v1.8.0

Ironclad/rivetv1.8.0Sep 22, 2023by abrenneke

AI Summary

This v1.8.0 release introduces significant improvements to async capabilities and node functionality in the Rivet graph processing system. Key additions include a new RivetUIContext object for UI-related functions, async support for getBody and getUIData, and a getChatNodeEndpoint method for handling additional headers and load balancing in chat nodes.

Key Highlights

New RivetUIContext object passed to getEditors, getBody, and getUIData for additional implementation context
getBody and getUIData now support async implementations
New getChatNodeEndpoint in ProcessContext/createProcessor for headers and load balancing (e.g., Azure OpenAI)
Prompt node can now compute token count and output it as a port
Fixed GraphProcessor total cost calculation to include split nodes

Breaking Changes

HTTP Call node output ID fix may break graphs that relied on the previous incorrect output ID

New Features

RivetUIContext object for getEditors, getBody, and getUIData
Async support for getBody and getUIData functions
getChatNodeEndpoint method for chat node header configuration and load balancing
Prompt node token count computation with port output

Full Release Notes

## New Features

- `getEditors`, `getBody`, and `getUIData` are passed a new `RivetUIContext` object with additional information that can be used in their implementations
- `getBody` and `getUIData` can now be async
- Add `getChatNodeEndpoint` to `ProcessContext`/`createProcessor` - can be used to pass additional headers to chat nodes, and for things like load balancing based on endpoint/model selected, if using Azure OpenAI for example.
- Prompt node can compute token count and output it as a port

## Changes & Fixes

- Fix total cost of GraphProcessor not including split nodes
- Fix remote debugger not sending `start` event correctly
- Improve Chat node error handling
- Fix output ID of HTTP Call node (may break some graphs that used it)