polish: refine H2EPR Explorer presentation

README.md

@@ -8,61 +8,56 @@ license: apache-2.0
 ---
 
 [![Project Website](https://img.shields.io/badge/Project_Website-Visit-176B70?style=flat-square)](https://agenticfinlab.github.io/H2EPR-Bench/)
-[![Public Dataset](https://img.shields.io/badge/Public_Dataset-Unified--3000-176B70?style=flat-square)](https://huggingface.co/datasets/AgenticFinLab/H2EPR-Bench)
+[![Public Dataset](https://img.shields.io/badge/Public_Dataset-3%2C000_events-176B70?style=flat-square)](https://huggingface.co/datasets/AgenticFinLab/H2EPR-Bench)
 [![FinMycelium System](https://img.shields.io/badge/FinMycelium-System-176B70?style=flat-square)](https://github.com/AgenticFinLab/FinMycelium)
 [![Reference EPGs (Gated)](https://img.shields.io/badge/Reference_EPGs-Gated-176B70?style=flat-square)](https://huggingface.co/datasets/AgenticFinLab/H2EPR-Bench-Gold)
 [![Release Repository](https://img.shields.io/badge/Release_Repository-Source-176B70?style=flat-square)](https://github.com/AgenticFinLab/H2EPR-Bench)
 
 # H²EPR-Bench Explorer
 
-H²EPR-Bench Explorer is the interactive browsing layer for the Unified-3000
-release of `AgenticFinLab/H2EPR-Bench`. The existing Docker Space runs a
-Streamlit interface for searching all 3,000 catalog events, inspecting current
-Draft EPG summaries and stage timelines, and previewing or downloading the
-public per-event Draft EPG for every event.
+Explore 3,000 real-world event processes without navigating raw Dataset files.
+Search by title, domain, category, or keyword; inspect graph statistics and
+ordered stages; open an interactive timeline; and preview or download the
+public Draft EPG for any event.
 
-**Release boundary:** public Draft EPGs are sanitized FinMycelium construction
-artifacts. Official benchmark scoring uses expert-adjudicated reference EPGs in
-the [manual-gated companion](https://huggingface.co/datasets/AgenticFinLab/H2EPR-Bench-Gold).
-The Explorer does not load reference EPGs or frozen evidence packages.
+## What you can explore
 
-The Explorer source code is Apache-2.0. Dataset content displayed by the app,
-including public Draft EPGs, remains CC BY-NC 4.0 under the Dataset release.
+- **3,000 event profiles** spanning six domains and 26 categories
+- **8,843 ordered stages** with temporal and graph-level summaries
+- **Interactive timelines** for calendar-based and relative-order processes
+- **Draft EPG JSON** with participants, actions, outcomes, and typed relations
+- **Direct links** to the Dataset and gated reference collection
 
-## Public data contract
+The Explorer presents the public FinMycelium Draft EPGs. Expert-finalized
+reference EPGs for official scoring are available through the
+[gated companion](https://huggingface.co/datasets/AgenticFinLab/H2EPR-Bench-Gold).
 
-The Explorer validates five Parquet tables from one release identity:
+## Data connection
+
+The app reads five Parquet views from
+[AgenticFinLab/H2EPR-Bench](https://huggingface.co/datasets/AgenticFinLab/H2EPR-Bench):
 
 | Table | Role |
 |---|---|
-| `event_gallery.parquet` | Lightweight event discovery metadata |
-| `event_catalog.parquet` | Stable event discovery metadata |
-| `event_instances.parquet` | Public artifact-access fields |
-| `finalcascade_summary.parquet` | Draft graph counts and event-level temporal summary |
-| `event_stages.parquet` | Ordered stage rows for every Draft EPG |
-
-Selected Draft EPGs are loaded directly from
-`draft_events/<H2EPR-ID>/draft_epg.json`. The path is derived only after the
-canonical event ID has been validated. `manifests/draft_source_hashes.csv`
-binds every direct file to its canonical source-payload digest and sanitized
-record digest; the Explorer verifies both identities and the canonical JSON
-digest before displaying a graph.
+| `event_gallery.parquet` | Lightweight event discovery |
+| `event_catalog.parquet` | Complete event registry |
+| `event_instances.parquet` | Metadata and artifact access fields |
+| `finalcascade_summary.parquet` | Event-level graph and temporal summaries |
+| `event_stages.parquet` | Ordered Draft-EPG stages |
 
-Remote reads and immutable per-file links are pinned to Dataset commit
-`6156a6bb3b838143401cb3e5709f708e5d6e802c`. They fail closed if that fixed
-release cannot be resolved and never fall back to a floating branch.
+Event-level graphs are loaded from
+`draft_events/<H2EPR-ID>/draft_epg.json`. Remote reads are pinned to Dataset
+commit `4b0f0f4000db3ba9b6e1a720e5b5cfbaae68353c`, and each file is verified
+against the release hash registry before display.
 
