@wener/ai

AI utilities, provider-agnostic schemas, and MCP (Model Context Protocol) server definitions.

Install

pnpm add @wener/ai

Modules

AI Provider Schemas

Zod-based schema definitions for major AI providers, covering Chat Completions, Responses API, streaming, and tool calling.

Each provider offers two schema flavors:

• Strict schemas (@wener/ai/<provider>/schema) — precise z.object() definitions for strong validation
• Generic/loose schemas (@wener/ai/<provider>) — z.looseObject() definitions that allow extra properties to pass through, suitable for protocol conversion, proxying, and auditing

// Generic (loose) — good for proxies and protocol conversion
import {
  CreateChatCompletionRequestSchema,
  CreateChatCompletionResponseSchema,
  CreateChatCompletionStreamChunkSchema,
  CreateResponseRequestSchema,
  UsageSchema,
} from '@wener/ai/openai';

// Strict — good for direct API interaction
import {
  ChatCompletionRequestSchema,
  ChatCompletionResponseSchema,
  ChatCompletionChunkSchema,
} from '@wener/ai/openai/schema';

Supported providers:

| Import path | Provider | Notes | | --- | --- | --- | | @wener/ai/openai | OpenAI | Chat Completions + Responses API | | @wener/ai/openai/schema | OpenAI | Strict schemas | | @wener/ai/anthropic | Anthropic | Claude-compatible schemas | | @wener/ai/anthropic/schema | Anthropic | Strict schemas | | @wener/ai/google | Google | Gemini-compatible schemas | | @wener/ai/google/schema | Google | Strict schemas |

MCP Server Definitions

Reusable, factory-based MCP server definitions that follow a standard McpServerDef pattern. Each server provides tools and resources accessible via the Model Context Protocol.

import { defineMcpServer, type McpServerDef, type McpServerInstance } from '@wener/ai/mcp';

Prometheus

PromQL-based metrics querying with instant and range query support.

import { createPrometheusMcpServer } from '@wener/ai/mcp/prometheus';

const { server, close } = createPrometheusMcpServer({
  url: 'http://prometheus:9090',
  username: 'admin', // optional
  password: 'secret', // optional
});

Tools: query (instant & range PromQL queries), list_metrics, metric_metadata, label_values

Resources: promql-guide — PromQL syntax reference

Tencent CLS

Tencent Cloud Log Service for log search, analysis, and pattern clustering.

import { createTencentClsMcpServer } from '@wener/ai/mcp/tencent-cls';

const { server, close } = createTencentClsMcpServer({
  clientId: 'AKID...',
  clientSecret: '...',
  region: 'ap-shanghai', // optional, defaults to ap-shanghai
});

Tools: search (CQL log search), cluster_logs (Drain3-based pattern clustering), log_context (surrounding context), list_topics, describe_topic, get_histograms

Resources: search-guide — CQL search syntax reference

SQL Database

Multi-dialect SQL database access via Kysely.

import { createSqlMcpServer } from '@wener/ai/mcp/sql';

const { server, close } = createSqlMcpServer({
  url: 'mysql://user:pass@localhost:3306/mydb',
});

Tools: query (execute SQL), list_tables, describe_table

Supported dialects: MySQL, PostgreSQL, SQLite, MSSQL

Peer dependencies: kysely, and the dialect driver (mysql2, tedious, etc.)

Feishu / Lark

Messaging and document operations for Feishu (飞书) / Lark.

import { createFeishuMcpServer } from '@wener/ai/mcp/feishu';

const { server, close } = createFeishuMcpServer({
  appId: 'cli_...',
  appSecret: '...',
  domain: 'feishu', // 'feishu' | 'lark' | custom domain
});

Tools: IM (messaging), document operations

Peer dependency: @larksuiteoapi/node-sdk

Apollo Config

Read application configurations from Apollo Config configuration center.

import { createApolloConfigMcpServer } from '@wener/ai/mcp/apolloconfig';

const { server, close } = createApolloConfigMcpServer({
  url: 'http://apollo-config:8080',
  appId: 'my-app',
  cluster: 'default', // optional
});

Tools: Configuration reading and namespace listing

MCP Relay

Proxy requests to another MCP server, supporting both Streamable HTTP and SSE transports.

import { createRelayMcpServer } from '@wener/ai/mcp/relay';

const { server, close } = createRelayMcpServer({
  url: 'http://other-mcp-server:3000/mcp',
  transport: 'http', // 'http' | 'sse'
  headers: { Authorization: 'Bearer ...' }, // optional
});

Tools: relay_list_tools, relay_call_tool, relay_list_resources, relay_read_resource

Custom MCP Server Definition

Use defineMcpServer to create your own server definition that integrates with @wener/mcps infrastructure:

import { defineMcpServer, type McpServerInstance } from '@wener/ai/mcp';

interface MyOptions {
  apiKey: string;
}

export const MyMcpServerDef = defineMcpServer<MyOptions>({
  name: 'my-service',
  title: 'My Service',
  description: 'My custom MCP server',
  version: '1.0.0',
  tags: ['custom'],
  validateOptions(options) {
    if (!options.apiKey) return { valid: false, error: 'Missing apiKey' };
    return { valid: true };
  },
  getCacheKey(options) {
    return `my::${options.apiKey}`;
  },
  create(options): McpServerInstance {
    const server = new McpServer({ name: 'my-service', version: '1.0.0' });
    // register tools, resources...
    return { server, async close() { await server.close(); } };
  },
});

Peer Dependencies

Heavy or optional dependencies are declared as optional peer dependencies to avoid eager loading:

| Dependency | Required by | | --- | --- | | kysely | @wener/ai/mcp/sql | | mysql2 | SQL MCP with MySQL dialect | | tedious | SQL MCP with MSSQL dialect | | tarn | Connection pooling for SQL | | @larksuiteoapi/node-sdk | @wener/ai/mcp/feishu |

License

ISC