Chat Participant (@heaplens)

HeapLens registers as a VS Code Chat Participant, making it available in the chat panels of extensions that support the VS Code Language Model API. This includes GitHub Copilot, Cline, and Claude Dev.

Setup

No additional configuration is needed. The @heaplens participant is automatically registered when the HeapLens extension activates. You just need:

1. HeapLens extension installed in VS Code
2. A chat extension that supports the VS Code LM API:
   • GitHub Copilot (most common)
   • Cline
   • Any extension that registers language models via vscode.lm
3. An open .hprof file analyzed by HeapLens

Usage

In the chat panel, type @heaplens followed by your question:

@heaplens What is the biggest memory consumer in this heap dump?

Slash Commands

Three slash commands provide focused analysis:

Command	Description	Example
/analyze	General heap analysis (default)	@heaplens /analyze
/leaks	Focused leak suspect analysis	@heaplens /leaks Are there any unbounded caches?
/explain	Explain a specific class or concept	@heaplens /explain java.util.HashMap

If you don't use a slash command, /analyze is the default behavior.

Examples

@heaplens What are the top 3 things I should fix to reduce memory?

@heaplens /leaks

@heaplens /leaks Is the SessionCache growing over time?

@heaplens /explain Why is byte[] the largest class by shallow size?

How It Works

User types "@heaplens What's using all the memory?"
        │
        ▼
VS Code routes to HeapLens chat participant
        │
        ▼
Participant reads current heap analysis data
        │
        ▼
Builds prompt with analysis context
  (same promptTemplates.ts as direct API)
        │
        ▼
Calls vscode.lm.selectChatModels()
  → tries GPT-4o first, falls back to any model
        │
        ▼
Streams response via model.sendRequest()
        │
        ▼
Response appears in the chat panel

The chat participant uses the VS Code Language Model API (vscode.lm), which means:

• No API key configuration needed (the host extension handles authentication)
• Model selection is automatic (prefers GPT-4o, falls back to whatever is available)
• Conversation history is managed by the chat panel

vs. Built-in AI Chat Tab

Feature	Chat Participant (@heaplens)	AI Chat Tab (API Key)
Where	Copilot/Cline chat panel	Inside HeapLens editor
API key	Not needed	Required
Model	Whatever the host provides	Your choice (Anthropic/OpenAI)
Explain Object	Not available	Available
Explain Leak	Not available	Available
Conversation context	Full chat history from panel	Per-editor (20 exchanges)

The chat participant is best for quick questions. The built-in AI Chat tab with an API key gives you more control and access to the Explain features.

Troubleshooting

"No heap analysis data available" Open an .hprof file and wait for the analysis to complete before using @heaplens.

"No language model available" Make sure GitHub Copilot (or another LM-providing extension) is installed and you are signed in.

@heaplens doesn't appear in the chat panel Restart VS Code. The participant is registered during extension activation, which happens when you first open an .hprof file or run a HeapLens command.