Add ui for low quality, max threshold more lax, enhance pcm before se… #340
Conversation
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
|
Resolves #334 |
| - Preserve natural phrasing: maintain contractions and informal tone if present, unless clarity demands adjustment. | ||
| - Maintain accuracy: do not invent or omit key details like dates, names, or numbers. | ||
| - Produce clean prose: use complSmiley faceete sentences, correct punctuation, and paragraph breaks only where needed for readability. | ||
| - Produce clean prose: use complete sentences, correct punctuation, and paragraph breaks only where needed for readability. |
There was a problem hiding this comment.
didn't we fix this before?
Comment on lines
+38
to
+48
| editingPrompt: ` You are a Command-Interpreter assistant. Your job is to take a raw speech transcript-complete with hesitations, false starts, "umm"s and self-corrections-and treat it as the user issuing a high-level instruction. Instead of merely polishing their words, you must: | ||
| 1. Extract the intent: identify the action the user is asking for (e.g. "write me a GitHub issue," "draft a sorry-I-missed-our-meeting email," "produce a summary of X," etc.). | ||
| 2. Ignore disfluencies: strip out "uh," "um," false starts and filler so you see only the core command. | ||
| 3. Map to a template: choose an appropriate standard format (GitHub issue markdown template, professional email, bullet-point agenda, etc.) that matches the intent. | ||
| 4. Generate the deliverable: produce a fully-formed document in that format, filling in placeholders sensibly from any details in the transcript. | ||
| 5. Do not add new intent: if the transcript doesn't specify something (e.g. title, recipients, date), use reasonable defaults (e.g. "Untitled Issue," "To: [Recipient]") or prompt the user for the missing piece. | ||
| 6. Produce only the final document: no commentary, apologies, or side-notes-just the completed issue/email/summary/etc. | ||
| 7. Your response MUST contain ONLY the resultant text. DO NOT include: | ||
| - Any markers like [START/END CURRENT NOTES CONTENT] | ||
| - Any explanations, apologies, or additional text | ||
| - Any formatting markers like --- or \`\`\` |
There was a problem hiding this comment.
Is this to try to keep the prompt from leaking?
There was a problem hiding this comment.
This was just the result of running bun generate:constants
| - Resolve corrections smoothly: when the speaker self-corrects ("let's do next week... no, next month"), choose the final phrasing. | ||
| - Preserve natural phrasing: maintain contractions and informal tone if present, unless clarity demands adjustment. | ||
| - Maintain accuracy: do not invent or omit key details like dates, names, or numbers. | ||
| - Produce clean prose: use complete sentences, correct punctuation, and paragraph breaks only where needed for readability. |
There was a problem hiding this comment.
Oh we did fix it before, we just never regenerated
| * - Applies a gentle high-pass filter (~80 Hz) | ||
| * - Peak normalizes to ~-3 dBFS with a capped gain | ||
| */ | ||
| function enhancePcm16(pcm: Buffer, sampleRate: number): Buffer { |
There was a problem hiding this comment.
Nice 😎 can we pull this into an audio utility file?
fulltimemike
approved these changes
Oct 15, 2025
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.
…nding to groq
After investigation, the audio silence ended up being a red herring. We have been successfully sending all audio, even shortform, to the server. The issue is the perceived quality on groqs end. Our threshold of -.55 was too strong and for short audio we ended up tripping this threshold frequently.
What's updated in this PR: