Skip to content

text-complete.md

Latest commit

 

History

History
116 lines (96 loc) · 3.23 KB

text-complete.md

File metadata and controls

116 lines (96 loc) · 3.23 KB
Raw

Text Complete

Guide to the Text Complete plugin, including LLM adapter setup, prompt configuration, and AI-assisted text generation for resource fields.

Text Complete

This plugin allows you to auto-complete text and string fields using OpenAI GPT models via the Responses API.

Installation

pnpm add @adminforth/text-complete --save
pnpm add @adminforth/completion-adapter-openai-responses --save

Go to https://platform.openai.com/, open Dashboard -> API keys -> Create new secret key. Paste value in your .env file:

...
OPENAI_API_KEY=your_key

Add plugin installation to any text or string column.

For example let's add it for title and description in aparts resource configuration which we created in Getting Started tutorial.

//diff-add
import TextCompletePlugin from '@adminforth/text-complete';
//diff-add
import CompletionAdapterOpenAIResponses from "@adminforth/completion-adapter-openai-responses";

export const admin = new AdminForth({
  ...
  resourceId: 'aparts',
  columns: [
    ...
  ],
  plugins: [
    ...
//diff-add
    new TextCompletePlugin({
//diff-add
      fieldName: 'title',
//diff-add
      adapter: new CompletionAdapterOpenAIResponses({
//diff-add
        openAiApiKey: process.env.OPENAI_API_KEY as string,
//diff-add
        model: 'gpt-5-nano', // default "gpt-5-nano"
//diff-add
        extraRequestBodyParameters: {
//diff-add
            temperature: 0.7 //Model temperature
//diff-add
        }
//diff-add
      }),
//diff-add
    }),
//diff-add
    new TextCompletePlugin({
//diff-add
      fieldName: 'description',
//diff-add
      adapter: new CompletionAdapterOpenAIResponses({
//diff-add
        openAiApiKey: process.env.OPENAI_API_KEY as string,
//diff-add
      }),
//diff-add
      // expert: {
//diff-add
        // maxTokens: 50, // token limit to generate for each completion. 50 is default
//diff-add
        // promptInputLimit: 500, // Limit in characters of edited field to be passed to Model. 500 is default value
//diff-add
        // debounceTime: 300, // Debounce time in milliseconds. 300 is default value
//diff-add
      // }
//diff-add
    }),

  ]

  ...

});

Here is how it looks:

!alt text

Initial prompt

If you want to generate field by your own prompt, you can use initialPrompt property in plugin configuration.

    new TextCompletePlugin({
      fieldName: 'description',
//diff-add
      initialPrompt: `I need to introduce apartment '{{title}}' to people who have a pets and afraid that apartment is not pet friendly. I need short but very attractive description if it. It should have CTAs and Emojis. write \n\n once you are done.`,
//diff-add
      expert: {
//diff-add
        stop: ['\n\n'],
//diff-add
        maxTokens: 500,
//diff-add
      },
    }),

Note: {{title}} is a placeholder for the value of the field title. It will be replaced with the actual value of the field when generating the prompt. You can use any field name from your resource configuration. Note: stop is a property of the model that tells it when to stop generating text. In this case, we are telling it to stop when it sees \n\n.