PR Reply Assistant

Draft context-aware, well-reasoned replies to GitHub pull request comments directly in VS Code/Cursor using the native Language Model API (vscode.lm) and your existing GitHub Copilot access.

Features

1) PR Comment Button

• Adds a Draft PR Reply icon button ($(comment-discussion)) to every PR comment's title toolbar.
• Reads the vscode.Comment text plus the parent CommentThread file and line range.
• Best-effort reads the GitHub PR title/body from the current branch for intent and scope context.
• Optionally prompts for extra one-off guidance before drafting.
• Runs the full multi-agent quality pipeline for Auto strategy (see Pipeline).
• Copies the final draft to clipboard. Metadata is hidden by default and remains available in the Output panel.
• Reuses the active draft run if you click the same comment again while generation is in progress.

2) Chat Participant (@prreply)

• Registers a sticky Copilot Chat participant named @prreply.
• Uses the model you select in the Chat UI (request.model).
• Accepts # file/location references for stronger code grounding.
• Infers tone and strategy intent from your prompt text (e.g. "push back firmly").
• Streams the draft back into the chat response, using the same metadata visibility setting as clipboard output.

3) Fix Grammar & Rephrase Comments

• Select any comment, review response, or text in the active editor, right-click, and select PR Reply Assistant: Fix Grammar or Rephrase (or run from the Command Palette).
• Instantly correct grammatical errors, spelling mistakes, or choose to rephrase the tone.
• Predefined Presets:
  • 📝 Fix Grammar & Spelling: Clean up typos and syntax issues while keeping the original phrasing.
  • 💼 Professional Rephrase: Elevate the tone to sound professional, constructive, and clear.
  • ⚡ Concise Rephrase: Condense long comments into short, punchy statements.
  • ❤️ Friendly Rephrase: Warm up comments to make them highly collaborative and supportive.
• Smart Small Model Filtering: Automatically detects and uses smaller/faster models (e.g., gpt-4o-mini, gemini-2.0-flash, or haiku) for near-instant execution.
• Interactive Review UX & Side-by-Side Diff: Displays a preview of the correction in a Quick Pick menu. Developers can launch a native side-by-side diff comparing the original and corrected text (retaining source syntax highlighting) to review all adjustments before applying them.
• Comment Prefix Protection: Safely preserves comment syntax (//, /*, #) if included in the selection.

4) Multi-agent quality pipeline

The Auto strategy runs a structured, evidence-based pipeline:

Safety gates run after the pipeline:

• Anchor gate — if the host did not supply an exact comment line range, falls back to a conservative best-effort reply.
• Safety gate — overrides a naive "agree" when imperative mutation evidence is present in symbols touched by the diff.
• Effort router — chooses fast, standard, or deep context collection based on anchor quality, strategy, and prompt breadth.

5) Deep context retrieval

The Planner has access to a suite of context tools it calls on demand:

• code_context_around_comment — code lines + diff hunks + related symbol snippets around the comment anchor
• pull_request_description — best-effort GitHub PR title/body for current-branch intent
• read_file_range / read_full_file — arbitrary workspace file reads
• git_diff_file / git_diff_pr / git_log — file-level available-change diffs and commit history
• symbol_evidence — structured write/read/mutation analysis for symbols in scope
• comprehensive_context — full-file content, page diff, reference-impact analysis, and available-change summary combined
• web_search — lightweight DuckDuckGo summary for external documentation when needed

Context depth setting controls pre-seeding:

• Standard — Planner gathers context on demand. Fast.
• Deep — available-change diagnostics and detailed diffs are pre-seeded into the evidence pack before the pipeline starts. Slower but broader.

6) Tone and strategy presets

Tone presets (set in Settings or optional setup):

Strategy presets:

7) Token usage reporting

• Prompt / completion / total token counts are always written to the extension Output panel.
• Set prReplyAssistant.outputDetail to compact or full if you want metadata included in clipboard/chat output.

Requirements

• VS Code or Cursor with engines.vscode: ^1.100.0.
• GitHub Copilot with Chat access enabled.
• Signed in to GitHub/Copilot inside VS Code.

Optional Setup

First use works immediately with defaults. Re-run setup at any time with PR Reply Assistant: Run Setup if you want to customize:

1. Language — choose or type the language for draft replies (e.g. English, Türkçe, Español).
2. Model — pick the Copilot language model to use (persisted to Settings).
3. Context depth — Standard or Deep.
4. Personal tone examples — optionally paste a few previous PR replies so drafts better match your style.

