whisper: Improve support for local custom Whisper models

whisper/CHANGELOG.md

@@ -1,5 +1,10 @@
 # Changelog
 
+## 2.5.0
+
+- Added configuration mapping to access local models.
+- Updated documentation about `custom_model` usage.
+
 ## 2.4.0
 
 - Add "auto" for model and beam size (0) to select values based on CPU

whisper/DOCS.md

@@ -67,6 +67,10 @@ Available models:
 ### Option: `custom_model`
 
 Path to a converted model directory, or a CTranslate2-converted Whisper model ID from the HuggingFace Hub like "Systran/faster-distil-whisper-small.en". 
+To use a local custom Whisper model, first create a `models` subdirectory in the add-on's configuration directory if it does not already exist. Then copy your model directory into:
+`/addon_configs/core_whisper/models/<your-model-dir>`.
+Then, set the `custom_model` path to:
+`/config/models/<your-model-dir>`. For a local model, the path must start with `/config/models/`, as this is how the add-on accesses your Home Assistant configuration directory through the container's mounted volume.
 
 ### Option: `beam_size`
 
@@ -82,7 +86,8 @@ See [this discussion](https://github.com/openai/whisper/discussions/963) for an
 
 ## Backups
 
-Whisper model files can be quite large, so they are automatically excluded from backups. The models will be re-downloaded when the backup is restored.
+Whisper model files can be large, so they are automatically excluded from backups and re-downloaded on restore for remote models.
+After restoring a backup with a local custom Whisper model, manually copy your model directory again.
 
 ## Support
 

whisper/config.yaml

@@ -1,5 +1,5 @@
 ---
-version: 2.4.0
+version: 2.5.0
 slug: whisper
 name: Whisper
 description: Speech-to-text with Whisper
@@ -12,6 +12,8 @@ discovery:
   - wyoming
 backup_exclude:
   - "models*"
+map:
+  - addon_config:rw
 options:
   model: auto
   language: en