upgini 1.2.167a4220.dev8__tar.gz → 1.2.167a4220.dev9__tar.gz

{upgini-1.2.167a4220.dev8 → upgini-1.2.167a4220.dev9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: upgini
-Version: 1.2.167a4220.dev8
+Version: 1.2.167a4220.dev9
 Summary: Intelligent data search & enrichment for Machine Learning
 Project-URL: Bug Reports, https://github.com/upgini/upgini/issues
 Project-URL: Homepage, https://upgini.com/

upgini-1.2.167a4220.dev9/src/upgini/__about__.py

@@ -0,0 +1 @@
+__version__ = "1.2.167.a4220.dev9"

{upgini-1.2.167a4220.dev8 → upgini-1.2.167a4220.dev9}/src/upgini/features_enricher.py

@@ -64,6 +64,7 @@ from upgini.metadata import (
     FeaturesMetadataV2,
     FileColumnMeaningType,
     FileColumnMetadata,
+    GeneratedFeatureMetadata,
     ModelTaskType,
     RuntimeParameters,
     SearchKey,
@@ -98,7 +99,13 @@ from upgini.utils.deduplicate_utils import (
     clean_full_duplicates,
     remove_fintech_duplicates,
 )
-from upgini.report.assemble import assemble_report_data, autofe_rows_for_report, build_search_results
+from upgini.report.assemble import (
+    _generated_feature_names,
+    _parse_generated_feature,
+    assemble_report_data,
+    autofe_rows_from_description,
+    build_search_results,
+)
 from upgini.report.html import generate_html_report
 from upgini.report.stats import make_report_frame
 from upgini.utils.display_utils import (
@@ -2479,11 +2486,14 @@ class FeaturesEnricher(TransformerMixin):
             generated_names=self._column_name_aliases(
                 self.fit_generated_features or [], self.fit_columns_renaming or {}
             ),
+            generated_features=self._search_task.get_all_generated_features() if self._search_task else [],
+            joined_ads_features_count=(
+                self._search_task.get_joined_ads_features_count() if self._search_task else None
+            ),
+            joined_ads_count=self._search_task.get_joined_ads_count() if self._search_task else None,
         )
-        autofe = autofe_rows_for_report(
+        autofe = autofe_rows_from_description(
             self.get_autofe_features_description(features_meta=features_meta),
-            features_meta,
-            self._is_ensemble_feature,
             self.bundle,
         )
         return features, sources, model_shap, summary, autofe
@@ -2503,15 +2513,29 @@ class FeaturesEnricher(TransformerMixin):
             return False
         return bool(aliases & self._column_name_aliases(list(ensemble_names), renaming))
 
+    def _match_autofe_feature_meta(
+        self,
+        generated_meta: GeneratedFeatureMetadata,
+        features_meta: list[FeaturesMetadataV2],
+        used_names: set[str],
+    ) -> FeaturesMetadataV2 | None:
+        available = [meta for meta in features_meta or [] if meta.name not in used_names]
+        by_name = {meta.name: meta for meta in available}
+        for name in _generated_feature_names(generated_meta):
+            matched = by_name.get(name)
+            if matched is not None:
+                return matched
+        return None
+
     def _ensemble_generated_metadata(self):
         autofe_meta = self._search_task.get_autofe_metadata() if self._search_task else None
         if not autofe_meta:
             return []
         matched = []
         for meta in autofe_meta:
-            op = EnsembleModel.from_formula(meta.formula.split("(")[0])
-            if op is not None:
-                matched.append((meta, op))
+            feature = _parse_generated_feature(meta)
+            if feature is not None and isinstance(feature.op, EnsembleModel):
+                matched.append((meta, feature.op))
         return matched
 
     def _ensemble_feature_names_from_metadata(self) -> set[str]:
@@ -5864,6 +5888,7 @@ if response.status_code == 200:
 
         selected_features_meta.sort(key=lambda m: (-m.shap_value, m.name))
         psi_keepers = self._psi_stability_keepers() if self.psi_values is not None else set()
+        single_ensemble = len(self._ensemble_generated_metadata()) == 1
 
         for feature_meta in selected_features_meta:
             original_name = original_names_dict.get(feature_meta.name, feature_meta.name)
@@ -5873,6 +5898,10 @@ if response.status_code == 200:
             ) or original_name in generated_aliases
             is_client_feature = original_name in clients_features_df.columns and not is_generated_feature
 
+            # Model-input ads/AutoFE rows are only for the HTML report; the selected output is the score.
+            if single_ensemble and not is_client_feature and not self._is_ensemble_feature(feature_meta.name):
+                continue
+
             if not is_client_feature and not is_generated_feature:
                 self.external_source_feature_names.append(original_name)
 
@@ -5887,8 +5916,6 @@ if response.status_code == 200:
                 if original_name in self.unstable_features or feature_meta.name in self.unstable_features:
                     continue
 
-            # TODO make a decision about selected features based on special flag from mlb
-
             if original_shaps.get(feature_meta.name, 0.0) == 0.0 and not self._should_show_client_feature_in_report(
                 is_client_feature, feature_meta.name
             ):
@@ -5961,52 +5988,34 @@ if response.status_code == 200:
             elif features_meta is None:
                 features_meta = self._search_task.get_all_features_metadata_v2()
 
