Create enhanced AudioToTextEnhancedProvider that also reformats text from transcription#363
Open
lukasdotcom wants to merge 6 commits intomainfrom
Open
Create enhanced AudioToTextEnhancedProvider that also reformats text from transcription#363lukasdotcom wants to merge 6 commits intomainfrom
lukasdotcom wants to merge 6 commits intomainfrom
Conversation
…g handler Signed-off-by: Lukas Schaefer <lukas@lschaefer.xyz>
marcelklehr
reviewed
May 7, 2026
marcelklehr
reviewed
May 7, 2026
marcelklehr
reviewed
May 7, 2026
There was a problem hiding this comment.
Pull request overview
This PR extends the app’s Task Processing providers by introducing a paragraph-reformatting text provider and an “enhanced” audio-to-text provider that chains transcription → paragraph reformatting (when the Nextcloud task type is available).
Changes:
- Add
ReformatParagraphsProvider(core:text2text:reformatparagraphs) that inserts paragraph breaks based on LLM-returned “anchor” lines. - Add
AudioToTextEnhancedProviderthat runsAudioToTextProviderand then invokes the reformat-paragraphs task as a follow-up step. - Register the new providers conditionally based on task type availability and add a unit test + psalm reference for the new task type class.
Reviewed changes
Copilot reviewed 6 out of 6 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
lib/TaskProcessing/ReformatParagraphsProvider.php |
New synchronous provider that requests anchor lines from the LLM and inserts paragraph breaks into the original input text. |
lib/TaskProcessing/AudioToTextEnhancedProvider.php |
New provider extending AudioToTextProvider and post-processing transcription via task processing manager. |
lib/TaskProcessing/AudioToTextProvider.php |
Makes dependencies protected to enable subclass access from the enhanced provider. |
lib/AppInfo/Application.php |
Registers the enhanced audio provider and the new reformat provider when the task type exists. |
tests/unit/Providers/OpenAiProviderTest.php |
Adds a unit test for ReformatParagraphsProvider and gates it by task type availability. |
psalm.xml |
Adds TextToTextReformatParagraphs as a referenced class for analysis. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| if (isset($input['model']) && is_string($input['model'])) { | ||
| $model = $input['model']; | ||
| } else { | ||
| $model = $this->appConfig->getValueString(Application::APP_ID, 'default_completion_model_id', Application::DEFAULT_MODEL_ID, lazy: true) ?: Application::DEFAULT_MODEL_ID; |
Member
Author
There was a problem hiding this comment.
This could be changed in the other providers too
Signed-off-by: Lukas Schaefer <lukas@lschaefer.xyz>
Signed-off-by: Lukas Schaefer <lukas@lschaefer.xyz>
…from transcription Signed-off-by: Lukas Schaefer <lukas@lschaefer.xyz> # Conflicts: # lib/TaskProcessing/ReformatParagraphsProvider.php
lukasdotcom
force-pushed
the
better-transcript
branch
2 times, most recently
from
df57549 to
fb23749
Compare
Signed-off-by: Lukas Schaefer <lukas@lschaefer.xyz>
lukasdotcom
force-pushed
the
better-transcript
branch
from
fb23749 to
a85993e
Compare
marcelklehr
reviewed
May 11, 2026
| } | ||
|
|
||
| public function getOptionalOutputShape(): array { | ||
| return $this->audioToTextProvider->getOptionalOutputShape(); |
Member
There was a problem hiding this comment.
Can we hardcode these? Otherwise changes in the audioToTextProvider will automatically change this provider as well, which feels like to much magic.
Member
Author
There was a problem hiding this comment.
Some of them it makes sense to change like for the output ones, but I would say the input ones should not be changed because they are passed into the audioToTextProvider directly anyway and should always be the same.
Member
There was a problem hiding this comment.
Alright, that makes sense, yes.
marcelklehr
reviewed
May 11, 2026
| . 'A break is allowed only when the subject matter changes significantly. ' | ||
| . 'Output format: For each identified paragraph, return only the first 8 to 12 words verbatim from the input. ' | ||
| . 'Structure: Return exactly one paragraph per line. Do not include bullets, html tags, numbering, summaries, quotes, or any additional text. ' | ||
| . 'Single topic: If the text covers only one topic, return exactly one line.'; |
Member
There was a problem hiding this comment.
Tested with output from stt_whisper2 and Mistral Small 24B from IONOS and it doesn't seem to work for me :/
Member
There was a problem hiding this comment.
I get the complete blob back from the task type, same as the input
Member
Author
There was a problem hiding this comment.
Interesting what prompt did you use. I used the one below and tested it on Mistral Small 24B and it splits it into two paragraphs:
Details
Dogs have been humanity's most devoted companions for thousands of years, evolving from wild wolves into loyal members of our families. Renowned for their incredible intelligence, adaptability, and emotional connection, they serve in countless roles ranging from herding flocks and searching for disaster victims to providing comfort to those with anxiety. Whether they are playful puppies chasing a ball, energetic adults running off-leash, or quiet guardians watching over the home, dogs bring a unique blend of energy, affection, and unconditional love into our lives. Their ability to read human body language and emotions has fostered a deep bond between species, making them not just pets, but true partners in our daily journeys. Cats are captivating creatures that have held a unique place in human history for millennia, revered both as loyal companions and mysterious symbols of grace. With their sleek, muscular bodies and agile movements, they are masters of the hunt, capable of pouncing with pinpoint accuracy even in the dimmest light. Beyond their physical prowess, cats possess an enigmatic personality that ranges from aloof independence to devoted affection, often choosing their owners with a quiet intuition. Whether lounging lazily in a sunbeam, watching the world with wide, golden eyes, or performing a silent, graceful leap, cats embody a blend of elegance and mystery that continues to enchant people everywhere.
Member
There was a problem hiding this comment.
I tried with the transcript of the last company call from may 5th
Signed-off-by: Lukas Schaefer <lukas@lschaefer.xyz>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Built off of #362 and creates an AudioToTextEnhancedProvider that can be used instead that also reformats text. (582825f is the important commit)