@monsoft/ai

AI utilities for structured outputs, text generation, and content analysis. Built on top of the Vercel AI SDK.

Features

• Core AI Functions - Wrapper functions with centralized configuration and telemetry
  • coreGenerateObject - Generate structured JSON output
  • coreGenerateText - Generate text responses
  • coreStreamObject - Stream structured JSON output
  • coreStreamText - Stream text responses
• Model Management - Pre-configured AI models with capability metadata
• Extraction Functions - High-level functions for common AI tasks
• Telemetry - Built-in Langfuse integration for observability
• Full TypeScript support with type-safe schemas
• ESM module format

Installation

npm install @monsoft/ai
# or
pnpm add @monsoft/ai
# or
yarn add @monsoft/ai

Usage

Core AI Functions

import {
  coreGenerateObject,
  coreGenerateText,
  coreStreamText,
} from "@monsoft/ai";
import { z } from "zod";

// Generate structured output
const result = await coreGenerateObject({
  schema: z.object({
    name: z.string(),
    age: z.number(),
  }),
  system: "You are a helpful assistant",
  prompt: "Extract the name and age from: John is 25 years old",
});

console.log(result.object); // { name: 'John', age: 25 }

// Generate text
const textResult = await coreGenerateText({
  system: "You are a helpful assistant",
  prompt: "Explain what TypeScript is in one sentence",
});

console.log(textResult.text);

// Stream text (for chat applications)
const stream = coreStreamText({
  system: "You are a helpful assistant",
  messages: [{ role: "user", content: "Hello!" }],
  smoothStreaming: true,
});

for await (const chunk of stream.textStream) {
  process.stdout.write(chunk);
}

Subpath Imports

The package provides subpath exports for more granular imports:

// Core functions only
import { coreGenerateObject } from "@monsoft/ai/core";

// Functions
import { extractOrganizationProfile } from "@monsoft/ai/functions";

// Models
import { AVAILABLE_MODELS, getModelById } from "@monsoft/ai/models";

// Prompts
import { organizationExtractionPrompt } from "@monsoft/ai/prompts";

// Telemetry
import { telemetryConfig } from "@monsoft/ai/telemetry";

Model Selection

import {
  AVAILABLE_MODELS,
  getModelById,
  getModelsByTier,
  getRecommendedModels,
  DEFAULT_CHAT_MODEL_ID,
} from "@monsoft/ai/models";

// Get all available models
console.log(AVAILABLE_MODELS);

// Get a specific model
const model = getModelById("gpt-4o");

// Get models by tier (economy, standard, premium)
const premiumModels = getModelsByTier("premium");

// Get recommended models for a capability
const recommendedForChat = getRecommendedModels("chat");

Organization Profile Extraction

import { extractOrganizationProfile } from "@monsoft/ai/functions";

const profile = await extractOrganizationProfile(websiteContent, {
  modelId: "gpt-4o-mini", // Optional: override default model
});

console.log(profile);
// {
//   name: 'Company Name',
//   description: '...',
//   industry: '...',
//   ...
// }

API Reference

Core Functions

coreGenerateObject<T>(options)

Generate structured JSON output matching a Zod schema.

| Option | Type | Description | | --------- | -------------- | ------------------------------------------------ | | schema | ZodSchema<T> | Zod schema for the output structure | | system | string | System prompt | | prompt | string | User prompt | | modelId | string | Optional model ID (uses default if not provided) |

Returns: Promise<GenerateObjectResult<T>>

coreGenerateText(options)

Generate text responses.

| Option | Type | Description | | ---------------------- | --------------------------- | ------------------------------- | | system | string | System prompt | | prompt or messages | string or CoreMessage[] | Input prompt or message history | | modelId | string | Optional model ID |

Returns: Promise<GenerateTextResult>

coreStreamText(options)

Stream text responses (for real-time chat).

| Option | Type | Description | | ----------------- | --------------- | ----------------------- | | system | string | System prompt | | messages | CoreMessage[] | Message history | | smoothStreaming | boolean | Enable smooth streaming | | modelId | string | Optional model ID |

Returns: StreamTextResult

coreStreamObject<T>(options)

Stream structured JSON output.

| Option | Type | Description | | --------- | -------------- | ------------------------- | | schema | ZodSchema<T> | Zod schema for the output | | system | string | System prompt | | prompt | string | User prompt | | modelId | string | Optional model ID |

Returns: StreamObjectResult<T>

Model Types

type AIModel = {
  id: string;
  name: string;
  provider: "openai" | "anthropic" | "google";
  tier: "economy" | "standard" | "premium";
  capabilities: ("chat" | "extraction" | "analysis" | "coding")[];
  contextWindow: number;
  maxOutput: number;
};

Environment Variables

The package uses the following environment variables:

| Variable | Description | | --------------------- | ----------------------------------- | | OPENAI_API_KEY | OpenAI API key | | LANGFUSE_SECRET_KEY | Langfuse secret key (for telemetry) | | LANGFUSE_PUBLIC_KEY | Langfuse public key (for telemetry) |

License

MIT