Changelog | November, 2024
Changelog | November, 2024
Changelog | November, 2024
November 24th, 2024
We’re excited to announce the beta release of the Vellum Workflows SDK! The Workflows SDK allows you to define the control flow of your AI systems using Python and push/pull the definition of your Workflows to/from Vellum’s UI.
The Workflows SDK is currently in closed beta. If you’d like to participate and provide feedback, please contact us.
For more information, check out the Vellum Workflows SDK docs.
We’ll be sharing more on the Workflows SDK in the coming weeks, so stay tuned!
November 23rd, 2024
We’ve added two new APIs for retrieving details about a Prompt/Workflow Deployment at a specific history point.
You can key off of either a history_id or the name of a Release Tag to retrieve the state of a Deployment
at that point in time.
For more info, check out our API reference:
These APIs are available in our SDKs beginning version 0.9.15.
November 21st, 2024
Prompt Sandboxes have received a major facelift, making it easier to reason about a Prompt’s input variables and values, as well as making it possible to more easily resize panels to suit your needs.
The Prompt Editor tab now consolidates the “Input Variables” and “Scenarios” panels, so that you now define variables and provide values in the same place. All panels are resizeable.
Comparison Mode makes use of the same variable input elements, making it possible to rename input variables right next to where you’d provide their values.
Lastly, Chat Mode has gotten major updates, such that you can now resize any panel and all UI elements look consistent across all modes.
November 20th, 2024
We’ve added support for the latest snapshot of GPT-4o (2024-11-20) to Vellum.
November 16th, 2024
We now have new API endpoints for listing all Release Tags associated with a Prompt/Workflow Deployment.
For details, check out our API Reference:
These APIs are available in our SDKs beginning version 0.9.5.
November 15th, 2024
We’ve made a number of small quality of life improvements throughout Vellum.
We now remember the sort order you set on index pages. If you sort by a column, leave the page, and come back, the sort order will be preserved.
We used to require that all CSV files uploaded for Test Cases have a header row with all expected columns. Now, you can choose to upload a CSV with a header row
containing only a subset of columns and Vellum will set the values of unspecified columns to null.
We now support semicolons as delimiters in CSV files uploaded for use as Test Cases.
We used to limit the max number of concurrent Test Case runs to 12. You can now crank up the concurrency to 36.
We now use the full height of the screen when displaying subworkflows in the workflow editor.
When copy/pasting Nodes in a Workflow, we used to prefix the label of each cloned Node with “Copy of.” This is no longer the case.
When on the detail page of a Workflow Execution, the information displayed was static. If the Workflow was still running, you had to refresh the page to see changes. Now, the page will update as the Workflow progresses.
November 5th, 2024
We’ve added audio input capabilities via our API for the following models:
You can now pass audio input to these models using a base64 encoded data URL.
You can read more about audio input capabilities for these models in OpenAI’s documentation.
To learn more about how to use audio input with Vellum, read our documentation.
November 5th, 2024
We’ve added support for Anthropic’s Haiku 3.5 10-22-2024 model.
Anthropic’s Claude Haiku 3.5 /api-reference/prompts/execute-prompt#request.body.inputs.CHAT_HISTORY.value.content.AUDIO.type