@ebowwa/teammates-glm

GLM 4.7 provider plugin for @ebowwa/teammates. Enables multi-agent coordination using GLM models.

Features

• GLM Integration: Uses existing @ebowwa/ai GLM client
• Composable: Plugin architecture - no changes to core teammates package
• Model Selection: Support for GLM-4.7, GLM-4.6, GLM-4.5, GLM-4.5-air
• Streaming: Full support for streaming responses
• Document Analysis: Pre-built workflows for common tasks

Installation

bun add @ebowwa/teammates-glm

Quick Start

import { GLMProvider } from '@ebowwa/teammates-glm';

// Create provider
const provider = new GLMProvider({
  defaultModel: 'GLM-4.7',
  defaultTemperature: 0.7,
});

// Simple chat
const result = await provider.chat('Explain quantum computing');
console.log(result.content);

// Create teammates
const team = await provider.createTeam('research', [
  {
    teamName: 'research',
    name: 'analyzer',
    prompt: 'Analyze the data',
    color: 'blue',
  },
]);

Environment Setup

Set your GLM API key:

export Z_AI_API_KEY='your-api-key'
# or
export ZAI_API_KEY='your-api-key'
# or
export GLM_API_KEY='your-api-key'

Or configure via Doppler for production use.

API Reference

GLMProvider

// Constructor
const provider = new GLMProvider({
  apiKey?: string,          // Defaults to Z_AI_API_KEY
  baseURL?: string,         // Defaults to GLM endpoint
  defaultModel?: string,    // Defaults to 'GLM-4.7'
  defaultTemperature?: number,  // Defaults to 0.7
  defaultMaxTokens?: number,
});

// Chat methods
await provider.chat(prompt, options?)
await provider.chatMessages(messages, options?)

// Streaming
for await (const chunk of provider.chatStream(prompt, options?)) {
  console.log(chunk);
}

// Teammate creation
await provider.createTeammate({
  teamName: 'my-team',
  name: 'analyst',
  prompt: 'Your task...',
  model?: 'GLM-4.7',
  systemPrompt?: 'Instructions...',
  color?: 'blue',
  temperature?: 0.7,
  maxTokens?: 4096,
});

// Batch teammate creation
await provider.createTeam(teamName, [teammateOptions, ...], description?)

// Document analysis workflow
await provider.analyzeDocument(documentPath, teamName?)

Example Workflows

Document Analysis Team

import { GLMProvider } from '@ebowwa/teammates-glm';

const provider = new GLMProvider();

const team = await provider.analyzeDocument(
  '/path/to/research.md',
  'my-analysis'
);

// Team will have 2 teammates:
// - analyzer (blue): Extracts key information
// - summarizer (green): Creates executive summary

Custom Team with Multiple Agents

const provider = new GLMProvider();

const team = await provider.createTeam(
  'code-review',
  [
    {
      teamName: 'code-review',
      name: 'security-reviewer',
      prompt: 'Review code for security vulnerabilities',
      color: 'red',
      model: 'GLM-4.7',
    },
    {
      teamName: 'code-review',
      name: 'performance-reviewer',
      prompt: 'Review code for performance issues',
      color: 'yellow',
      model: 'GLM-4.5-air',  // Faster model
    },
    {
      teamName: 'code-review',
      name: 'style-reviewer',
      prompt: 'Review code for style and conventions',
      color: 'green',
    },
  ],
  'Multi-perspective code review'
);

Streaming Chat

const provider = new GLMProvider();

for await (const chunk of provider.chatStream('Write a haiku about AI')) {
  process.stdout.write(chunk);
}

Dependencies

This plugin composes with:

• @ebowwa/ai - GLM client implementation
• @ebowwa/teammates - Multi-agent coordination
• @ebowwa/codespaces-types - Shared type definitions

Architecture

┌─────────────────────────────────────┐
│     @ebowwa/teammates-glm          │
│     (GLM Provider Plugin)           │
└──────────────┬──────────────────────┘
               │
               ├──► @ebowwa/ai (GLMClient)
               │        └──► Z.AI API
               │
               └──► @ebowwa/teammates
                        └──► Team coordination

License

MIT

Author

Ebowwa Labs [email protected]