openrunner-sdk 2.1.0__tar.gz → 2.2.0__tar.gz

{openrunner_sdk-2.1.0 → openrunner_sdk-2.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: openrunner-sdk
-Version: 2.1.0
+Version: 2.2.0
 Summary: OpenRunner SDK - W&B-compatible ML experiment tracking client
 Project-URL: Homepage, https://github.com/jqueguiner/openrunner
 Project-URL: Repository, https://github.com/jqueguiner/openrunner

{openrunner_sdk-2.1.0 → openrunner_sdk-2.2.0}/openrunner/media.py

@@ -310,9 +310,99 @@ class Table:
         data = df.values.tolist()
         return cls(columns=columns, data=data)
 
+    def set_reference_column(self, column: str = "sample_id") -> None:
+        """Mark a column as the dataset reference/sample ID.
+
+        This tells the UI to treat this column as a link back to the
+        original dataset, enabling sample traceability.
+
+        Args:
+            column: Name of the column containing sample IDs/references.
+                    Must already exist in the table.
+        """
+        if column not in self.columns:
+            raise ValueError(f"Column '{column}' not in table. Columns: {self.columns}")
+        self._reference_column = column
+
+    @classmethod
+    def from_predictions(
+        cls,
+        *,
+        sample_ids: list[str],
+        references: list[str],
+        hypotheses: list[str],
+        audio_paths: list[str] | None = None,
+        metadata: dict[str, list] | None = None,
+        extra_columns: dict[str, list] | None = None,
+    ) -> "Table":
+        """Create a predictions table with sample traceability.
+
+        Convenience constructor for evaluation results. Ensures consistent
+        column naming (sample_id, ref, hyp, audio) that the UI recognizes
+        for diff highlighting and audio playback.
+
+        Args:
+            sample_ids: Unique IDs linking each row to the source dataset
+            references: Ground truth transcriptions
+            hypotheses: Model predictions
+            audio_paths: Optional audio file paths for inline playback
+            metadata: Optional per-sample metadata dict (e.g. {"speaker": [...], "duration": [...], "language": [...]})
+            extra_columns: Alias for metadata (deprecated, use metadata)
+
+        Example:
+            table = Table.from_predictions(
+                sample_ids=["file_001", "file_002"],
+                references=["hello world", "press one"],
+                hypotheses=["hello word", "press 1"],
+                audio_paths=["audio/001.wav", "audio/002.wav"],
+                metadata={
+                    "duration_s": [3.2, 1.8],
+                    "speaker": ["agent", "ivr"],
+                    "language": ["en", "es"],
+                    "source_file": ["call_123.wav", "call_456.wav"],
+                },
+            )
+            openrunner.log({"predictions": table})
+        """
+        columns = ["sample_id"]
+        n = len(sample_ids)
+        data: list[list[Any]] = [[sid] for sid in sample_ids]
+
+        if audio_paths:
+            columns.append("audio")
+            for i, p in enumerate(audio_paths):
+                data[i].append(p)
+
+        columns.append("hyp")
+        for i, h in enumerate(hypotheses):
+            data[i].append(h)
+
+        columns.append("ref")
+        for i, r in enumerate(references):
+            data[i].append(r)
+
+        # Merge metadata and extra_columns
+        all_extra = {}
+        if metadata:
+            all_extra.update(metadata)
+        if extra_columns:
+            all_extra.update(extra_columns)
+
+        for col_name, col_data in all_extra.items():
+            columns.append(col_name)
+            for i, v in enumerate(col_data[:n]):
+                data[i].append(v)
+
+        t = cls(columns=columns, data=data)
+        t._reference_column = "sample_id"
+        return t
+
     def _serialize(self) -> dict[str, Any]:
         """Serialize to a dict for JSON transmission."""
-        return {"columns": self.columns, "data": self._data}
+        result: dict[str, Any] = {"columns": self.columns, "data": self._data}
+        if hasattr(self, "_reference_column"):
+            result["reference_column"] = self._reference_column
+        return result
 
 
 class Audio:

{openrunner_sdk-2.1.0 → openrunner_sdk-2.2.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "openrunner-sdk"
-version = "2.1.0"
+version = "2.2.0"
 description = "OpenRunner SDK - W&B-compatible ML experiment tracking client"
 readme = "README.md"
 license = {text = "MIT"}