[#1] ✨ - Multiple AI providers#2
Merged
Merged
Conversation
…pt transformation - Updated .eslintrc.js to ignore coverage and test directories. - Expanded package.json to include new transformation provider commands and configuration options for multiple AI providers (OpenAI, Anthropic, Google Gemini, Azure, and Ollama). - Enhanced VSCodeConfigRepository to manage settings for different transformation providers and their respective API keys. - Updated README.md to document new features and configuration options for prompt transformation. - Refactored transformation logic to support dynamic provider selection via PromptTransformerFactory. - Updated pnpm-lock.yaml for new dependencies related to additional AI providers.
|
To view this pull requests documentation preview, visit the following URL: docs.page/vypdev/cursor-whisper~2 Documentation is deployed and generated using docs.page. |
vypbot
changed the title
1 task
Contributor
✨ Feature Actions
Debug log🚀 Happy coding! Made with ❤️ by vypdev/copilot |
…vements - Updated .eslintrc.js to ignore additional test patterns. - Refactored logging in extension activation for consistency. - Improved error handling in audio recording classes. - Enhanced prompt transformation logic with new validation and error handling. - Updated test cases for Azure and Ollama prompt transformers. - Adjusted configuration repository to enable prompt transformation by default. - Streamlined code formatting and structure across various files for better readability.
…r-0.1.0.vsix binary
vypbot
reviewed
May 23, 2026
…atus bar updates - Integrated configuration validation on startup to ensure API key and transformation provider are set correctly. - Updated StartRecordingUseCase to validate configuration before starting recording. - Added transformation provider label management in the status bar for better user feedback. - Improved error handling for missing or invalid configuration during recording.
…nt and error handling - Enhanced configuration validation during startup to ensure all required settings are correctly configured. - Updated StartRecordingUseCase to include pre-recording configuration checks. - Improved status bar feedback with dynamic transformation provider labels. - Strengthened error handling for scenarios involving missing or invalid configurations during recording.
…improvements - Introduced a setup wizard for first-time users to configure OpenAI API key and optional prompt optimization provider. - Updated configuration validation to ensure required settings are completed during startup. - Improved status bar feedback to indicate setup status and prompt users for incomplete configurations. - Enhanced command prompts for configuring API keys and transformation providers with clearer messaging. - Updated documentation to reflect new setup process and configuration options.
…s and enhance user experience - Replaced the setup wizard command with a direct configuration panel command for improved accessibility. - Updated configuration validation messages to guide users towards the new configuration panel. - Enhanced status bar items to reflect the new configuration command, promoting a more intuitive user interface. - Adjusted test cases to align with the updated command structure and ensure consistent behavior across the extension.
…d configuration validation - Introduced new commands for starting and stopping transcription and prompt optimization recordings. - Updated package.json to reflect deprecated recording commands and added new keybindings. - Enhanced configuration validation to ensure OpenAI API key is set for transcription and prompt optimization. - Improved status bar feedback to indicate recording modes and configuration requirements. - Refactored StartRecordingUseCase to handle different recording modes and session management. - Updated tests to cover new recording functionalities and validation scenarios.
…ates - Updated settings item text to include 'Settings' for clarity. - Added command for opening the configuration panel directly from the status bar. - Implemented logic to display a warning label when the configuration checklist is incomplete. - Enhanced tests to verify the new setup state behavior and UI updates for the status bar items.
…master' branch; modify button text in configuration panel for clarity
- Added a new configuration option for the transformation system prompt in package.json. - Updated the ConfigurationPanel to allow users to customize and reset the system prompt. - Enhanced prompt transformer classes to utilize the configured system prompt. - Improved tests to validate the new system prompt functionality and ensure correct behavior across different transformers. - Updated the default system prompt to provide clearer transformation instructions.
