Avoid collisions in common CSV names, and allow instantiating deployment+recording metadata directly from a dataframe.

PiperOrigin-RevId: 876277128

Author: sdenton4

perch_hoplite/agile/README.md

@@ -45,22 +45,22 @@ directory:
         `int`, and `bytes`.
     *   `description`: An optional description of the field.
 
-2.  **`deployments_metadata.csv`**: This file contains metadata for each
+2.  **`hoplite_deployments_metadata.csv`**: This file contains metadata for each
     deployment. The first column must be the deployment identifier (which
     corresponds to the directory name if audio files are in
     `deployment/recording.wav`
     format), and subsequent columns should match `field_name`s from
     `metadata_description.csv` where `metadata_level` is `deployment`.
 
-3.  **`recordings_metadata.csv`**: This file contains metadata for each
+3.  **`hoplite_recordings_metadata.csv`**: This file contains metadata for each
     recording. The first column must be the recording identifier (e.g.
     `deployment/recording.wav`),
     and subsequent columns should match `field_name`s from
     `metadata_description.csv` where `metadata_level` is `recording`.
 
 ### Example
 
-**`metadata_description.csv`**
+**`hoplite_metadata_description.csv`**
 
 ```csv
 field_name,metadata_level,type,description
@@ -71,15 +71,15 @@ file_id,recording,str,Recording identifier.
 mic_type,recording,str,Microphone type.
 ```
 
-**`deployments_metadata.csv`**
+**`hoplite_deployments_metadata.csv`**
 
 ```csv
 deployment_name,habitat,latitude
 DEP01,"forest",47.6
 DEP02,"grassland",45.1
 ```
 
-**`recordings_metadata.csv`**
+**`hoplite_recordings_metadata.csv`**
 
 ```csv
 file_id,mic_type
@@ -98,7 +98,8 @@ and `Recording` objects returned by the database interface (e.g.,
 
 If you have existing annotations for your audio data, Hoplite can ingest these
 during the embedding process. Annotations should be stored in CSV files named
-`annotations.csv` alongside your audio data. Each `annotations.csv` should
+`hoplite_annotations.csv` alongside your audio data. Each
+`hoplite_annotations.csv` should
 contain columns for `recording` (the filename or file_id of the audio),
 `start_offset_s`, `end_offset_s`, `label`, and `label_type` ('positive',
 'negative', or 'uncertain'). When embeddings are generated, Hoplite will find
@@ -107,7 +108,7 @@ appropriate time windows.
 
 ### Example
 
-**`annotations.csv`**
+**`hoplite_annotations.csv`**
 
 ```csv
 recording,start_offset_s,end_offset_s,label,label_type

perch_hoplite/agile/embed.py

@@ -220,6 +220,7 @@ def _get_or_insert_deployment_id(
     )
     if not deployments:
       md = self.metadata[project_name].get_deployment_metadata(deployment_name)
+      md.pop('deployment', None)
       return self.db.insert_deployment(
           name=deployment_name,
           project=project_name,
@@ -243,6 +244,7 @@ def _get_or_insert_recording_id(
     )
     if not recordings:
       md = self.metadata[dataset_name].get_recording_metadata(filename)
+      md.pop('recording', None)
       return (
           self.db.insert_recording(
               filename=filename,
@@ -296,7 +298,7 @@ def add_annotations(self, target_dataset_name: str | None = None):
       if not agile_md.annotations:
         continue
       for file_id, annotation_list in agile_md.annotations.items():
-        depl_name = os.path.split(file_id)[0]
+        depl_name = file_id.split('/')[0]
         if not depl_name:
           logging.warning(
               'Could not get deployment name from file_id %s, skipping.',