Breaking changes

v0.9.0 Pipeline node can have multiple post-pipeline graphs. The post pipeline node port name has changed to have index numbers.

v0.8.0 Iterator node output is no longer embedded in outputs. Its more flat to make it consistent to the inputs.

v0.7.0 pipelineInput is available to all stages. However it is always available as pipelineInput. Even to the first stage.

v0.3.0 has renamed the Iterator Node (from Iterator Plugin Node). You'll have to readd the node

Description

This is an utilities plugin for rivet

$3

The iterator Node will map through an array of object inputs objectDataValues[][]. The plugin will then call the subgraph you'd like to execute for item in the array. Additionally

- it has a chunkSize option to limit the number of concurrent executions.
- it has a cache option to cache subgraph outputs of successful item runs.
- graph to run for each item in the array

#### Using the Iterator Node

Once install you can use it as show in the examples below. Make sure the inputs are DataValues.

Inputs must be an array of objects to iterate over. Each object in the array should be a ObjectDataValue {type: 'object', value: }; where is of the format {type: object , value: {: }} The graph input id should match the graph's input ports. The input value should be a DataValue.

Ouputs will be an array of ObjectDataValue type: object , value: {:

}

$3

The PipelineNode will take and input and run it through a pipeline of graphs (stages). Each stage output should be the next graph's input. The pipeline node has a optional pre/post stage to run before and after the pipeline stages. Additionally, you have option to loop through the pipeline stages multiple with the loop number option.

- it has a cache option to cache subgraph outputs of successful item runs.
- it has a loop number option to loop through the pipeline stages multiple. default 1

#### The stage graphs

- pre optional graph
- The output of the pre graph will be the input to the first stage graph
- stage x graphs
- The output of the x graph will be the input to the next stage y graph
- For example stage 1 output will be the input to stage 2
- You can loop through the stages x multiple times with the loop number option. The output of the last loop will be the final output of the pipeline node.
- post optional graph
- The output of the post graph will be the final output of the pipeline node

#### Graph Inputs

- The input to the first stage has id pipelineInput
- The inputs to all following stages is the output of the prior stage
- All stages have access to id pipelineInput

#### An example

An example of executing a pipeline with 3 stages and a pre/post stage.

- run pre stage
- output of pre-stage => stage 0
- run stage 0
- output of stage 0 => stage 1
- run stage 1
- ouput of stage 1 => stage 2
- run stage 2
- if no loop: output of stage 2 => post-stage
- if loop: output of stage 2 => stage 0. Repeat loop number times
- output of stage n => post-stage

> All stages will have access to initial pipelineInput

$3

The node will allow you to query pinecone for vectors. Allows filters with metadata and sparse vectors. It also allows you to get scores back from the api. You have access to alpha which is the weight of the sparse vector 1 vs dense vector 0

$3

This node will allow you to upsert into pinecone. It allows you to upsert sparse vectors and metadata as well.

Installation

See Rivet Plugins for more details

Source

This was forked from: This project is an example of a Rivet plugin. It is a minimal TypeScript Rivet plugin that adds a single node called Example Plugin Node.

> See Rivet Example Plugin on details on plugin architecture and building it

Use pnpm to build and install