v1.21.0

Ironclad/rivetv1.21.0Feb 13, 2025by abrenneke

AI Summary

Major libraries update introducing conditional nodes for all node types, document loading (PDF), audio playback, and expanded AI model support across Anthropic, Google, and OpenAI.

Key Highlights

  • Conditional Nodes - every node can have a condition input
  • Document node for PDF loading
  • Play Audio node
  • Expanded AI model support (Gemini 2.0, OpenAI o1, GPT-4o Audio)
  • Anthropic PDF processing and Citations support

New Features

  • Conditional Nodes for all node types
  • Document node for PDF loading
  • Play Audio node
  • Chat Loop convenience node
  • Read All Files convenience node
  • To Markdown Table node
  • Cron node for scheduled graph execution
  • Anthropic PDF processing support
  • Anthropic Citations support
  • Anthropic endpoint override
  • Gemini 2.0 models support
  • Generative AI API Key support for Google
  • OpenAI o1 and o3-mini models
  • GPT-4o Audio output support
  • OpenAI predicted outputs
  • Server-side token counting

Full Release Notes

## New Features

- Conditional Nodes - Now every node can optionally have a condition input toggling whether the node runs
- {{tokens}} can be escaped by doing {{{token}}}
- New node: Document - Load documents such as PDF files into rivet for passing to e.g. Anthropic
- New node: Play Audio - Plays audio data directly!
- New convenience node: Chat Loop - Easily set up a chat loop with user input!
- New convenience node: Read All Files - Read all files in a directory!
- New node: To Markdown Table - Converts an array of objects into a markdown table!
- New node: Cron - Runs a graph at a specified time or interval!
- Anthropic Plugin: Support for PDF processing!
- Anthropic Plugin: Support for Citations!
- Anthropic Plugin: Endpoint can be overriden in the settings or the Chat Anthropic node
- Google Plugin: Support for all Gemini 2.0 models!
- Google Plugin: Support for using a Generative AI API Key instead of application credentials, the new preferred method
- OpenAI: Support for o1 and o3-mini models!
- OpenAI: Support for GPT 4o Audio! Get output as audio!
- OpenAI: Support for predicted outputs!
- OpenAI: Use the server for token counts instead of calculating them locally

## Changes & Fixes

- Added DocumentDataValue type
- Added DirectoryBrowserEditorDefinition type
- User input nodes are now handled like all other nodes, and any node can call `context.requestUserInput` to request user input.
- `RivetUIContext` now has access to the `nativeApi` configured.
- Added setting to throttle chat node updates - can improve performance