Render Pipeline API

The Render Pipeline API is the execution endpoint for multi-step AI pipelines defined in Slop Machine. Unlike Buckets (which are pre-templated and locked down to dynamic variables), Pipelines transfer full runtime control to the developer, allowing dynamic prompts, multi-step chaining (e.g. text generation → video generation → audio → video text compositing), and custom metadata persistence.

Pipelines are identified by their pipelineId.

Endpoint Details

• Method: GET or POST
• Base URL: https://us-central1-slopmachine-12bfb.cloudfunctions.net/renderPipeline
• Identification: pipelineId passed in query parameters or request body.

Usage in Components (SlopImage, SlopVideo, SlopText)

The React and Svelte components seamlessly support Pipelines by providing the pipelineId and runtime prompt:

import { SlopImage } from "@slopmachine/react";

<SlopImage
  pipelineId="pipe_abc123"
  prompt="1960s retro robot eating pizza on Mars with a flying pig"
  aspectRatio="16:9"
  metadata={{ userId: "usr_42", campaign: "launch_v1" }}
/>

When pipelineId is passed to <SlopImage> or <SlopVideo>, the SDK automatically appends redirect=true, allowing the media tag to transparently redirect to the final rendered asset URL.

Programmatic Execution (executePipeline)

For backend workflows or programmatic client execution returning the full typed execution breakdown:

import { executePipeline } from "@slopmachine/core";

const result = await executePipeline({
  pipelineId: "pipe_abc123",
  prompt: "Generate a vibrant summer fashion reel",
  variables: { location: "Tokyo" },
  metadata: { batchId: "b-987" }
});

console.log("Status:", result.status);
console.log("Primary Output URL:", result.url);
console.log("Step Breakdown:", result.stepResults);
console.log("Fuel Cost:", result.totalFuelCost);
console.log("Execution Time (ms):", result.totalComputeTimeMs);

Request Parameters

Parameter	Type	Required	Default	Description
pipelineId	string	Yes		The ID identifying the pipeline.
siloId	string	No		Optional Silo ID to optimize direct document lookup.
prompt	string	No		Runtime prompt passed into the pipeline's entry step (e.g., text, image, or video prompt).
variables	object	No	{}	Dynamic variables to interpolate into step prompts and configuration templates ({key}).
metadata	object	No	{}	Arbitrary key-value metadata to attach to the execution result in Firestore.
sync	boolean	No	true	When true, waits for the pipeline execution to complete before returning.
redirect	boolean	No	false	When true (and sync=true), responds with an HTTP 302 Found redirect directly to the primary media output.
resultId	string	No		If provided, returns the cached result of a previously executed pipeline.

Response Schema (PipelineResult)

interface PipelineResult {
  id: string;
  pipelineId: string;
  siloId: string;
  url?: string;
  text?: string;
  data?: any;
  resultType: string; // "image" | "video" | "text" | "audio" | "composite" | "unknown"
  status: "completed" | "failed" | "pending";
  stepResults: Array<{
    stepId: string;
    stepName?: string;
    type: string;
    outputUrl?: string;
    outputText?: string;
    outputData?: any;
    computeTimeMs?: number;
    fuelCost?: number;
    error?: string;
  }>;
  totalComputeTimeMs: number;
  totalFuelCost: number;
  metadata?: Record<string, any>;
  timestamp: any;
  error?: string;
}