-            def get_feature_by_name(name: str):
-                if not name or not features_meta:
-                    return None
-                for candidate in features_meta:
-                    if candidate.name == name:
-                        return candidate
-                return None
-
             descriptions = []
+            used_names: set[str] = set()
             for m in autofe_meta:
-                orig_to_hashed = {base_column.original_name: base_column.hashed_name for base_column in m.base_columns}
-
-                autofe_feature = (
-                    Feature.from_formula(m.formula)
-                    .set_display_index(m.display_index)
-                    .set_alias(m.alias)
-                    .set_op_params(m.operator_params or {})
-                    .rename_columns(orig_to_hashed)
-                )
+                try:
+                    orig_to_hashed = {
+                        base_column.original_name: base_column.hashed_name for base_column in m.base_columns
+                    }
+                    autofe_feature = (
+                        Feature.from_formula(m.formula)
+                        .set_display_index(m.display_index or None)
+                        .set_alias(m.alias)
+                        .set_op_params(m.operator_params or {})
+                        .rename_columns(orig_to_hashed)
+                    )
+                except Exception:
+                    self.logger.exception(f"Failed to parse AutoFE formula: {m.formula}")
+                    continue
 
                 is_ts = isinstance(autofe_feature.op, TimeSeriesBase)
 
                 if autofe_feature.op.is_vector and not is_ts:
                     continue
 
-                description = {}
-
-                lookup_names = [autofe_feature.get_display_name(shorten=True, unhash=True)]
-                if m.alias:
-                    lookup_names.extend([m.alias, f"f_autofe_{m.alias}"])
-                feature_meta = None
-                for lookup_name in lookup_names:
-                    feature_meta = get_feature_by_name(lookup_name)
-                    if feature_meta is not None:
-                        break
-                if feature_meta is None and m.alias and features_meta:
-                    prefix = f"f_autofe_{m.alias}"
-                    matches = [
-                        candidate
-                        for candidate in features_meta
-                        if candidate.name == prefix or candidate.name.startswith(prefix + "_")
-                    ]
-                    if len(matches) == 1:
-                        feature_meta = matches[0]
+                feature_meta = self._match_autofe_feature_meta(m, features_meta, used_names)
                 if feature_meta is None or not feature_meta.shap_value:
                     continue
+
+                description = {}
                 description["shap"] = feature_meta.shap_value
                 source = feature_meta.data_source or ""
                 description[self.bundle.get("autofe_descriptions_sources")] = source.replace(
@@ -6028,6 +6037,7 @@ if response.status_code == 200:
                     sorted(autofe_feature.get_all_operand_names())
                 )
 
+                used_names.add(feature_meta.name)
                 descriptions.append(description)
 
             if len(descriptions) == 0:

{upgini-1.2.167a4220.dev8 → upgini-1.2.167a4220.dev9}/src/upgini/metadata.py

@@ -346,6 +346,8 @@ class ProviderTaskMetadataV2(BaseModel):
     zero_hit_rate_search_keys: Optional[List[str]] = None
     features_used_for_embeddings: Optional[List[str]] = None
     shuffle_kfold: Optional[bool] = None
+    joined_ads_features_count: Optional[int] = None
+    joined_ads_count: Optional[int] = None
     generated_features: Optional[List[GeneratedFeatureMetadata]] = None
 
 

{upgini-1.2.167a4220.dev8 → upgini-1.2.167a4220.dev9}/src/upgini/report/assemble.py

@@ -7,7 +7,9 @@ from typing import Callable, Optional
 import pandas as pd
 
 from upgini.__about__ import __version__
-from upgini.metadata import FeaturesMetadataV2
+from upgini.autofe.feature import Feature
+from upgini.autofe.vector import CatboostModel, EnsembleModel, OnnxModel
+from upgini.metadata import BaseColumnMetadata, FeaturesMetadataV2, GeneratedFeatureMetadata
 from upgini.report.data import (
     FeatureRow,
     ModelFeatureShap,
@@ -101,6 +103,9 @@ def build_search_results(
     features_meta: list[FeaturesMetadataV2],
     is_ensemble: Callable[[str], bool],
     generated_names: Optional[set[str]] = None,
+    generated_features: Optional[list[GeneratedFeatureMetadata]] = None,
+    joined_ads_features_count: Optional[int] = None,
+    joined_ads_count: Optional[int] = None,
 ) -> tuple[list[FeatureRow], list[SourceRow], list[ModelFeatureShap], SearchResultsSummary]:
     generated_names = generated_names or set()
     model_rows: list[FeatureRow] = []
@@ -123,44 +128,19 @@ def build_search_results(
         for index, row in enumerate(model_rows, start=1)
     ]
     sources = _source_rows(found_rows)
-    contributed = len({(row.provider, row.source) for row in found_rows if row.provider})
-    stable = sum(1 for row in model_rows if row.psi is not None and row.psi < _STABILITY_PSI)
-    summary = SearchResultsSummary(
-        relevant_features=len(found_rows),
-        model_features=len(model_rows) or None,
-        data_sources=len(sources),
-        stable_features_share=(stable / len(model_rows)) if model_rows else None,
-        contributed_sources=contributed,
-        stable_features=stable if model_rows else None,
+    summary = _overview_summary(
+        features_meta=features_meta,
+        model_rows=model_rows,
+        found_rows=found_rows,
+        sources=sources,
+        is_ensemble=is_ensemble,
+        generated_features=generated_features or [],
+        joined_ads_features_count=joined_ads_features_count,
+        joined_ads_count=joined_ads_count,
     )
     return model_rows, sources, model_shap, summary
 
