Scale - AI-Powered Developer Platform for VS Code

Scale is an AI-powered developer platform extension for Visual Studio Code that helps you understand code, debug issues, and ship faster. It deeply understands your codebase and how you build software, providing intelligent assistance through Chat, Next Edit, best practices, and Code Completions.

Features

Chat

Scale's Chat enables you to work with your codebase using natural language. You can:

• Ask questions about your code and get contextual answers
• Get help debugging issues
• Generate entire functions and tests
• Understand complex code patterns
• Receive guidance on best practices

The Chat feature understands your codebase in real-time, providing responses that are specific to your project.

AI Agent

Scale's AI Agent takes the extension to the next level by allowing Scale to make modifications directly to your codebase. The Agent can:

• Create and implement entire features
• Fix bugs across multiple files
• Refactor code to improve quality
• Set up project structures and boilerplate
• Write comprehensive tests
• Generate documentation

The Agent breaks down complex tasks into a step-by-step plan and implements each step while keeping you informed about the changes being made.

Code Completions

Scale provides personalized in-line completions that truly understand your codebase, dependencies, and external APIs:

• Completions reflect your code style, dependencies, idioms, and best practices
• Grab any SDK, function, or API without breaking your flow
• Find and delete unused or non-functioning code
• Make edits across multiple lines or entirely different files
• Write a comment using natural language and let completions implement it
• Completions pick up where your chat left off for continuity

Next Edit

The Next Edit feature allows you to navigate through your codebase more efficiently:

• Jump to the next edit with a keystroke
• Navigate through related changes
• Quickly move between files that are part of the same feature or fix

Context Engine

Scale's Context Engine provides real-time indexing of your codebase:

• Understands the structure and relationships in your code
• Tracks dependencies between files and modules
• Analyzes code patterns and best practices
• Provides context for AI-powered features

Model Performance Analytics

Scale includes comprehensive analytics for monitoring and optimizing language model performance:

• Track response times, token usage, and costs for all model interactions
• View detailed performance metrics by model and feature
• Collect user feedback on model responses to improve accuracy
• Export analytics data for further analysis
• Visualize performance trends with interactive charts

External Integrations

Scale can integrate with external services to add additional context to your requests and take actions on your behalf:

Git Integration

• Branch-aware context for more relevant assistance
• Commit history analysis for better code understanding
• File status tracking for targeted help with changes
• Repository structure awareness for comprehensive assistance

Issue Tracker Integration

• GitHub Issues: Access and create issues directly from VS Code
• Jira: Work with issues and project management
• Linear: Manage tasks and track progress
• Issue context for AI features to provide more relevant assistance

Package Manager Integration

• Dependency analysis for npm, yarn, and pnpm projects
• Vulnerability detection and remediation suggestions
• Package context for AI features to understand project dependencies
• Script management and execution

Documentation Integration

• Confluence: Access documentation and knowledge base
• Notion: Integrate with team wikis and documentation
• Documentation generation and updating

Getting Started

Installation

1. Open Visual Studio Code
2. Go to the Extensions view (Ctrl+Shift+X)
3. Search for "Scale"
4. Click Install

Configuration

1. Open VS Code settings (Ctrl+,)
2. Search for "Scale"
3. Enter your API key
4. Configure enabled features and integrations

Using Chat

1. Click on the Scale icon in the activity bar
2. Select "Chat" mode
3. Type your question or request
4. Receive contextual responses based on your codebase

Using Agent

1. Click on the Scale icon in the activity bar
2. Select "Agent" mode
3. Describe the task you want the Agent to perform
4. Review the plan and watch as the Agent implements the changes

Using Code Completions

Code completions will appear automatically as you type. You can:

1. Accept a completion with Tab
2. Cycle through alternatives with Alt+]
3. Dismiss a completion with Esc

Using Next Edit

1. Make changes to your code
2. Press Alt+N (or your configured keybinding) to jump to the next edit location
3. Continue navigating through your edits with the same shortcut