…uracy - Changed markdown description in package.json to link to the Configuration Guide instead of the Provider Guide. - Revised progress documentation to reflect the current status of automated testing and documentation structure. - Updated README to clarify provider setup instructions and project status references. - Removed outdated references to provider setup documentation in quickstart and ADR files. - Enhanced architecture overview to better describe the system's functionality and integration points.
vypbot
reviewed
May 24, 2026
| await useCase.execute('promptimize'); | ||
| await vscode.window.showInformationMessage('Recording started'); | ||
| } catch (error) { | ||
| if (error instanceof MissingApiKeyError) { |
Contributor
There was a problem hiding this comment.
MissingApiKeyError and ConfigError catch branches are unreachable dead code
Severity: low
Location: src/presentation/commands/StartRecordingCommand.ts:44
The catch block at lines 44-48 handles MissingApiKeyError and ConfigError, but neither is thrown within the try block. validateConfigurationForPromptimize returns a value (does not throw) and useCase.execute() only throws RecordingError. These branches will never execute.
Suggested fix:
Remove the MissingApiKeyError and ConfigError catch branches since they can never be triggered.
Written by vypdev/copilot for commit 51cd340232e0fb179ab054d6111870adfde909f1. This will update automatically on new commits.
…uter providers - Added OpenCode and OpenRouter as new transformation providers in package.json. - Updated configuration options to include OpenCode base URL and model identifier. - Enhanced documentation to reflect the addition of new providers and their setup instructions. - Improved validation and error handling for OpenCode and OpenRouter configurations. - Updated tests to cover new provider functionalities and ensure correct behavior across the extension.
…nsion - Introduced Cursor as a new transformation provider in package.json and updated related configuration options. - Enhanced documentation to include setup instructions for the Cursor provider and its models. - Updated validation logic to ensure proper configuration of Cursor API keys and models. - Added tests to cover the new Cursor provider functionality and ensure correct behavior across the extension. - Improved user interface elements to facilitate selection and configuration of the Cursor model.
vypbot
reviewed
May 24, 2026
| const result = await Agent.prompt(fullPrompt, { | ||
| apiKey, | ||
| model: { id: model }, | ||
| local: { cwd: process.cwd() }, |
Contributor
There was a problem hiding this comment.
CursorPromptTransformer uses process.cwd() which is unreliable in VS Code extension host
Severity: medium
Location: src/infrastructure/transformation/CursorPromptTransformer.ts:59
CursorPromptTransformer.transform() at line 59 passes local: { cwd: process.cwd() } to Agent.prompt(). In a VS Code extension host, process.cwd() often returns the VS Code installation directory (e.g. /usr/share/code) rather than the user's workspace folder. This causes the Cursor SDK to operate relative to the wrong directory, potentially breaking file references and workspace-aware operations.
Suggested fix:
Replace process.cwd() with vscode.workspace.workspaceFolders?.[0]?.uri?.fsPath || process.cwd() to use the actual workspace folder when available, falling back to process.cwd() only when no workspace is open.
Written by vypdev/copilot for commit edd9c47e6addd66b0af2b47ab05fce927aec9e9a. This will update automatically on new commits.
vypbot
reviewed
May 24, 2026
|
|
||
| async transform( | ||
| transcription: string, | ||
| context?: import('../../application/ports/IPromptTransformer').PromptContext |
Contributor
There was a problem hiding this comment.
ConfigurablePromptTransformer uses inline dynamic import for type annotation
Severity: low
Location: src/infrastructure/transformation/PromptTransformerFactory.ts:185
ConfigurablePromptTransformer.transform() at line 185 uses import('../../application/ports/IPromptTransformer').PromptContext as a dynamic import expression for a type annotation. This non-standard pattern may confuse bundlers (webpack), IDEs, and type checkers. The type should be imported at the top of the file via a regular import type statement.
Suggested fix:
Add import type { PromptContext } from '../../application/ports/IPromptTransformer' at the top of the file (alongside the existing import { IPromptTransformer } ... on line 2 or as a separate import type), and replace the inline import(...) expression with the simple name PromptContext.
