🧠 Improve with AI
What's New:
Create better prompts with AI assistance in both Prompts and Workflows.- AI-generated system prompts to guide LLM behavior
- AI-generated user prompts with clear instructions, steps, and output formats
- Ability to iterate on prompts with additional AI improvement requests
- Available in both standalone prompts and workflow prompts
Benefits:
- Quickly create 90% optimized prompts without prompt engineering expertise
- Clear prompt structure with instructions, steps, and formatting
- Reduced hallucinations with proper prompt construction
- Streamlined prompt creation for beginners and experts alike
🧩 Maximum Character Control for Website Scraping
What's New:
Set character limits on website scrape operations in workflows.- Control exactly how much content is scraped from websites
- Default conservative 7,500 character limit
- Customizable limits to fit your workflow needs
- Helps manage token consumption in complex workflows
Benefits:
- Prevents hitting LLM token limits in multi-step workflows
- Better control over input context size
- Improved reliability for workflows with multiple data sources
- More predictable workflow execution
📊 Advanced Dataset Configurations
What's New:
Enhanced control and transparency for vector database storage.- Choose between default (speed-optimized) or advanced chunking strategies
- Configure custom chunk length and overlap settings
- Remove sensitive data like emails and phone numbers
- Run custom prompts to clean or format data before storage
- Detailed metrics on chunks, paragraphs, sentences, tokens, and costs
Benefits:
- Complete transparency in how data is stored and chunked
- Better control over vector database storage
- Clear visibility into storage costs and token usage
- Improved data privacy with sensitive information removal
🔍 Test Dataset Searches
What's New:
Search your datasets directly in the dashboard without using agent credits.- Three search options: Hybrid, Vector, and Full Text
- Configurable chunk limits (1-20 results per search)
- Adjustable similarity thresholds for result quality control
- Real-time testing of your datasets before using in agents
Benefits:
- Verify dataset content before using in production
- Save credits by testing without agent invocation
- Fine-tune search parameters for optimal results
- Better understanding of how vector and text search differ