Using Model Performance Analytics

1. Click on the Scale icon in the activity bar
2. Select "Model Analytics" from the view container
3. View performance metrics for different models and features
4. Use the time range selector to filter data
5. Export analytics data for further analysis
6. Provide feedback on model responses when prompted

Requirements

• Visual Studio Code 1.60.0 or higher
• An active Scale API key

Extension Settings

This extension contributes the following settings:

• scale.apiKey: API key for Scale AI services
• scale.enableCodeCompletions: Enable AI-powered code completions
• scale.enableNextEdit: Enable Next Edit feature
• scale.integrations: Configure external service integrations
  • scale.integrations.git: Enable Git integration
  • scale.integrations.github: Enable GitHub integration
  • scale.integrations.jira: Enable Jira integration
  • scale.integrations.linear: Enable Linear integration
  • scale.integrations.npm: Enable npm integration
  • scale.integrations.yarn: Enable yarn integration
  • scale.integrations.pnpm: Enable pnpm integration
• scale.git: Git integration settings
  • scale.git.includeBranchContext: Include branch context in AI requests
  • scale.git.includeCommitHistory: Include commit history in AI requests
  • scale.git.includeFileStatus: Include file status in AI requests
  • scale.git.maxCommitHistory: Maximum number of commits to include in history
• scale.github: GitHub integration settings
  • scale.github.repository: GitHub repository in format "owner/repo"
  • scale.github.token: GitHub personal access token
• scale.jira: Jira integration settings
  • scale.jira.url: Jira instance URL
  • scale.jira.username: Jira username
  • scale.jira.token: Jira API token
  • scale.jira.project: Jira project key
• scale.analytics.trackModelPerformance: Enable tracking of model performance metrics
• scale.analytics.collectFeedback: Enable collection of user feedback on model responses
• scale.analytics.feedbackPromptFrequency: Control how often to prompt for feedback
• scale.analytics.costTracking: Enable tracking of estimated costs for model usage
• scale.analytics.modelCosts: Configure cost per 1000 tokens for different models
• scale.benchmarking: Performance benchmarking settings
  • scale.benchmarking.enabled: Enable performance benchmarking
  • scale.benchmarking.defaultIterations: Default number of iterations for benchmarks
  • scale.benchmarking.defaultWarmupIterations: Default number of warmup iterations
  • scale.benchmarking.trackMemory: Track memory usage during benchmarks
  • scale.benchmarking.retentionDays: Number of days to retain benchmark results

Known Issues

• The extension is currently in beta and may have occasional performance issues with very large codebases
• Some language-specific features may have varying levels of support

Release Notes

0.3.0

• Enhanced ecosystem integrations with Git, GitHub Issues, and package managers
• Added comprehensive performance benchmarking system
• Improved context window optimization with integration-aware chunking
• Enhanced documentation with user guide, developer guide, and API reference
• Implemented CI/CD pipeline for automated builds and releases
• Added cross-platform testing and quality assurance improvements

0.2.0

• Added Model Performance Analytics
• Implemented feedback collection for model responses
• Added support for custom models and providers
• Optimized context window handling for large codebases
• Improved error handling and logging

0.1.0

• Initial release of Scale
• Added Chat, Agent, Code Completions, and Next Edit features
• Implemented Context Engine for codebase understanding
• Added support for external integrations

Privacy and Security

Scale takes your privacy and security seriously:

• Your code never leaves your machine unless explicitly sent to the Scale API
• API keys are stored securely in VS Code's secret storage
• You have full control over which features and integrations are enabled
• All communication with Scale services is encrypted

Feedback and Support

We welcome your feedback and suggestions for improving Scale:

• File issues on our GitHub repository
• Contact support at support@scale.com
• Join our community on Discord

License

Copyright (c) 2023 Scale AI, Inc. All rights reserved.

Powered by Scale AI