> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.vellum.ai/changelog/2024/2024-12/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.vellum.ai/_mcp/server. # Changelog | December, 2024 ## Reject on Error Toggle for Guardrail Nodes **December 20th, 2024** You can now catch and handle errors in Guardrail Nodes just as you can most other Node types using the new "Reject on Error" toggle. ## Optional Input Variables for Prompts & Workflows **December 19th, 2024** You can now specify whether an input variable to a Prompt or Workflow is required or optional. This determines whether API calls made to invoke the Prompt or Workflow must include a value for that input variable. If the input variable is optional, the API call can omit its value and still be successful. If required, the API call must include a value for that input variable or else it will fail with a 400 status code response. Additionally, you may set explicit default values for optional input variables. If an API call omits a value for an optional input variable, the default value will be used instead. Newly created input variables going forward are optional by default with a `null` default value. This protects against breaking changes for existing API calls that do not include values for newly added variables. You can see a demo of it in action here: ## Support for JSON Files in Document Indexes **December 18th, 2024** We now support uploading `.json` files to Document Indexes for indexing and searching. ## Map End-User Feedback to Ground Truth Variables in Evals **December 18th, 2024** Capturing end-user feedback on a deployed AI system is a powerful way of understanding whether your system is performing as expected. However, even more valuable can be using this collected feedback in conjunction with evals. We've added the ability to map end-user feedback to ground truth variables in evals. This allows you to build up your eval datasets and the ground-truth data within, with real-world feedback from your end-users. Check out a demo of it in action below: ## Newly Added Models for Azure * **December 14th, 2024**: We've added support for the newest [GPT-4o/GPT-4o Mini model](https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models?tabs=python-secure%2Cglobal-standard%2Cstandard-chat-completions#gpt-4o-and-gpt-4-turbo) on Azure: * [GPT-4o 08/06](https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models#gpt-4o-0806) * [GPT-4o 11/20](https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models#gpt-4o-1120) * [GPT-4o Mini 07/18](https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models#gpt-4o-mini-0718) ## Link Search Results to Document and Expose Document Metadata **December 14th, 2024** When using the Search functionality within a Document Index, you can now see a link to the document that the search result is from. In addition to the link, you can also see the Document's metadata in the search results. ![Search Results with Document Link and Metadata](https://storage.googleapis.com/vellum-public/help-docs/changelogs/2024-12/document_link_and_metadata.png) ## Newly Added Model for Google * **December 11th, 2024**: Added support for [Gemini 2.0 Flex](https://ai.google.dev/gemini-api/docs/models/gemini#gemini-2.0-flash) via Google's Vertex AI. ## Newly Added Support for Together AI Models **December 11th, 2024** We've added support for Together AI's [Together AI](https://docs.together.ai/docs/introduction) models. Along with the launch of the Together AI integration, we've added support for [Llama 3.3 70B Instruct Turbo](https://docs.together.ai/docs/serverless-models). ## Newly Added Model for Groq * **December 8th, 2024**: We've added support for [Llama 3.3 70B Versatile](https://console.groq.com/docs/models#production-models) via Groq's production models. ## Newly Added Model for Fireworks AI * **December 8th, 2024**: We've added support for [Llama 3.3 70B Instruct](https://fireworks.ai/models/fireworks/llama-v3p3-70b-instruct) via Fireworks AI. ## Newly Added Models for AWS Bedrock * **December 5th, 2024**: With the launch of [AWS Bedrock's Nova Models](https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html), we've added support for Nova Micro v1, Nova Lite v1, and Nova Pro v1 in the `us-east-1` region. ## Newly Added Support for SambaNova AI Models **December 5th, 2024** We've added support for [SambaNova AI](https://docs.sambanova.ai/home/latest/index.html) as one of our newest model hosts! Along with the launch of the SambaNova integration, we've added the following model: * [Llama 3.1 405B Instruct](https://community.sambanova.ai/docs?topic=193#p-256-llama-31-family-3) For more information, check out the [SambaNova API reference](https://docs.sambanova.ai/sambastudio/latest/api-reference.html). ## Performance Improvements to the Workflow Editor **December 4th, 2024** The Workflow Editor UI could feel sluggish when working on sufficiently large Workflows consisting of many Nodes. We've made a number of broad sweeping optimizations so that editing Workflows in the UI should generally feel snappier and more responsive. ## Newly Added Models for OpenRouter * **December 17th, 2024**: We've added support for [Llama 3 Lumimaid 8B](https://openrouter.ai/neversleep/llama-3-lumimaid-8b) via OpenRouter. * **December 17th, 2024**: We've added support for [Llama 3.3 70B Instruct](https://openrouter.ai/meta-llama/llama-3.3-70b-instruct) via OpenRouter. * **December 17th, 2024**: We've added support for [Eva Llama 3.33 70b](https://openrouter.ai/eva-unit-01/eva-llama-3.33-70b) via OpenRouter. * **December 17th, 2024**: We've added support for [Mistral Large 2411](https://openrouter.ai/mistralai/mistral-large-2411) via OpenRouter. * **December 5th, 2024**: We've added support for [Eva Qwen 2.5 72B](https://openrouter.ai/eva-unit-01/eva-qwen-2.5-72b) via OpenRouter. * **December 2nd, 2024**: We've added support for [Cohere's Command R+ 08-2024](https://openrouter.ai/cohere/command-r-plus-08-2024) model through OpenRouter.