 
-def autofe_rows_for_report(
-    df: Optional[pd.DataFrame],
-    features_meta: list[FeaturesMetadataV2],
-    is_ensemble: Callable[[str], bool],
-    bundle: ResourceBundle,
-) -> list[dict[str, str]]:
-    rows = autofe_rows_from_description(df, bundle)
-    listed = {row["generatedFeature"] for row in rows if row.get("generatedFeature")}
-    for meta in features_meta:
-        if is_ensemble(meta.name) or meta.source != "generated":
-            continue
-        if not _has_nonzero_shap_value(meta.shap_value) or meta.name in listed:
-            continue
-        row = _feature_row(meta.name, meta)
-        rows.append(
-            {
-                "sources": row.source,
-                "generatedFeature": row.name,
-                "sourceFeatures": "",
-                "functions": "",
-            }
-        )
-    return rows
-
-
 def autofe_rows_from_description(df: Optional[pd.DataFrame], bundle: ResourceBundle) -> list[dict[str, str]]:
     if df is None or df.empty:
         return []
@@ -275,11 +255,20 @@ def _feature_row(
         drift=meta.drift_score,
         coverage=meta.hit_rate,
         provider=info.internal_provider,
-        source=info.internal_source,
+        source=_feature_source(meta, is_generated, info.internal_source),
         stability_status=_psi_status(meta.psi_value),
     )
 
 
+def _feature_source(meta: FeaturesMetadataV2, is_generated: bool, fallback: str) -> str:
+    if is_generated:
+        if meta.data_source:
+            return meta.data_source
+        if meta.data_sources:
+            return ", ".join(meta.data_sources)
+    return fallback
+
+
 def _source_rows(rows: list[FeatureRow]) -> list[SourceRow]:
     grouped: dict[tuple[str, str], list[FeatureRow]] = {}
     for row in rows:
@@ -311,3 +300,136 @@ def _psi_status(psi: Optional[float]) -> Optional[str]:
 
 def _has_nonzero_shap_value(shap: Optional[float]) -> bool:
     return shap is not None and not pd.isna(shap) and shap != 0
+
+
+def _overview_summary(
+    *,
+    features_meta: list[FeaturesMetadataV2],
+    model_rows: list[FeatureRow],
+    found_rows: list[FeatureRow],
+    sources: list[SourceRow],
+    is_ensemble: Callable[[str], bool],
+    generated_features: list[GeneratedFeatureMetadata],
+    joined_ads_features_count: Optional[int],
+    joined_ads_count: Optional[int],
+) -> SearchResultsSummary:
+    top_level = _unique_fold_model_columns(generated_features)
+    features_by_name = {meta.name: meta for meta in features_meta}
+    autofe_by_name = _autofe_generated_by_name(generated_features, is_ensemble)
+    if top_level:
+        counts = {"external": 0, "original": 0, "autofe": 0}
+        ads_ids: set[str] = set()
+        psis: list[Optional[float]] = []
+        for column in top_level:
+            kind = _top_level_kind(column, features_by_name, autofe_by_name, is_ensemble)
+            counts[kind] += 1
+            ads_ids.update(_leaf_ads_ids(column, autofe_by_name))
+            meta = features_by_name.get(column.hashed_name) or features_by_name.get(column.original_name)
+            psis.append(None if meta is None else meta.psi_value)
+        model_features = len(top_level)
+        external, original, autofe = counts["external"], counts["original"], counts["autofe"]
+        contributed = len(ads_ids)
+    else:
+        model_features = len(model_rows) or None
+        external = original = autofe = None
+        contributed = len({(row.provider, row.source) for row in found_rows if row.provider})
+        psis = [row.psi for row in model_rows]
+    share, stable, psi_count = _stability_from_psi(psis)
+    return SearchResultsSummary(
+        relevant_features=len(found_rows),
+        features_found=joined_ads_features_count if joined_ads_features_count is not None else len(found_rows),
+        model_features=model_features,
+        external_features=external,
+        original_features=original,
+        autofe_features=autofe,
+        data_sources=len(sources),
+        joined_sources=joined_ads_count if joined_ads_count is not None else len(sources),
+        stable_features_share=share,
+        contributed_sources=contributed,
+        stable_features=stable,
+        psi_features=psi_count,
+    )
+
+
+def _unique_fold_model_columns(generated_features: list[GeneratedFeatureMetadata]) -> list[BaseColumnMetadata]:
+    unique: dict[str, BaseColumnMetadata] = {}
+    for meta in generated_features:
+        feature = _parse_generated_feature(meta)
+        if feature is None or not isinstance(feature.op, (CatboostModel, OnnxModel)):
+            continue
+        for column in meta.base_columns:
+            unique.setdefault(column.hashed_name or column.original_name, column)
+    return list(unique.values())
+
+
+def _autofe_generated_by_name(
+    generated_features: list[GeneratedFeatureMetadata], is_ensemble: Callable[[str], bool]
+) -> dict[str, GeneratedFeatureMetadata]:
+    by_name: dict[str, GeneratedFeatureMetadata] = {}
+    for meta in generated_features:
+        feature = _parse_generated_feature(meta)
+        if feature is None or isinstance(feature.op, (CatboostModel, EnsembleModel, OnnxModel)):
+            continue
+        name = feature.get_display_name(shorten=True, unhash=True, cache=False)
+        if name and not is_ensemble(name):
+            by_name.setdefault(name, meta)
+    return by_name
+
+
+def _parse_generated_feature(meta: GeneratedFeatureMetadata) -> Feature | None:
+    try:
+        parsed = Feature.from_formula(meta.formula)
+        if not isinstance(parsed, Feature):
+            return None
+        orig_to_hashed = {column.original_name: column.hashed_name for column in meta.base_columns}
+        return (
+            parsed.set_display_index(meta.display_index or None)
+            .set_alias(meta.alias)
+            .set_op_params(meta.operator_params or {})
+            .rename_columns(orig_to_hashed)
+        )
+    except Exception:
+        return None
+
+
+def _generated_feature_names(meta: GeneratedFeatureMetadata) -> set[str]:
+    feature = _parse_generated_feature(meta)
+    if feature is None:
+        return set()
+    return {feature.get_display_name(shorten=True, unhash=True, cache=False)}
+
+
+def _top_level_kind(
+    column: BaseColumnMetadata,
+    features_by_name: dict[str, FeaturesMetadataV2],
+    autofe_by_name: dict[str, GeneratedFeatureMetadata],
+    is_ensemble: Callable[[str], bool],
+) -> str:
+    names = [column.hashed_name, column.original_name]
+    if any(name in autofe_by_name for name in names if name):
+        return "autofe"
+    meta = next((features_by_name[name] for name in names if name in features_by_name), None)
+    if meta is not None and not is_ensemble(meta.name):
+        if meta.source == "generated":
+            return "autofe"
+        if meta.source == "ads":
+            return "external"
+        if meta.source == "etalon":
+            return "original"
+    if column.ads_definition_id:
+        return "external"
+    return "original"
+
+
+def _leaf_ads_ids(column: BaseColumnMetadata, autofe_by_name: dict[str, GeneratedFeatureMetadata]) -> set[str]:
+    autofe = autofe_by_name.get(column.hashed_name) or autofe_by_name.get(column.original_name)
+    leaves = autofe.base_columns if autofe is not None else [column]
+    return {leaf.ads_definition_id for leaf in leaves if leaf.ads_definition_id}
+
+
+def _stability_from_psi(psis: list[Optional[float]]) -> tuple[Optional[float], Optional[int], Optional[int]]:
+    values = [psi for psi in psis if psi is not None and not pd.isna(psi)]
+    if not values:
+        return None, None, None
+    stable = sum(1 for psi in values if psi < _STABILITY_PSI)
+    return stable / len(values), stable, len(values)

