Skip to content

chat-summarization.md

Latest commit

 

History

History
115 lines (84 loc) · 4.51 KB

chat-summarization.md

File metadata and controls

115 lines (84 loc) · 4.51 KB
Raw

Chat Summarization

Chat Summarization allows you to compress long conversations into concise AI-generated summaries. This helps save context tokens and makes it easier to continue work on complex topics without losing important information.

Overview

When conversations grow long, they consume more context tokens with each message. Chat Summarization uses your configured Chat Assistant provider to create an intelligent summary that preserves:

  • Key decisions and conclusions
  • Technical details and code references
  • Important context for continuing the conversation

Key Features:

  • One-click compression: Summarize directly from the chat toolbar
  • Preserves original: Creates a new chat file, keeping the original intact
  • Smart summaries: AI extracts the most relevant information
  • Markdown formatted: Summaries are well-structured and readable

Using Chat Summarization

Compressing a Chat

  1. Open any chat with conversation history
  2. Click the Compress button (📦) in the chat top bar
  3. Wait for the AI to generate the summary
  4. A new chat opens with the compressed summary

What Gets Preserved

The summarization process:

  • Maintains chronological flow of the discussion
  • Keeps technical details, code snippets, and file references
  • Preserves key decisions and conclusions
  • Aims for 30-40% of the original conversation length

What Gets Filtered

The following message types are excluded from summarization:

  • Tool call results (file reads, searches)
  • File edit blocks
  • Thinking/reasoning blocks

How It Works

┌─────────────────────────────────────────────────────────────┐
│                    CHAT SUMMARIZATION                       │
├─────────────────────────────────────────────────────────────┤
│  1. Original chat messages are collected                    │
│  2. Tool/thinking messages are filtered out                 │
│  3. AI generates a structured summary                       │
│  4. New chat file is created with summary as first message  │
│  5. Original chat remains unchanged                         │
└─────────────────────────────────────────────────────────────┘

File Naming

Compressed chats are saved with a unique suffix:

original_chat.json → original_chat_a1b2c.json

Both files appear in your chat history, allowing you to switch between the full conversation and the summary.

Best Practices

  1. Summarize at natural breakpoints: Compress after completing a major task or topic
  2. Review the summary: Ensure important details were captured before continuing
  3. Keep originals: Don't delete original chats until you've verified the summary is sufficient
  4. Use for long sessions: Most beneficial for conversations with 20+ messages

When to Use

Good candidates for summarization:

  • Long debugging sessions with resolved issues
  • Feature implementation discussions with final decisions
  • Research conversations where conclusions were reached
  • Any chat approaching context limits

Consider keeping full history for:

  • Ongoing work that may need exact message references
  • Conversations with important code snippets you'll copy
  • Discussions where the reasoning process matters

Configuration

Chat Summarization uses your current Chat Assistant settings:

  • Provider: Same as Chat Assistant (Settings → QodeAssist → General)
  • Model: Same as Chat Assistant
  • Template: Same as Chat Assistant

No additional configuration is required.

Troubleshooting

Compression Button Not Visible

  • Ensure you have an active chat with messages
  • Check that the chat top bar is visible

Compression Fails

  • Verify your Chat Assistant provider is configured correctly
  • Check network connectivity
  • Ensure the model supports chat completions

Summary Missing Details

  • The AI aims for 30-40% compression; some details may be condensed
  • For critical information, keep the original chat
  • Consider summarizing smaller conversation segments

Related Documentation