Product Update - Experiments, Custom Evaluators, Cost Tracking

Product Update - Prompt & Model A/B Testing

Sep 30, 2024

|

0 min read

Share:

Product Update - Prompt & Model A/B Testing

Sep 30, 2024

|

0 min read

Share:

Product Update - Experiments, Custom Evaluators, Cost Tracking

We're thrilled to announce a series of significant updates and new features for Literal AI. Our team has been working hard to enhance your experience and provide you with more powerful tools for AI development and management. Let's dive into what's new!

🚀 New Features

Prompt Playground Enhancements

  • Amazon Bedrock Support: Expand your model options with Amazon Bedrock integration.

  • Structured Output: Specify any JSON schema to structure LLM outputs, both in the Playground and in prompts.

  • Keyboard Shortcuts: Streamline your workflow with new keyboard shortcuts.

Advanced Evaluation Tools

  • Custom LLM-as-a-Judge: Create custom evaluator prompts for production monitoring.

  • Improved Score Display: Human/AI reviews now show category labels instead of values.

Experimentation

  • "Run Experiment" Feature: Evaluate prompts managed on Literal AI directly on the platform.

Logging

  • Environment-Specific Logging: Log to specific environments (dev, staging, prod).

Cost Tracking and Optimization

  • Model Costs: Monitor actual costs associated with your logged generations.

  • Total Cost Chart: New dashboard addition showing input and output token costs.

💡 Platform Improvements

Many UI/UX & performance improvements across the platform !

Dashboard Updates

  • New tiles for recently ingested Runs & Scores with one-click data access.

  • Add descriptions to individual values in score schemas.

Performance and Stability

  • Increased worker stability and improved query efficiency.

  • Enhanced multi-threaded asynchronous step ingestion for faster performance.

Evaluation

  • Human/AI reviews display category labels instead of values

  • Optimized score management for faster edits and fewer queries

Dataset Management

  • Enhanced dataset table with a side-panel Item View.

TypeScript SDK

  • Introduced decorators with metadata, tags, and stepid fields.

  • Improved compatibility with the LangChain ecosystem.

  • Simplified Vercel AI SDK integration.

Python SDK

  • Extended LlamaIndex instrumentation support.

  • Improved LangChain/LangGraph integration.

  • Updated Mistral AI integration.

  • Enhanced tag and metadata handling for LangChain logs.

📚 New Resources

  • Continuous Improvement Guide: Learn how to improve your LLM applications over time.

  • LangGraph Example: Check out our new end-to-end example with langgraph in our cookbook.

🔗 Chainlit Realtime Voice API

We're constantly working to improve Literal AI and provide you with the best tools for your AI development needs. Thank you for your continued support and feedback. Stay tuned for more updates!

For detailed information on these features, please visit our documentation.

Happy coding!

Ship AI with confidence

Gain visibility on your AI application

Create an account instantly to get started or contact us to self host Literal AI for your business.

Ship AI with confidence

Gain visibility on your AI application

Create an account instantly to get started or contact us to self host Literal AI for your business.

Ship AI with confidence

Gain visibility on your AI application

Create an account instantly to get started or contact us to self host Literal AI for your business.

Ship AI with confidence

Gain visibility on your AI application

Create an account instantly to get started or contact us to self host Literal AI for your business.