Usage

Draft from a PR comment

1. Open a pull request in a VS Code UI that surfaces comment threads (e.g. the GitHub Pull Requests extension).
2. Click the Draft PR Reply icon in the comment header.
3. If askForExtraInstructions is enabled, add optional guidance (or leave blank).
4. The pipeline runs, then the draft reply is copied to clipboard.
5. Paste into the PR response box.

Fix Grammar or Rephrase Comments in the Editor

1. Highlight any comment, review draft, or text in your active editor.
2. Right-click and choose PR Reply Assistant: Fix Grammar or Rephrase (or search for it in the Command Palette Cmd+Shift+P / Ctrl+Shift+P).
3. Select your desired mode from the Quick Pick list (e.g., Fix Grammar & Spelling, Rephrase: Professional, Rephrase: Concise, Rephrase: Friendly).
4. Review the corrected text preview. You can choose Compare Changes (Show Diff) to open a side-by-side comparison with syntax highlighting. Once satisfied, select Replace Selection to update the document, or Copy to Clipboard.

Use @prreply in Copilot Chat

Open Copilot Chat and start your message with @prreply:

@prreply Draft a reply to this comment:
"This function should be split — it does too many things at once."

Attach a file or range with # for stronger grounding:

@prreply #src/utils.ts Push back firmly — the complexity is necessary here.

Extension Settings

Architecture

src/
├── extension.ts          # activate / deactivate — wires commands and chat participant
├── agents.ts             # multi-agent quality draft pipeline (Planner, Decider, Critic, Arbiter, Writer)
├── constants.ts          # shared string/numeric constants
├── draftRegistry.ts      # in-flight draft de-duplication helpers
├── presets.ts            # tone, strategy, and context-mode preset data and helpers
├── settings.ts           # readUserSettings()
├── errors.ts             # user-facing error message formatting
├── llmClient.ts          # collectResponseWithUsage, token counting
├── modelResolver.ts      # model selection and fallback logic
├── onboarding.ts         # optional setup wizard
├── utils.ts              # escapeRegExp, truncateText, execFileAsync
├── context/
│   ├── anchor.ts         # buildAnchorEvidence — seed evidence from comment location
│   ├── chatRequest.ts    # resolveChatRequestContext, prompt inference helpers
│   ├── code.ts           # getCodeContext, buildCodeContextFromDocumentRange
│   ├── comment.ts        # extractCommentData, getThreadConversationContext
│   ├── comprehensive.ts  # getComprehensiveContext — full-page, ref-impact, PR changes
│   ├── evidence.ts       # getSymbolEvidenceContext — symbol write/read/mutation analysis
│   ├── git.ts            # git diff and deep context helpers
│   ├── prDescription.ts  # best-effort GitHub PR title/body retrieval
│   └── web.ts            # fetchDuckDuckGoSummary
└── pipeline/
    ├── draft.ts          # runAutoDraftPipeline, runForcedDraftPipeline, buildForcedStrategyPrompt
    ├── effort.ts         # deterministic fast / standard / deep effort router
    ├── format.ts         # formatSingleDraftOutput, humanizeStrategy
    ├── gates.ts          # applyAnchorGate, applySafetyGate
    ├── progress.ts       # humanizeProgressMessage
    ├── tools.ts          # buildContextToolRegistry — tool specs for the Planner
    └── types.ts          # AutoDecisionResult, PipelineDetailSummary

Release Notes

This README describes the current 1.1.0 release. See CHANGELOG.md for the full version history.

1.1.0

• Drafts copy or stream only the reply by default; compact/full metadata remains available through prReplyAssistant.outputDetail.
• Re-clicking the same PR comment while a draft is running reuses the active draft instead of opening stacked progress dialogs.
• Deterministic effort routing chooses fast, standard, or deep context collection based on anchor quality, strategy, and prompt breadth.
• Optional personal tone examples and best-effort PR title/body context improve draft fit without changing the core workflow.
• Local git evidence is described as available-change context, avoiding claims that it is a verified GitHub PR diff.

1.0.0

• Initial marketplace-ready release with the PR comment action, @prreply chat participant, multi-agent pipeline, context tools, strategy/tone presets, setup flow, language support, token reporting, and model persistence.