build-code

Converts TypeScript functions to Breadboard runJavascript components. Code is bundled and schema is automaticaly generated from TypeScript types.

Usage

1. Install:

   npm i @breadboard-ai/build-code

2. Create a module with the following exports:

   • A type/interface called Inputs
   • A type/interface called Outputs
   • A function called run

   For example, the file src/js-components/is-foo.ts:

   export interface Inputs {
     str: string;
   }
   
   export interface Outputs {
     isFoo: boolean;
   }
   
   export const run = ({ str }: Inputs): Outputs => {
     return { isFoo: strIsFoo(str) };
   };
   
   // Note that this function is outside the scope of `run`. This would
   // normally be a problem with the the standard Breadboard `code`
   // function, but it's OK here because we bundle! Imports work too.
   function strIsFoo(str: string): boolean {
     return str === "foo";
   }

3. Use the build-code binary:

   {
     "scripts": {
       "generate:js-components": "build-code --tsconfig=./tsconfig.json --out=./src/generated ./src/js-components/*.ts"
     }
   }

4. Import the generated module, which in this case will be at src/generated/is-foo.ts, and use it with the Breadboard Build API. The module will contain a strongly-typed function based on the name of the source module:

   import { isFoo } from "./generated/is-foo.js";
   import { input, board } from "@breadboard-ai/build";
   
   const str = input();
   const isFooInst = isFoo({ str });
   export const myBoard = board({
     inputs: { str },
     outputs: { isFoo: isFooInst.outputs.isFoo },
   });

5. Upon serialization to BGL, a runJavascript component will be created with bundled code and automatically generated schemas:

   {
     "nodes": {
       "id": "runJavascript-0",
       "type": "runJavascript",
       "configuration": {
         "raw": true,
         "name": "run",
         "code": "export const run = ({ str }) => {\n  return { isFoo: strIsFoo(str) };\n};\n\nfunction strIsFoo(str) {\n  return str === \"foo\";\n}\n",
         "inputSchema": {
           "type": "object",
           "properties": {
             "str": {
               "type": "string"
             }
           },
           "required": ["str"],
           "additionalProperties": false
         },
         "outputSchema": {
           "type": "object",
           "properties": {
             "isFoo": {
               "type": "boolean"
             }
           },
           "required": ["isFoo"],
           "additionalProperties": false
         }
       }
     }
   }