{upgini-1.2.167a4220.dev8 → upgini-1.2.167a4220.dev9}/src/upgini/report/data.py

@@ -66,11 +66,17 @@ class SourceRow:
 @dataclass
 class SearchResultsSummary:
     relevant_features: Optional[int] = None
+    features_found: Optional[int] = None
     model_features: Optional[int] = None
+    external_features: Optional[int] = None
+    original_features: Optional[int] = None
+    autofe_features: Optional[int] = None
     data_sources: Optional[int] = None
+    joined_sources: Optional[int] = None
     stable_features_share: Optional[float] = None
     contributed_sources: Optional[int] = None
     stable_features: Optional[int] = None
+    psi_features: Optional[int] = None
 
 
 @dataclass

{upgini-1.2.167a4220.dev8 → upgini-1.2.167a4220.dev9}/src/upgini/report/html.py

@@ -145,16 +145,30 @@ def _summary_cards(summary: SearchResultsSummary) -> list[dict]:
     if summary.contributed_sources is not None:
         sources_caption = f"{summary.contributed_sources} contributed"
     stability_caption = ""
-    if summary.stable_features is not None and summary.model_features is not None:
-        stability_caption = f"{summary.stable_features} of {summary.model_features} are stable"
+    if summary.stable_features is not None and summary.psi_features is not None:
+        stability_caption = f"{summary.stable_features} of {summary.psi_features} are stable"
+    features_found = summary.features_found if summary.features_found is not None else summary.relevant_features
+    data_sources = summary.joined_sources if summary.joined_sources is not None else summary.data_sources
     return [
-        {"label": "Features found", "value": _dash(summary.relevant_features), "caption": ""},
-        {"label": "Used in model", "value": _dash(summary.model_features), "caption": ""},
-        {"label": "Data sources", "value": _dash(summary.data_sources), "caption": sources_caption},
+        {"label": "Features found", "value": _dash(features_found), "caption": ""},
+        {"label": "Used in model", "value": _dash(summary.model_features), "caption": _used_in_model_caption(summary)},
+        {"label": "Data sources", "value": _dash(data_sources), "caption": sources_caption},
         {"label": "Stability", "value": _dash(stability), "caption": stability_caption},
     ]
 
 
+def _used_in_model_caption(summary: SearchResultsSummary) -> str:
+    if summary.external_features is None and summary.original_features is None and summary.autofe_features is None:
+        return ""
+    return " · ".join(
+        [
+            f"{summary.external_features or 0} external",
+            f"{summary.original_features or 0} original",
+            f"{summary.autofe_features or 0} AutoFE",
+        ]
+    )
+
+
 def _features_payload(features: list[FeatureRow]) -> list[dict]:
     return [
         {

{upgini-1.2.167a4220.dev8 → upgini-1.2.167a4220.dev9}/src/upgini/report/template.html