 ## Local development
 
-The app can use the frozen local Unified-3000 release tree without any network
-access:
+Point the app at a complete local Dataset tree:
 
-```bash
+~~~bash
 export H2EPR_EXPLORER_LOCAL_DATASET_DIR=/path/to/H2EPR-Bench
 streamlit run app.py
-```
+~~~
 
-For local verification, `H2EPR_EXPLORER_LOCAL_DATASET_DIR` is required. In the
-published Space, the source constant fixes the immutable Dataset revision; the
-application exposes no runtime revision override, so one process cannot mix
-releases.
+The Explorer source is Apache-2.0. Dataset content displayed by the app is
+released under CC BY-NC 4.0.

app.py

@@ -21,7 +21,6 @@ from h2epr_explorer.constants import (
     PUBLIC_DATASET_REPO,
     PUBLIC_DATASET_REVISION_URL,
     PROJECT_WEBSITE_URL,
-    RELEASE_BOUNDARY_NOTICE,
     SOURCE_REPOSITORY_URL,
 )
 from h2epr_explorer.data_loader import (
@@ -111,10 +110,9 @@ div[data-testid="stMetric"] {
 st.markdown('<div class="h2epr-kicker">H²EPR-Bench · public release explorer</div>', unsafe_allow_html=True)
 st.markdown('<div class="h2epr-title">Event-process graph browser</div>', unsafe_allow_html=True)
 st.markdown(
-    '<div class="h2epr-subtitle">Browse 3,000 public event records, Draft EPG summaries, and stage timelines from one validated Unified-3000 release.</div>',
+    '<div class="h2epr-subtitle">Search 3,000 real-world events, inspect their process structure, and open every public Draft EPG.</div>',
     unsafe_allow_html=True,
 )
-st.info(RELEASE_BOUNDARY_NOTICE)
 
 try:
     release = load_release()
@@ -224,7 +222,7 @@ event_links = build_event_links(selected_event)
 st.caption(filter_summary_text(len(filtered_rows), len(catalog_rows)))
 
 tabs = st.tabs(
-    ["Catalog", "Event detail", "Timeline", "Stages", "Draft EPG JSON", "Access and boundary"]
+    ["Catalog", "Event detail", "Timeline", "Stages", "Draft EPG JSON", "About"]
 )
 
 with tabs[0]:
@@ -309,15 +307,16 @@ with tabs[4]:
 with tabs[5]:
     st.markdown(
         f"""
-### Release boundary
+### About H²EPR-Bench
 
 - Project Website: [H²EPR-Bench]({PROJECT_WEBSITE_URL}).
 - Public Dataset: [`{PUBLIC_DATASET_REPO}`]({PUBLIC_DATASET_REVISION_URL}).
 - FinMycelium System: [multi-agent event reconstruction system]({FINMYCELIUM_URL}).
 - Reference EPGs (Gated): [`{GOLD_COMPANION_REPO}`]({GOLD_COMPANION_URL}).
 - Release Repository: [`AgenticFinLab/H2EPR-Bench`]({SOURCE_REPOSITORY_URL}).
-- Public Draft EPGs are sanitized FinMycelium construction artifacts, not scoring references.
-- Official benchmark scoring uses expert-adjudicated reference EPGs in the gated companion.
-- This Explorer loads neither reference EPGs nor frozen evidence packages.
+
+The Explorer presents FinMycelium Draft EPGs for browsing and method
+development. Official benchmark scoring uses the expert-finalized reference
+EPGs in the gated companion.
 """
     )

src/h2epr_explorer/constants.py

@@ -11,7 +11,7 @@ GOLD_COMPANION_URL = "https://huggingface.co/datasets/AgenticFinLab/H2EPR-Bench-
 SOURCE_REPOSITORY_URL = "https://github.com/AgenticFinLab/H2EPR-Bench"
 AGENTICFINLAB_URL = "https://agenticfinlab.github.io/"
 
