.com/?file_id=1eoeyKHeNX_qYq6m8oL37XLQMEoLCK7Xv02sBSGAGbwg#0"
+    )
+
+
 PAGES = {
+    "About & Skills": render_about,
     "Overview": render_overview,
     "Benchmarks": render_benchmarks,
     "Cost (Build vs Buy)": render_cost,
     "Distribution Drift": render_drift,
     "Reports": render_reports,
 }
+
+render_provenance()
 PAGES[page]()

dashboard/data_loader.py

@@ -31,9 +31,9 @@ def _read_json(path: Path) -> Optional[Any]:
         return None
 
 
-def _latest(pattern: str) -> Optional[Path]:
-    """Return the most recently modified file matching a glob under the repo."""
-    matches = glob.glob(str(REPO_ROOT / pattern))
+def _latest(pattern: str, root: Path = REPO_ROOT) -> Optional[Path]:
+    """Return the most recently modified file matching a glob under ``root``."""
+    matches = glob.glob(str(root / pattern))
     if not matches:
         return None
     return Path(max(matches, key=os.path.getmtime))
@@ -55,6 +55,40 @@ def load_metrics_history(root: Path = REPO_ROOT) -> List[Dict[str, Any]]:
     return sorted(data, key=lambda r: r.get("timestamp", ""))
 
 
+def _all_timestamps(root: Path = REPO_ROOT) -> List[str]:
+    """Collect ISO timestamps from every result file that records one."""
+    stamps: List[str] = []
+
+    bench = load_benchmark(root).get("benchmark_info", {}) or {}
+    stamps.append(bench.get("timestamp"))
+
+    stamps.append((load_drift(root) or {}).get("timestamp"))
+
+    history = load_metrics_history(root)
+    if history:
+        stamps.append(history[-1].get("timestamp"))
+
+    summary = _read_json(root / "evaluation" / "latest_evaluation_summary.json") or {}
+    stamps.append(summary.get("evaluation_timestamp"))
+
+    cost_info = (load_cost_analysis(root).get("analysis_info", {}) or {})
+    stamps.append(cost_info.get("timestamp"))
+
+    return [s for s in stamps if s]
+
+
+def data_as_of(root: Path = REPO_ROOT) -> Optional[str]:
+    """Return the date (YYYY-MM-DD) of the most recent recorded pipeline run.
+
+    Used to tell viewers the dashboard shows a static snapshot, not live data.
+    ISO-8601 strings sort lexicographically, so max() gives the latest.
+    """
+    stamps = _all_timestamps(root)
+    if not stamps:
+        return None
+    return max(stamps)[:10]
+
+
 def metric_cards(metrics: Dict[str, Any]) -> List[Dict[str, Any]]:
     """Shape the headline metrics into display cards with pass/fail vs thresholds.
 
@@ -115,7 +149,7 @@ def metric_cards(metrics: Dict[str, Any]) -> List[Dict[str, Any]]:
 # --------------------------------------------------------------------------- #
 def load_benchmark(root: Path = REPO_ROOT) -> Dict[str, Any]:
     """Load the most recent corrected pipeline benchmark."""
-    path = _latest("benchmarks/results/CORRECTED_pipeline_benchmark_*.json")
+    path = _latest("benchmarks/results/CORRECTED_pipeline_benchmark_*.json", root)
     return _read_json(path) if path else {}
 
 
@@ -152,7 +186,7 @@ def benchmark_failures(benchmark: Dict[str, Any]) -> List[Dict[str, Any]]:
 # --------------------------------------------------------------------------- #
 def load_cost_analysis(root: Path = REPO_ROOT) -> Dict[str, Any]:
     """Load the most recent cloud vs infrastructure cost analysis."""
-    path = _latest("benchmarks/results/cost_analysis_*.json")
+    path = _latest("benchmarks/results/cost_analysis_*.json", root)
     return _read_json(path) if path else {}
 
 

tests/unit/test_dashboard_data.py

@@ -99,3 +99,27 @@ def test_real_reports_listed():
     names = {p.name for p in dl.list_reports()}
     assert "benchmarks.md" in names
     assert "xbrl_cross_verification_report.md" in names
+
+
+# --- snapshot-date provenance ----------------------------------------------- #
+def test_data_as_of_none_for_empty_root(tmp_path):
+    assert dl.data_as_of(tmp_path) is None
+
+
+def test_data_as_of_picks_latest_date(tmp_path):
+    # Only a metrics history file exists; data_as_of should return its latest date.
+    hist_dir = tmp_path / "evaluation" / "metrics"
+    hist_dir.mkdir(parents=True)
+    (hist_dir / "metrics_history.json").write_text(
+        '[{"timestamp": "2024-01-01T10:00:00"}, '
+        '{"timestamp": "2025-06-15T12:00:00"}]',
+        encoding="utf-8",
+    )
+    assert dl.data_as_of(tmp_path) == "2025-06-15"
+
+
+def test_data_as_of_real_data_is_a_date():
+    as_of = dl.data_as_of()
+    assert as_of is not None
+    # YYYY-MM-DD shape
+    assert len(as_of) == 10 and as_of[4] == "-" and as_of[7] == "-"