Written by vypdev/copilot for commit edd9c47e6addd66b0af2b47ab05fce927aec9e9a. This will update automatically on new commits.
vypbot
reviewed
May 24, 2026
| return { client: this.client, model }; | ||
| } | ||
|
|
||
| const apiKey = new ApiKey(apiKeyStr); |
Contributor
There was a problem hiding this comment.
OpenRouterPromptTransformer uses OpenAI-specific ApiKey validation
Severity: high
Location: src/infrastructure/transformation/OpenRouterPromptTransformer.ts:45
OpenRouterPromptTransformer.ensureClient() at line 45 calls new ApiKey(apiKeyStr) which requires the key to start with sk- (OpenAI format). OpenRouter API keys use a different format and prefix, so this throws an uncaught Error('API key must start with sk-') before any transformation can occur, completely blocking OpenRouter provider usage.
Suggested fix:
Remove the ApiKey value object usage from OpenRouterPromptTransformer. The ApiKey class is OpenAI-specific (validates sk- prefix). For OpenRouter, simply validate that the key is non-empty without the sk- prefix check, consistent with how AnthropicPromptTransformer and GooglePromptTransformer handle API keys.
Written by vypdev/copilot for commit edd9c47e6addd66b0af2b47ab05fce927aec9e9a. This will update automatically on new commits.
…configuration - Refactored status bar item creation to include explicit names for 'transcribe', 'promptimize', and 'settings'. - Adjusted alignment values for status bar items to ensure consistent positioning. - Enhanced unit tests to verify the correct creation and configuration of status bar items. - Updated .vscodeignore to exclude additional test files and documentation from the build.
…oved documentation - Added a new configuration option for `transcriptionHint` in package.json to allow users to specify optional vocabulary for improved recognition accuracy. - Updated markdown descriptions for several configuration options to clarify their current status and future implementation plans. - Revised README and documentation to include details about the new recording modes and updated keyboard shortcuts. - Enhanced progress documentation to reflect the latest changes in configuration and user experience. - Improved user interface elements to facilitate better understanding of the new features and settings.
…nhancements - Added support for the token-costs npm package to fetch real-time provider pricing data. - Updated ConfigurationPanel to include dynamic provider comparison data based on token-costs. - Enhanced UI to display real-time pricing updates and fallback static data when necessary. - Revised provider comparison logic to improve user experience and data accuracy. - Updated configuration and documentation to reflect the new pricing strategy and architecture.
…s and documentation - Changed the extension name from Cursor Whisper to Promptimize in package.json and all related files. - Updated command names and titles to reflect the new branding throughout the extension. - Revised documentation, including README, quick start guide, and configuration details, to align with the new name. - Adjusted issue templates to reference Promptimize instead of Cursor Whisper. - Enhanced user experience by ensuring all references to the old name are replaced for consistency.
- Changed descriptions in hotfix and release issue templates to reference Promptimize instead of Copilot. - Updated disclaimers to clarify that only members of the Promptimize team can create these issues.
…ports - Expanded options in bug report, feature request, chore task, and help request templates to cover more specific areas of concern and actions. - Improved categorization of affected areas and actions to better guide users in reporting issues and requesting features. - Added new options related to audio recording, transcription, and configuration to align with recent updates in the extension.
Contributor
✨ Feature Actions
Debug log🚀 Happy coding! Made with ❤️ by vypdev/copilot |
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
2 participants
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.