-DEFAULT_PUBLIC_DATASET_REVISION: str | None = "6156a6bb3b838143401cb3e5709f708e5d6e802c"
+DEFAULT_PUBLIC_DATASET_REVISION: str | None = "4b0f0f4000db3ba9b6e1a720e5b5cfbaae68353c"
 PUBLIC_DATASET_REVISION = DEFAULT_PUBLIC_DATASET_REVISION
 PUBLIC_DATASET_REVISION_URL = (
     f"{PUBLIC_DATASET_URL}/tree/{PUBLIC_DATASET_REVISION}"
@@ -189,9 +189,3 @@ PROFILE_COLUMNS = (
     "gold_reference_access_level",
     "finalcascade_access_level",
 )
-
-RELEASE_BOUNDARY_NOTICE = (
-    "Public Draft EPGs are FinMycelium construction artifacts. Official benchmark "
-    "scoring uses reference EPGs in the manual-gated companion repository; this "
-    "Explorer loads neither reference EPGs nor frozen evidence packages."
-)

src/h2epr_explorer/data_loader.py

@@ -61,7 +61,7 @@ class DatasetRevisionUnavailable(DatasetTransportError):
 
 
 class ReleaseContractError(ExplorerDataError):
-    """The loaded files do not form the expected Unified-3000 release."""
+    """The loaded files do not form the expected H²EPR-Bench release."""
 
 
 class DraftAssetMissing(ExplorerDataError):
@@ -73,7 +73,7 @@ class DraftIntegrityError(ExplorerDataError):
 
 
 class InvalidEventId(ValueError):
-    """An identifier is not a canonical Unified-3000 event ID."""
+    """An identifier is not a canonical H²EPR-Bench event ID."""
 
 
 @dataclass
@@ -87,7 +87,7 @@ class ExplorerRelease:
         validate_event_id(event_id)
         rows = self.events.loc[self.events["event_id"].eq(event_id)]
         if rows.empty:
-            raise InvalidEventId(f"Unknown Unified-3000 event ID: {event_id}")
+            raise InvalidEventId(f"Unknown H²EPR-Bench event ID: {event_id}")
         if len(rows) != 1:
             raise ReleaseContractError(f"Duplicate event identity in joined view: {event_id}")
         return rows.iloc[0]
@@ -109,10 +109,10 @@ def _as_local_root(local_dataset_dir: Path | str | None = None) -> Path | None:
 
 def validate_event_id(event_id: str) -> str:
     if not isinstance(event_id, str) or not re.fullmatch(EVENT_ID_PATTERN, event_id):
-        raise InvalidEventId(f"Invalid Unified-3000 event ID: {event_id!r}")
+        raise InvalidEventId(f"Invalid H²EPR-Bench event ID: {event_id!r}")
     number = int(event_id.rsplit("-", 1)[1])
     if not EVENT_ID_MIN <= number <= EVENT_ID_MAX:
-        raise InvalidEventId(f"Unified-3000 event ID is out of range: {event_id}")
+        raise InvalidEventId(f"H²EPR-Bench event ID is out of range: {event_id}")
     return event_id
 
 
@@ -266,7 +266,7 @@ def _expected_event_ids() -> list[str]:
 def _require_exact_event_order(frame: pd.DataFrame, column: str, table_name: str) -> None:
     if frame[column].tolist() != _expected_event_ids():
         raise ReleaseContractError(
-            f"{table_name} does not contain the exact numeric Unified-3000 event order"
+            f"{table_name} does not contain the exact numeric H²EPR-Bench event order"
         )
 
 
@@ -346,7 +346,7 @@ def build_explorer_view(
     summary: pd.DataFrame,
     source_hashes: pd.DataFrame,
 ) -> pd.DataFrame:
-    """Build the exact one-row-per-event Unified-3000 Explorer view."""
+    """Build the exact one-row-per-event H²EPR-Bench Explorer view."""
 
     tables = (
         (gallery, EVENT_GALLERY_SCHEMA, "event_gallery", "public_event_id"),
@@ -451,7 +451,7 @@ def build_explorer_view(
             source_hashes, on="public_event_id", how="left", validate="one_to_one"
         )
     except Exception as exc:
-        raise ReleaseContractError("Unified-3000 Explorer join multiplicity failure") from exc
+        raise ReleaseContractError("H²EPR-Bench Explorer join multiplicity failure") from exc
 
     _require_row_count(joined, EXPECTED_EVENT_COUNT, "joined Explorer view")
     _require_no_nulls(joined, "joined Explorer view")