@@ -7,7 +7,7 @@
   <style>
     :root{--ink:#15211f;--muted:#6f7976;--line:#dce4e1;--paper:#fff;--bg:#f3f6f5;--blue:#1645ee;--green:#53cc61;--cyan:#20aab4;--warn:#e19132;--danger:#d45b5b;--shadow:0 18px 60px rgba(18,45,39,.08)}
     *{box-sizing:border-box}html{scroll-behavior:smooth}body{margin:0;background:var(--bg);color:var(--ink);font-family:Inter,ui-sans-serif,-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif;font-size:14px;line-height:1.45}.shell{max-width:1440px;margin:auto;min-height:100vh}.topbar{height:72px;padding:0 42px;display:flex;align-items:center;gap:25px;background:rgba(255,255,255,.92);backdrop-filter:blur(16px);border-bottom:1px solid var(--line);position:sticky;top:0;z-index:20}.brand{display:flex;align-items:center;gap:14px}.wordmark{font-size:25px;font-weight:800;letter-spacing:-.07em;line-height:1;background:linear-gradient(135deg,#0138f8,#5bd357);-webkit-background-clip:text;background-clip:text;color:transparent}.wordmark b{font-weight:850}.wordmark span{font-weight:650}.wordmark.small{font-size:20px;min-width:66px}.brand-divider{width:1px;height:25px;background:var(--line)}.brand-label{font-weight:700;color:#45514f}.top-meta{display:flex;gap:8px;margin-left:auto;align-items:center}.chip{background:#eef2f1;border:1px solid #e0e7e5;border-radius:999px;padding:7px 11px;color:#53605d;font-size:12px}.button{border:0;border-radius:10px;padding:10px 15px;font:inherit;font-weight:700;cursor:pointer;transition:.18s ease}.button:hover{transform:translateY(-1px)}.button.primary{background:var(--ink);color:white}.button.ghost{background:white;border:1px solid var(--line);color:var(--ink)}.layout{display:grid;grid-template-columns:210px minmax(0,1fr);gap:0}.sidenav{position:sticky;top:72px;height:calc(100vh - 72px);padding:30px 22px;border-right:1px solid var(--line);background:#f7f9f8}.nav-title{font-size:10px;letter-spacing:.14em;text-transform:uppercase;color:#98a29f;margin:0 10px 12px}.nav-link{display:flex;align-items:center;gap:10px;color:#66716f;text-decoration:none;padding:10px;border-radius:9px;margin:2px 0;font-weight:600}.nav-link:hover,.nav-link.active{background:white;color:var(--ink);box-shadow:0 4px 18px rgba(25,49,43,.06)}.nav-dot{width:7px;height:7px;border-radius:50%;background:#c4cecb}.nav-link.active .nav-dot{background:linear-gradient(135deg,var(--blue),var(--green))}.run-note{position:absolute;bottom:24px;left:22px;right:22px;padding:14px;background:white;border:1px solid var(--line);border-radius:12px;font-size:11px;color:var(--muted)}.run-note strong{display:block;color:var(--ink);font-size:12px;margin-bottom:4px}.main{min-width:0;padding:38px 44px 80px}.eyebrow{display:flex;align-items:center;gap:8px;text-transform:uppercase;letter-spacing:.12em;color:#74807d;font-size:11px;font-weight:800}.status-dot{width:8px;height:8px;border-radius:50%;background:var(--green);box-shadow:0 0 0 4px #e7f8e9}.hero-title{display:flex;justify-content:space-between;align-items:flex-end;gap:30px;margin:12px 0 28px}.hero-title h1{font-size:36px;line-height:1.08;letter-spacing:-.04em;margin:0;max-width:720px}.hero-title p{color:var(--muted);margin:10px 0 0;max-width:690px}.run-stamp{text-align:right;min-width:190px;color:var(--muted);font-size:12px}.run-stamp strong{display:block;color:var(--ink);font-size:14px;margin-bottom:3px}.metric-hero{position:relative;overflow:hidden;background:#172421;color:white;border-radius:22px;padding:28px 30px;box-shadow:var(--shadow);margin-bottom:18px}.metric-hero:after{content:"";position:absolute;width:400px;height:400px;right:-150px;top:-250px;border-radius:50%;background:radial-gradient(circle,rgba(71,215,117,.24),transparent 68%)}.metric-head{position:relative;z-index:1;display:flex;justify-content:space-between;gap:20px}.metric-tabs{display:flex;gap:5px;background:rgba(255,255,255,.08);border-radius:11px;padding:4px}.metric-tab{border:0;background:transparent;color:#aebbb7;padding:8px 12px;border-radius:8px;font-weight:700;cursor:pointer}.metric-tab.active{background:white;color:#13201e}.scores{position:relative;z-index:1;display:grid;grid-template-columns:1fr 1fr 1.2fr;gap:25px;margin-top:32px;align-items:end}.score-label{text-transform:uppercase;letter-spacing:.11em;color:#8fa09b;font-size:10px;font-weight:800}.score-value{font-size:39px;line-height:1;margin-top:8px;font-variant-numeric:tabular-nums;letter-spacing:-.04em}.uplift .score-value{color:#73e57b}.uplift p{color:#aebcb8;margin:8px 0 0;font-size:12px}.summary-grid{display:grid;grid-template-columns:repeat(4,1fr);gap:14px;margin:18px 0 34px}.summary{background:white;border:1px solid var(--line);border-radius:15px;padding:17px 18px}.summary .k{color:var(--muted);font-size:11px;text-transform:uppercase;letter-spacing:.09em;font-weight:800}.summary .v{font-size:23px;font-weight:750;margin-top:5px;letter-spacing:-.02em}.summary .d{font-size:11px;color:#88928f;margin-top:4px}.section{scroll-margin-top:94px;margin-top:36px}.section-head{display:flex;align-items:end;justify-content:space-between;gap:20px;margin-bottom:15px}.section-head h2{font-size:22px;letter-spacing:-.025em;margin:0}.section-head p{color:var(--muted);margin:5px 0 0;font-size:12px}.panel{background:white;border:1px solid var(--line);border-radius:17px;box-shadow:0 4px 20px rgba(18,45,39,.035)}.chart-panel{padding:20px}.chart-head{display:flex;justify-content:space-between;align-items:center;gap:16px;margin-bottom:10px}.legend{display:flex;gap:17px;flex-wrap:wrap;font-size:11px;color:var(--muted)}.legend i{display:inline-block;width:9px;height:9px;border-radius:50%;margin-right:6px}.control{border:1px solid var(--line);background:#fafcfb;color:var(--ink);border-radius:9px;padding:8px 10px;font:inherit;font-size:12px}.canvas-wrap{height:280px;position:relative}canvas{width:100%;height:100%}.callout{display:flex;gap:12px;align-items:flex-start;padding:14px 16px;background:#fff8ed;border:1px solid #f1dcc0;border-radius:13px;color:#6b4b20}.callout strong{display:block;color:#4c361a;margin-bottom:2px}.callout-icon{width:27px;height:27px;display:grid;place-items:center;background:#f4c77f;border-radius:8px;font-weight:900;color:#65410c;flex:none}.stability-grid{display:grid;grid-template-columns:minmax(0,1.5fr) minmax(260px,.5fr);gap:14px}.heatmap{overflow:auto;padding:16px}.heat-table{border-collapse:separate;border-spacing:5px;width:100%;font-size:11px}.heat-table