@yamlresume/ai

AI-powered generation and translation for YAMLResume.

Installation

npm install @yamlresume/ai

You also need a Vercel AI SDK provider and its API key. For example:

npm install @ai-sdk/openai

Supported providers

All providers expose an OpenAI-compatible endpoint and are configured through environment variables.

| Provider | Required env variable | Default model | Optional overrides | | ---------------------------------------------- | --------------------- | ------------------- | ----------------------------------------------- | | DeepSeek | DEEPSEEK_API_KEY | deepseek-v4-flash | YAMLRESUME_AI_MODEL, YAMLRESUME_AI_BASE_URL | | Kimi (Moonshot AI) | MOONSHOT_API_KEY | kimi-k2.6 | YAMLRESUME_AI_MODEL, YAMLRESUME_AI_BASE_URL | | Ollama (local) | OLLAMA_HOST | llama3.2 | YAMLRESUME_AI_MODEL, YAMLRESUME_AI_BASE_URL | | OpenAI | OPENAI_API_KEY | gpt-5 | YAMLRESUME_AI_MODEL, YAMLRESUME_AI_BASE_URL |

The provider is inferred from the env variables in the order shown above. When no provider is detected, Kimi is used as the default so that a missing API key produces a clear configuration error.

CLI usage

The yamlresume CLI uses @yamlresume/ai under the hood. Set a provider API key and run:

# Generate a new resume
export OPENAI_API_KEY=sk-...
yamlresume ai generate --position "Software Engineer" --language en resume.yml

Override the model or endpoint when needed:

export YAMLRESUME_AI_MODEL=gpt-5
export YAMLRESUME_AI_BASE_URL=https://custom.openai.endpoint/v1
yamlresume ai generate --position "Registered Nurse" --language en resume.yml

Translate an existing resume to another locale language:

# Translate a resume from its current locale.language to Simplified Chinese
export OPENAI_API_KEY=sk-...
yamlresume ai translate --to zh-hans resume.en.yml resume.zh-hans.yml

Generate a resume

import { generateResume } from "@yamlresume/ai";
import { openai } from "@ai-sdk/openai";

const model = openai("gpt-5");

const yaml = await generateResume("Registered Nurse", "en", {
  model,
});

Translate a resume

import { translateResume } from "@yamlresume/ai";

const sourceYaml = "..."; // existing YAMLResume content
const yaml = await translateResume(sourceYaml, "en", "zh-hans", {
  model,
});

Use Kimi (Moonshot AI)

Kimi exposes an OpenAI-compatible endpoint:

import { openai } from "@ai-sdk/openai";

const kimi = openai({
  apiKey: process.env.MOONSHOT_API_KEY,
  baseURL: "https://api.moonshot.cn/v1",
});

const yaml = await generateResume("Registered Nurse", "zh-hans", {
  model: kimi("moonshot-v1-8k"),
});

Validation and retries

Both functions parse the LLM output and validate it against ResumeSchema from @yamlresume/core. If validation fails, the request is retried up to maxRetries times (default: 2).

const yaml = await generateResume("Software Engineer", "en", {
  model,
  maxRetries: 3,
});