Skip to content

Pipelines & Jobs

Source: src/renderer/src/components/PipelinesJobs.tsx

Overview

Build and run sequential processing pipelines. Each pipeline chains steps that pass output to the next: LLM inference, shell commands, and text filters.

Pipeline Definition

FieldDescription
idUUID
nameDisplay name
nodesArray of processing steps

Step Types

TypeDescription
llmRun input through the loaded model
toolExecute a shell command ($NODE_INPUT env var)
conditionMatch keyword, output ifTrue/ifFalse text

Steps run sequentially — each step's output feeds into the next as input.

Operations

ActionDescription
CreateDefine a new pipeline with steps
EditModify name, reorder, configure steps
DeleteRemove a pipeline
ExecuteRun the pipeline with an input string

Execution

POST /api/agents/{id}/execute with { input: "..." }.

The backend runs each node in order. LLM nodes use the currently loaded model. Tool nodes run shell commands with a 30-second timeout. Results show per-step status and output.

Storage

Pipelines are stored in ~/.silicon-studio/agents/agents.json.

API

EndpointMethodDescription
/api/agents/GETList all pipelines
/api/agents/POSTCreate/update pipeline
/api/agents/{id}DELETEDelete pipeline
/api/agents/{id}/executePOSTExecute pipeline

Released under the MIT License.