@yamlresume/ai

AI-powered generation and translation for YAMLResume.

Installation

npm install @yamlresume/ai
Copy

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

npm install @ai-sdk/openai
Copy

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
Copy

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
Copy

Generate a resume

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

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

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

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({
  position: "Registered Nurse",
  language: "zh-hans",
  model: kimi("moonshot-v1-8k"),
});
Copy

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({
  position: "Software Engineer",
  language: "en",
  model,
  maxRetries: 3,
});
Copy