llmjs2 1.6.1 → 2.0.0

package/core/cli.js

@@ -1,275 +0,0 @@
-#!/usr/bin/env node
-
-const { app } = require('./server');
-const { router } = require('./router');
-const { Graph } = require('../chain');
-const yaml = require('yaml');
-const fs = require('fs');
-const path = require('path');
-
-class CLI {
-  constructor() {
-    this.args = process.argv.slice(2);
-    this.options = this.parseArgs();
-  }
-
-  parseArgs() {
-    const options = {
-      mode: 'start',
-      start: false,
-      port: 3000,
-      host: 'localhost',
-      config: null,
-      flowPath: null,
-      inputStr: '',
-      help: false
-    };
-
-    if (this.args.length > 0 && this.args[0] === 'run') {
-      options.mode = 'run';
-      if (this.args.length < 2) {
-        console.error('Missing flow file for run command');
-        this.showHelp();
-        process.exit(1);
-      }
-      options.flowPath = this.args[1];
-      options.inputStr = this.args.slice(2).join(' ');
-      return options;
-    }
-
-    for (let i = 0; i < this.args.length; i++) {
-      const arg = this.args[i];
-
-      switch (arg) {
-        case '-S':
-          options.start = true;
-          break;
-        case '-p':
-        case '--port':
-          options.port = parseInt(this.args[++i], 10);
-          break;
-        case '-H':
-        case '--host':
-          options.host = this.args[++i];
-          break;
-        case '-c':
-        case '--config':
-          options.config = this.args[++i];
-          break;
-        case '-h':
-        case '--help':
-          options.help = true;
-          break;
-        default:
-          if (arg.startsWith('-')) {
-            console.error(`Unknown option: ${arg}`);
-            this.showHelp();
-            process.exit(1);
-          } else {
-            console.error(`Unexpected positional argument: ${arg}`);
-            this.showHelp();
-            process.exit(1);
-          }
-      }
-    }
-
-    return options;
-  }
-
-  parseInputs(inputStr) {
-    const inputs = {};
-    if (!inputStr.trim()) return inputs;
-
-    const pairs = inputStr.split(';').map((s) => s.trim()).filter((s) => s);
-    for (const pair of pairs) {
-      const [key, value] = pair.split('=');
-      if (key && value) {
-        const cleanValue = value.replace(/^["']|["']$/g, '');
-        inputs[key.trim()] = cleanValue;
-      }
-    }
-    return inputs;
-  }
-
-  async runWorkflow() {
-    try {
-      const config = JSON.parse(fs.readFileSync(this.options.flowPath, 'utf8'));
-      const workflow = Graph.load(config).compile();
-      const inputs = this.parseInputs(this.options.inputStr);
-
-      console.log(`Running workflow: ${config.name || this.options.flowPath}`);
-      console.log(`Inputs: ${JSON.stringify(inputs, null, 2)}`);
-
-      const result = await workflow.run(inputs);
-
-      console.log('Workflow completed successfully');
-      console.log(JSON.stringify(result, null, 2));
-    } catch (error) {
-      console.error('Failed to run workflow:', error.message);
-      process.exit(1);
-    }
-  }
-
-  showHelp() {
-    console.log(`
-llmjs2 - OpenAI-Compatible API Server and Workflow Runner
-
-USAGE:
-  llmjs2 -S [options]          Start the API server
-  llmjs2 run <flow> [inputs]   Run a workflow from JSON file
-
-DESCRIPTION:
-  Starts an OpenAI-compatible API server with intelligent routing and guardrails,
-  or runs workflows defined in JSON files with agent steps and conditional logic.
-
-OPTIONS:
-  -S                           Start the server
-  -c, --config <file>          YAML config file with models, guardrails, and routing
-  -p, --port <port>            Port to listen on (default: 3000)
-  -H, --host <host>            Host to bind to (default: localhost)
-  -h, --help                   Show this help message
-
-EXAMPLES:
-  llmjs2 -S
-  llmjs2 -S --config core/config.yaml
-  llmjs2 -S --port 8080 --host 0.0.0.0
-  llmjs2 run chain/workflow-example.json
-  llmjs2 run chain/workflow-example.json input="hello";reference="world"
-`);
-  }
-
-  loadConfig() {
-    if (!this.options.config) {
-      return this.createDefaultConfig();
-    }
-
-    const configPath = path.resolve(this.options.config);
-    if (!fs.existsSync(configPath)) {
-      throw new Error(`Configuration file not found: ${configPath}`);
-    }
-
-    const configContent = fs.readFileSync(configPath, 'utf8');
-    const config = yaml.parse(configContent);
-    return this.resolveEnvVars(config);
-  }
-
-  createDefaultConfig() {
-    return {
-      model_list: [
-        {
-          model_name: 'default',
-          llm_params: {
-            model: process.env.OLLAMA_DEFAULT_MODEL || 'ollama/minimax-m2.5:cloud',
-            api_key: process.env.OLLAMA_API_KEY,
-            api_base: process.env.OLLAMA_BASE_URL
-          }
-        }
-      ],
-      guardrails: [],
-      router_settings: {
-        routing_strategy: 'default',
-        allow_unsafe_guardrails: false
-      }
-    };
-  }
-
-  resolveEnvVars(obj) {
-    if (typeof obj === 'string') {
-      if (obj.startsWith('os.environ/')) {
-        const envVar = obj.replace('os.environ/', '');
-        return process.env[envVar] || obj;
-      }
-      return obj;
-    }
-
-    if (Array.isArray(obj)) {
-      return obj.map((item) => this.resolveEnvVars(item));
-    }
-
-    if (obj && typeof obj === 'object') {
-      const resolved = {};
-      for (const [key, value] of Object.entries(obj)) {
-        resolved[key] = this.resolveEnvVars(value);
-      }
-      return resolved;
-    }
-
-    return obj;
-  }
-
-  createRouter(config) {
-    const settings = config.router_settings || {};
-    const route = router(config.model_list || [], settings.routing_strategy || 'default', {
-      allowUnsafeGuardrails: Boolean(settings.allow_unsafe_guardrails)
-    });
-
-    if (Array.isArray(config.guardrails) && config.guardrails.length > 0) {
-      route.setGuardrails(config.guardrails);
-    }
-
-    return route;
-  }
-
-  async run() {
-    if (this.options.help) {
-      this.showHelp();
-      return;
-    }
-
-    if (this.options.mode === 'run') {
-      await this.runWorkflow();
-      return;
-    }
-
-    if (!this.options.start && this.args.length === 0) {
-      this.options.start = true;
-    }
-
-    if (!this.options.start) {
-      this.showHelp();
-      return;
-    }
-
-    try {
-      const config = this.loadConfig();
-      const route = this.createRouter(config);
-
-      console.log('Starting llmjs2 server...');
-      console.log(`Configuration: ${this.options.config || 'default'}`);
-      console.log(`Models: ${route.getAvailableModels().join(', ')}`);
-      console.log(`Strategy: ${(config.router_settings || {}).routing_strategy || 'default'}`);
-      console.log(`Guardrails: ${(config.guardrails || []).length}`);
-
-      app.use(route);
-      app.listen(this.options.port, this.options.host);
-    } catch (error) {
-      console.error('Failed to start server:', error.message);
-      process.exit(1);
-    }
-  }
-}
-
-function loadEnvFile() {
-  const envPath = path.join(process.cwd(), '.env');
-  if (!fs.existsSync(envPath)) {
-    return;
-  }
-
-  const envContent = fs.readFileSync(envPath, 'utf8');
-  const envVars = envContent.split('\n').filter((line) => line.trim() && !line.startsWith('#'));
-
-  envVars.forEach((line) => {
-    const [key, ...rest] = line.split('=');
-    const value = rest.join('=');
-    if (key && value) {
-      process.env[key.trim()] = value.trim();
-    }
-  });
-}
-
-loadEnvFile();
-const cli = new CLI();
-cli.run().catch((error) => {
-  console.error(error.message || String(error));
-  process.exit(1);
-});

