Skip to main content

AI Summarizer

Redline includes a dedicated AI summarizer that automatically generates summaries for various content types. The summarizer uses a separate, typically faster and cheaper model than the main chat assistant.

What Gets Summarized​

Content TypeWhen SummarizedWhat's Generated
News ArticlesOn demand (click "Summarize")Summary, key points, sentiment, topics
NotesOn demandCondensed summary of note content
NarrativesOn demand or when updatedExecutive summary, key themes, key actors
RSS ArticlesDuring relevance filteringUsed to determine if article matches your criteria

News Article Summaries​

When you create a Newsfeed node from a URL, Redline scrapes the article content. You can then generate an AI summary:

  1. Select the Newsfeed node
  2. Click Summarize in the Context Panel
  3. The AI generates:
    • Summary - 2-3 sentence overview
    • Key Points - 3-5 bullet points
    • Sentiment - Overall tone (positive, negative, neutral)
    • Topics - Main themes covered

Note Summaries​

Long notes can be summarized to extract key information:

  1. Select a Note node with substantial content
  2. Click Summarize in the Context Panel
  3. The AI generates a condensed summary

This is useful for:

  • Summarizing meeting notes
  • Condensing research findings
  • Creating quick reference summaries

Narrative Summaries​

Narratives can have AI-generated executive summaries that synthesize all member nodes:

  1. Select a Narrative node
  2. Click Generate Summary in the Context Panel
  3. The AI analyzes all member nodes and creates:
    • Executive Summary - Overview of the storyline
    • Key Themes - Main topics across all nodes
    • Key Actors - Important people/organizations involved

The narrative summary is automatically updated when you regenerate it, reflecting any new nodes added to the narrative.

Summarizer Model​

The summarizer uses a dedicated model configured separately from your chat model. This allows you to use a faster, cheaper model for routine summarization while keeping a more capable model for complex research.

Default Models by Provider​

ProviderDefault Summarizer Model
Redline AIClaude 3.5 Haiku
ClaudeClaude 3.5 Haiku
OpenAIGPT-4o Mini
OpenRouterClaude 3.5 Haiku
OllamaLlama 3.2

Changing the Summarizer Model​

  1. Go to Settings > AI Provider
  2. Find the Summarizer Model dropdown
  3. Select your preferred model
  4. Click Save
Cost Optimization

Using a faster model like Claude 3.5 Haiku or GPT-4o Mini for summarization can significantly reduce costs while maintaining quality for straightforward summarization tasks.

RSS Relevance Filtering​

The summarizer powers RSS feed relevance filtering, helping you automatically filter incoming articles:

How It Works​

  1. When new articles arrive from an RSS feed
  2. The summarizer evaluates each article against your relevance prompt
  3. A relevance score (0-100) is calculated
  4. Only articles meeting your threshold are imported as nodes

Configuring Relevance Filtering​

When creating or editing an RSS Feed node:

  1. Set a Relevance Prompt describing what you're looking for
  2. Set a Relevance Threshold (0-100)
  3. Articles scoring above the threshold are imported

Example Relevance Prompt:

Articles about Tesla stock price movements, Elon Musk business decisions, or electric vehicle market trends in North America

Example Threshold: 70 (imports articles scoring 70% or higher relevance)

Tips for Effective Filtering​

  • Be specific - Vague prompts like "interesting news" produce inconsistent results
  • Include keywords - Mention specific terms, names, or topics
  • Set appropriate thresholds - Start at 50-60 and adjust based on results
  • Review periodically - Check what's being filtered out to tune your prompt

Token Usage​

Summarization uses tokens from your AI budget:

  • Input tokens - The content being summarized
  • Output tokens - The generated summary

Summarization typically uses fewer tokens than chat interactions since:

  • Responses are shorter and structured
  • No conversation history is included
  • Faster models have lower per-token costs

Troubleshooting​

Summary Not Generating​

  • Ensure you have an AI provider configured
  • Check that the node has content to summarize
  • For newsfeeds, wait for article scraping to complete first

Poor Quality Summaries​

  • Try a more capable summarizer model
  • Ensure the source content is clear and complete
  • For RSS filtering, refine your relevance prompt

RSS Articles Not Being Imported​

  • Lower your relevance threshold
  • Make your relevance prompt more inclusive
  • Check that the RSS feed is being polled (check feed status)