th{color:var(--muted);font-weight:600;padding:2px 6px;white-space:nowrap}.heat-table th:first-child{text-align:left}.heat-table td:first-child{font-weight:700;padding-right:12px;white-space:nowrap}.heat{height:30px;min-width:47px;text-align:center;border-radius:6px;font-weight:750}.s1{background:#dbf4de;color:#276b2f}.s2{background:#e9f4cf;color:#526826}.s3{background:#fff0c7;color:#7f641b}.s4{background:#f8d5c8;color:#914634}.stability-side{padding:20px}.donut{width:140px;height:140px;border-radius:50%;margin:8px auto 17px;background:conic-gradient(var(--green) 0 72%,#e6b556 72% 91%,#df7474 91%);position:relative}.donut:after{content:"";position:absolute;inset:25px;background:white;border-radius:50%}.donut-label{position:absolute;inset:0;z-index:1;display:grid;place-items:center;text-align:center;font-weight:800;font-size:21px}.donut-label small{display:block;color:var(--muted);font-size:10px;font-weight:600}.stat-list{display:grid;gap:8px}.stat-row{display:flex;justify-content:space-between;padding-top:8px;border-top:1px solid #edf1ef;color:var(--muted);font-size:12px}.stat-row b{color:var(--ink)}.toolbar{display:flex;gap:9px;flex-wrap:wrap}.search{min-width:230px;border:1px solid var(--line);background:white;border-radius:9px;padding:9px 11px;font:inherit}.table-wrap{overflow:auto}.data-table{width:100%;border-collapse:collapse;min-width:860px}.data-table th{text-align:left;padding:12px 14px;background:#f8faf9;color:#77827f;text-transform:uppercase;letter-spacing:.07em;font-size:10px;border-bottom:1px solid var(--line)}.data-table td{padding:13px 14px;border-bottom:1px solid #edf1ef;font-size:12px}.data-table tr:last-child td{border-bottom:0}.data-table tbody tr:hover{background:#f8fbfa}.feature-name{font-weight:750;color:#25312f}.source-name{color:#74807d}.bar{height:6px;width:90px;background:#edf1ef;border-radius:99px;overflow:hidden}.bar i{height:100%;display:block;background:linear-gradient(90deg,var(--blue),#43c878);border-radius:99px}.tag{display:inline-flex;align-items:center;border-radius:999px;padding:4px 8px;font-size:10px;font-weight:750}.tag.good{background:#e8f7ea;color:#337139}.tag.watch{background:#fff1d9;color:#855d18}.tag.risk{background:#fde7e4;color:#984a40}.source-cards{display:grid;grid-template-columns:repeat(3,1fr);gap:14px}.source-card{padding:19px}.source-top{display:flex;justify-content:space-between;gap:12px}.source-icon{width:38px;height:38px;border-radius:11px;display:grid;place-items:center;background:linear-gradient(135deg,#e8eeff,#e8f9e8);font-weight:850;color:var(--blue)}.source-card h3{font-size:14px;margin:12px 0 4px}.source-card p{font-size:11px;color:var(--muted);margin:0}.source-metrics{display:flex;gap:18px;margin-top:17px;padding-top:13px;border-top:1px solid #edf1ef}.source-metrics b{display:block;font-size:16px}.source-metrics span{color:var(--muted);font-size:10px}.notebook{padding:0;overflow:hidden}.cell-head{background:#f8faf9;border-bottom:1px solid var(--line);padding:10px 14px;font-family:ui-monospace,SFMono-Regular,Menlo,monospace;color:#68736f;font-size:11px}.cell-output{padding:18px;display:flex;align-items:center;gap:25px;flex-wrap:wrap}.cell-quality{display:flex;gap:24px}.cell-quality span{font-size:11px;color:var(--muted)}.cell-quality b{font-size:17px;color:var(--ink);display:block}.footer{text-align:center;color:#8a9592;font-size:11px;padding:40px 0 10px}.empty{display:none;padding:24px;text-align:center;color:var(--muted)}
-    .sample-table{min-width:700px}.toggle-legend{margin-bottom:10px}.series-toggle{border:0;background:transparent;color:var(--muted);cursor:pointer;font:inherit;font-size:11px;padding:5px 7px;border-radius:6px}.series-toggle:not(.active){opacity:.35;text-decoration:line-through}.threshold-legend{display:flex;gap:8px;justify-content:flex-end;margin-bottom:10px}.shap-panel{padding:16px}.shap-row{display:grid;grid-template-columns:minmax(170px,1fr) 2fr 60px;gap:12px;align-items:center;padding:9px 4px;border-bottom:1px solid #edf1ef}.shap-row:last-child{border:0}.provider{font-size:9px;margin-left:6px;padding:3px 6px;border-radius:99px;border:1px solid var(--line);color:var(--muted)}.shap-track{height:10px;background:#edf1ef;border-radius:99px;overflow:hidden}.shap-fill{display:block;height:100%;border-radius:99px}.shap-fill.upgini{background:#1645ee}.shap-fill.user{background:#20aab4}.shap-fill.autofe{background:#53cc61}.shap-fill.external{background:#7b5cff}.result-tabs{display:flex;gap:4px;margin-bottom:12px}.result-tab{border:1px solid var(--line);background:#eef2f1;padding:9px 13px;border-radius:9px;font:inherit;cursor:pointer;color:var(--muted)}.result-tab.active{background:var(--ink);color:#fff}.result-tab b{font-size:10px;margin-left:5px}.result-pane{display:none}.result-pane.active{display:block}.table-actions{display:flex;gap:8px;justify-content:flex-end;flex-wrap:wrap;margin-bottom:10px}.chart-tooltip{position:absolute;display:none;pointer-events:none;background:#172421;color:white;padding:9px 11px;border-radius:8px;font-size:10px;box-shadow:var(--shadow);white-space:nowrap}.sidenav{overflow:auto}.run-note{position:static;margin-top:22px}.nav-link{font-size:12px;padding:8px 10px}
+    .sample-table{min-width:700px}.toggle-legend{margin-bottom:10px}.series-toggle{border:0;background:transparent;color:var(--muted);cursor:pointer;font:inherit;font-size:11px;padding:5px 7px;border-radius:6px}.series-toggle:not(.active){opacity:.35;text-decoration:line-through}.threshold-legend{display:flex;gap:8px;justify-content:flex-end;margin-bottom:10px}.shap-panel{padding:16px}.shap-row{display:grid;grid-template-columns:minmax(170px,1fr) 2fr 60px;gap:12px;align-items:center;padding:9px 4px;border-bottom:1px solid #edf1ef}.shap-row:last-child{border:0}.provider{display:inline-block;font-size:9px;margin-left:6px;padding:3px 7px;border-radius:8px;border:1px solid var(--line);color:var(--muted);line-height:1.3;text-align:center;vertical-align:middle;white-space:normal}.shap-track{height:10px;background:#edf1ef;border-radius:99px;overflow:hidden}.shap-fill{display:block;height:100%;border-radius:99px}.shap-fill.upgini{background:#1645ee}.shap-fill.user{background:#20aab4}.shap-fill.autofe{background:#53cc61}.shap-fill.external{background:#7b5cff}.result-tabs{display:flex;gap:4px;margin-bottom:12px}.result-tab{border:1px solid var(--line);background:#eef2f1;padding:9px 13px;border-radius:9px;font:inherit;cursor:pointer;color:var(--muted)}.result-tab.active{background:var(--ink);color:#fff}.result-tab b{font-size:10px;margin-left:5px}.result-pane{display:none}.result-pane.active{display:block}.table-actions{display:flex;gap:8px;justify-content:flex-end;flex-wrap:wrap;margin-bottom:10px}.chart-tooltip{position:absolute;display:none;pointer-events:none;background:#172421;color:white;padding:9px 11px;border-radius:8px;font-size:10px;box-shadow:var(--shadow);white-space:nowrap}.sidenav{overflow:auto}.run-note{position:static;margin-top:22px}.nav-link{font-size:12px;padding:8px 10px}