📌 Summary
This PR adds support for multiple AI providers for prompt optimization, separating it from Whisper transcription (which remains exclusively OpenAI). Users can now choose from 8 providers — OpenAI, Anthropic, Google Gemini, Azure OpenAI, Ollama, OpenCode, OpenRouter, and Cursor — each with its own credentials, model selection, and configuration flow. A new configuration webview and setup wizard streamline the multi-provider experience, while the
RecordingSessionModesystem introduces two distinct recording modes: "transcribe" (raw Whisper output) and "promptimize" (Whisper + optional optimization).🎯 Related Issues / Tickets
🧩 Scope of Changes
PromptTransformerFactory— factory resolving the active provider from configurationConfigurationPanel— new VS Code webview for unified settings (Whisper key + optimization provider + system prompt)ProviderPricingService— real-time cost comparison viatoken-costsnpm package with static fallbackRecordingSessionMode—transcribe/promptimizemodes with separate start/stop commandsConfigureTransformationProviderCommand,TestTransformationCommand,OpenConfigurationPanelCommand,FirstTimeSetupCommandConfigurationValidationService— validates both Whisper and optimization setup with user-facing messagesRecordingStatusBarItem— enhanced tooltip showing both service statuses__tests__/— 15+ new test files covering providers, factory, validation, pricing, status bar, and integrationVSCodeConfigRepository— provider-specific settings (API keys via SecretStorage, model per provider)StartRecordingUseCase/StopRecordingUseCase— split into transcribe-only and promptimize variantsextension.ts— registers all new commands and servicespackage.json— 8 new commands, new configuration keys, new dependencies (@anthropic-ai/sdk,@google/generative-ai,@cursor/sdk,token-costs,ollama)README.mdand all docs — rebranded to Promptimize with multi-provider guidanceOpenAIPromptTransformer— simplified to match the new adapter patternRecordingStatusBarItem— two-service status display🛠️ Technical Details
IPromptTransformerport. ThePromptTransformerFactoryuses aswitchonTransformationProviderto instantiate the correct adapter — no dynamic dispatch or reflection.SecretStorageunderpromptimize.apiKey.{provider}. This avoids any secrets in settings.json.ProviderPricingServicefetches live pricing fromtoken-costs(daily cache, 3s timeout) and falls back to static data fromsrc/shared/constants/providerComparison.ts. Only OpenAI, Anthropic, Google, and OpenRouter support live pricing; Ollama, OpenCode, Azure, and Cursor always use static values.start/stopTranscribeRecordingandstart/stopPromptimizeRecording) allow users to bypass optimization entirely or apply it. The status bar tracks which mode is active viaRecordingSessionMode.baseURL, since their APIs are OpenAI-compatible. Cursor uses the@cursor/sdkwithAgent.prompt(). Ollama uses its own REST API directly.startRecording/stopRecordingcommands are deprecated but kept for migration. The default provider remains OpenAI, so existing single-provider setups work without changes.🔍 How to Test
F5).Promptimize: Setup Wizard— follow the flow to configure an OpenAI Whisper key.Promptimize: Open Configuration Panel— select a provider (e.g., Anthropic), enter its API key, and pick a model.Promptimize: Test Transformation— verify the selected provider responds correctly.Promptimize: Start Transcribe Recording— record audio, verify raw transcription output.Promptimize: Start Promptimize Recording— verify transcription + optimization output.Promptimize: Start Transcribe Recordingstill works without a provider configured.🧪 Test Coverage
PromptTransformerFactoryConfigurationValidationServiceProviderPricingServiceTransformationProvidervalue objectRecordingStatusBarItemFirstTimeSetupCommand,StartRecordingUseCasepromptimize.startRecording/promptimize.stopRecordingcommands are deprecated (marked in the UI). Users should migrate to theTranscribeorPromptimizevariants.promptimize.transformationModelis now scoped per-provider (e.g.,promptimize.anthropicModel,promptimize.ollamaModel). The old key is read as fallback for OpenAI.🚀 Deployment Notes
🔒 Security Considerations
📈 Performance Impact
ProviderPricingServiceuses a 1-hour cache with 3s timeout to avoid blocking the UI📝 Notes for Reviewers
The largest files are the configuration webview (
ConfigurationPanel.ts+ HTML + CSS) and the provider adapters. Focus review on:PromptTransformerFactory.ts:100–125— exhaustiveness check vianevertypeVSCodeConfigRepository.ts— SecretStorage key management and per-provider setting resolutionTransformationError)RecordingSessionModestate module is intentionally a simple singleton module (not injected) because it lives entirely in the sync presentation layer✅ Checklist