PerplexityPilot

PerplexityPilot is a VS Code extension that brings Perplexity models into your editor for Copilot‑like assistance: inline completions (ghost text), inline selection edits, quick fixes, and a simple “edit from prompt” command.

• Inline completions while you type
• Inline edit of selected code by instruction
• Quick Fix action to improve code
• Model switcher in Command Palette
• Secure Perplexity API key storage in VS Code Secret Storage

Requirements

• VS Code 1.102.0+
• Perplexity API key
• Node 18+ for development

Installation (development)

1. Clone this repo
2. Install dependencies: npm install
3. Build/watch: npm run watch or press F5 to launch the Extension Development Host
4. In the Dev Host, open a file and try the commands below

Commands

• PerplexityPilot: Edit Code from Prompt
  Prompts you for a goal and inserts a single generated line at the end of the active file.

• PerplexityPilot: Inline Edit Selection
  Select code, run the command, describe the change, and it replaces the selection.

• PerplexityPilot: Switch Model
  Quick-pick of supported Perplexity models and updates your setting.

Inline completions are enabled automatically when the extension activates. Use “Trigger Inline Suggestion” to force a suggestion if needed.

Settings

• perplexity.model
  Default: sonar
  Allowed: sonar, sonar-reasoning, sonar-pro, sonar-reasoning-pro, sonar-deep-research, r1-1776
  The extension validates and falls back to sonar if an unsupported model is selected.

• perplexity.maxTokens
  Default: 120
  Maximum tokens used for quick generations.

API Key

On first use, the extension prompts for your Perplexity API key and stores it in VS Code Secret Storage. To update it later, re-enter via the prompt when needed or clear it from Secret Storage.

Features

• Inline Completions
  Provides code suggestions inline (ghost text) based on the current line context.

• Inline Edit
  Turns your selection and instruction into a transformed code block, replacing the selection.

• Quick Fix / Code Action
  Lightbulb entry “PerplexityPilot: Improve this code” that routes to Inline Edit.

• Model Switcher
  Quick-pick of supported models with validation.

Usage Tips

• Inline completions work best when there’s some prefix on the current line. If nothing shows, try the “Trigger Inline Suggestion” command or continue typing a bit.
• Use Inline Edit for refactors or targeted changes. Select the smallest code region that captures the intent.
• Keep perplexity.maxTokens modest for snappier responses during editing; increase when you need longer output.

Example Workflow

1. Open a Python file
2. Select a function
3. Run “PerplexityPilot: Inline Edit Selection”
4. Instruction: “Add a docstring and type hints”
5. Review the replacement and undo if needed

Development

• Build once: npm run compile
• Watch mode: npm run watch
• Package a VSIX: npm run package (produces .vsix)
• Publish to Marketplace: use vsce with your publisher id

Project structure

• src/extension.ts
  Activates the extension and registers features

• src/perplexityClient.ts
  OpenAI SDK client configured for Perplexity API

• src/inlineCompletions.ts
  InlineCompletionItemProvider for ghost text

• src/codeActions.ts
  CodeActionProvider that surfaces a quick fix

• src/commands/inlineEdit.ts
  Selection-based transformation command

• src/commands/switchModel.ts
  Model quick-pick with