     .metric-sample-table{position:relative;z-index:1;margin-top:26px}.metric-sample-head,.metric-sample-row{display:grid;grid-template-columns:minmax(140px,.9fr) 1fr 1fr .8fr .8fr;gap:18px;align-items:center}.metric-sample-ci{display:block;margin-top:6px;color:#8fa09b;font-size:11px;font-weight:600;font-variant-numeric:tabular-nums}.metric-note{color:#8fa09b;font-size:11px;margin-top:6px}.partner-logo{height:30px;width:auto;display:block}.metric-sample-head{padding:0 14px 8px;color:#8fa09b;text-transform:uppercase;letter-spacing:.11em;font-size:10px;font-weight:800}.metric-sample-row{padding:15px 14px;border-top:1px solid rgba(255,255,255,.12)}.metric-sample-name{display:flex;align-items:center;gap:10px;font-size:16px;font-weight:760}.metric-sample-name i{width:9px;height:9px;border-radius:50%;background:var(--blue);box-shadow:0 0 0 4px rgba(22,69,238,.18)}.metric-sample-row[data-sample="eval_1"] .metric-sample-name i{background:var(--cyan);box-shadow:0 0 0 4px rgba(32,170,180,.18)}.metric-sample-row[data-sample="eval_2"] .metric-sample-name i{background:var(--green);box-shadow:0 0 0 4px rgba(83,204,97,.18)}.metric-sample-name small{display:block;color:#8fa09b;font-size:10px;font-weight:650;margin-top:2px}.metric-sample-value{font-size:27px;line-height:1;font-variant-numeric:tabular-nums;letter-spacing:-.035em}.metric-sample-uplift .metric-sample-value{color:#73e57b}.metric-sample-uplift p{margin:5px 0 0;color:#aebcb8;font-size:11px}.tall-chart{height:340px}.chart-tooltip b{font-size:11px}.chart-tooltip .tip-label{color:#aebbb7}.canvas-wrap canvas{cursor:crosshair}.search-id-chip{font-family:ui-monospace,SFMono-Regular,Menlo,monospace;max-width:360px;overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
     @media(max-width:980px){.layout{grid-template-columns:1fr}.sidenav{display:none}.main{padding:28px 22px 70px}.topbar{padding:0 22px}.scores{grid-template-columns:1fr 1fr}.uplift{grid-column:1/-1}.summary-grid{grid-template-columns:1fr 1fr}.stability-grid,.sample-grid{grid-template-columns:1fr}.source-cards{grid-template-columns:1fr 1fr}}
     @media(max-width:650px){.top-meta .chip,.brand-divider,.brand-label{display:none}.topbar{gap:10px}.button{padding:9px 10px}.hero-title{display:block}.run-stamp{text-align:left;margin-top:16px}.hero-title h1{font-size:30px}.metric-head{display:block}.metric-tabs{margin-top:15px;width:max-content}.summary-grid,.source-cards{grid-template-columns:1fr}.scores{gap:14px}.score-value{font-size:32px}.metric-sample-head{display:none}.metric-sample-row{grid-template-columns:1fr 1fr;gap:14px;padding:16px 2px}.metric-sample-name{grid-column:1/-1}.metric-sample-value{font-size:24px}.metric-sample-row>div:not(.metric-sample-name):before{display:block;color:#8fa09b;text-transform:uppercase;letter-spacing:.08em;font-size:8px;font-weight:800;margin-bottom:6px}.metric-sample-row>div:nth-child(2):before{content:"Baseline"}.metric-sample-row>div:nth-child(3):before{content:"Enriched"}.metric-sample-row>div:nth-child(4):before{content:"Uplift, abs"}.metric-sample-row>div:nth-child(5):before{content:"Uplift, %"}.main{padding-left:15px;padding-right:15px}.cell-output{gap:14px}}
@@ -127,6 +127,9 @@ const REPORT_DATA = __REPORT_DATA__;
   const pct1 = v => (v == null ? '—' : Number(v).toFixed(1) + '%');
   const signed3 = v => (v >= 0 ? '+' : '') + Number(v).toFixed(3);
   const setText = (id, text) => { const el = document.getElementById(id); if (el) el.textContent = text; };
+  const providerBadge = provider => provider
+    ? `<span class="provider">${String(provider).replace(/,\s*/g, '<br>')}</span>`
+    : '';
   function rgba(hex, alpha) {
     const h = hex.replace('#', '');
     const n = parseInt(h.length === 3 ? h.split('').map(x => x + x).join('') : h, 16);
@@ -507,7 +510,7 @@ const REPORT_DATA = __REPORT_DATA__;
     const maxImportance = Math.max(...features.map(f => f.importance)) || 1;
     const list = shapAll ? features : features.slice(0, (D.shap && D.shap.topN) || features.length);
     document.getElementById('shapList').innerHTML = list.map((f, i) =>
-      `<div class="shap-row"><div><b>#${i + 1} ${f.name}</b>${f.provider ? `<span class="provider">${f.provider}</span>` : ''}</div>` +
+      `<div class="shap-row"><div><b>#${i + 1} ${f.name}</b>${providerBadge(f.provider)}</div>` +
       `<div class="shap-track"><i class="shap-fill ${f.shapClass || 'upgini'}" style="width:${(f.importance / maxImportance * 100).toFixed(2)}%"></i></div>` +
       `<b>${fmt3(Math.abs(f.shap))}</b></div>`).join('');
     document.getElementById('showAllShap').textContent = shapAll ? `Show top ${(D.shap && D.shap.topN) || features.length}` : `Show all ${features.length}`;
@@ -534,7 +537,7 @@ const REPORT_DATA = __REPORT_DATA__;
     const body = document.getElementById('featureBody');
     body.innerHTML = rows.map(f =>
       `<tr data-name="${f.name}"><td><span class="feature-name" title="${f.name}">${f.name}</span></td>` +
-      `<td>${f.provider ? `<span class="provider">${f.provider}</span>` : ''}</td><td class="source-name">${f.source}</td>` +
+      `<td>${providerBadge(f.provider)}</td><td class="source-name">${f.source}</td>` +
       `<td>${shapLabel(f.shap)}</td><td><span class="tag ${f.status}">${fmt2(f.psi)}</span></td>` +
       `<td>${f.drift == null ? '—' : fmt2(f.drift)}</td><td>${pct1(f.coverage)}</td></tr>`).join('');
     document.getElementById('emptyState').style.display = rows.length ? 'none' : 'block';

{upgini-1.2.167a4220.dev8 → upgini-1.2.167a4220.dev9}/src/upgini/search_task.py

@@ -210,6 +210,31 @@ class SearchTask:
             if meta.generated_features is not None:
                 return meta.generated_features
 
+    def get_all_generated_features(self) -> List[GeneratedFeatureMetadata]:
+        if self.provider_metadata_v2 is None:
+            return []
+        generated: List[GeneratedFeatureMetadata] = []
+        for meta in self.provider_metadata_v2:
+            if meta.generated_features:
+                generated.extend(meta.generated_features)
+        return generated
+
+    def get_joined_ads_features_count(self) -> Optional[int]:
+        if self.provider_metadata_v2 is None:
+            return None
+        for meta in self.provider_metadata_v2:
+            if meta.joined_ads_features_count is not None:
+                return meta.joined_ads_features_count
+        return None
+
+    def get_joined_ads_count(self) -> Optional[int]:
+        if self.provider_metadata_v2 is None:
+            return None
+        for meta in self.provider_metadata_v2:
+            if meta.joined_ads_count is not None:
+                return meta.joined_ads_count
+        return None
+
     @staticmethod
     def _get_provider_summaries(summary: SearchTaskSummary) -> List[ProviderTaskSummary]:
         if summary.status in {

upgini-1.2.167a4220.dev8/src/upgini/__about__.py

@@ -1 +0,0 @@
-__version__ = "1.2.167.a4220.dev8"