package/core/config.yaml

@@ -1,149 +0,0 @@
-# Sample configuration for llmjs2 server
-# This file demonstrates all available configuration options
-
-# Model list defines available models and their providers
-model_list:
-
-
-  # Ollama models
-  - model_name: glm-5.1:cloud
-    llm_params:
-      model: ollama/glm-5.1:cloud
-
-  - model_name: minimax-m2.5
-    llm_params:
-      model: ollama/minimax-m2.5:cloud
-  # OpenRouter models (multiple with same name for load balancing)
-
-
-  - model_name: free-model
-    llm_params:
-      model: openrouter/openrouter/free
-
-
-# Guardrails for request/response processing
-guardrails:
-  # Content filtering before LLM calls
-  - name: content_filter
-    mode: pre_call
-    code: |
-      (processId, input) => {
-        const { model, messages } = input;
-
-        console.log(`[${processId}] Filtering content for model: ${model}`);
-
-        // Filter inappropriate content
-        const filteredMessages = messages.map(msg => {
-          if (msg.role === 'user' && msg.content) {
-            // Basic profanity filter
-            const filtered = msg.content
-              .replace(/badword/gi, '****')
-              .replace(/inappropriate/gi, '****');
-
-            return { ...msg, content: filtered };
-          }
-          return msg;
-        });
-
-        return { model, messages: filteredMessages };
-      }
-
-  # Rate limiting
-  - name: rate_limiter
-    mode: pre_call
-    code: |
-      (processId, input) => {
-        // Simple in-memory rate limiting (use Redis in production)
-        const now = Date.now();
-        const windowMs = 60000; // 1 minute
-        const maxRequests = 3; // 10 requests per minute
-
-        if (!global.rateLimitStore) {
-          global.rateLimitStore = new Map();
-        }
-
-        const userId = 'default_user'; // In real app, get from auth
-        const userRequests = global.rateLimitStore.get(userId) || [];
-
-        // Clean old requests
-        const recentRequests = userRequests.filter(time => now - time < windowMs);
-
-        if (recentRequests.length >= maxRequests) {
-          throw new Error(`Rate limit exceeded. Maximum ${maxRequests} requests per minute.`);
-        }
-
-        recentRequests.push(now);
-        global.rateLimitStore.set(userId, recentRequests);
-
-        console.log(`[${processId}] Rate limit check passed: ${recentRequests.length}/${maxRequests}`);
-        return input;
-      }
-
-  # Request logging
-  - name: request_logger
-    mode: pre_call
-    code: |
-      (processId, input) => {
-        const { model, messages } = input;
-
-        const logData = {
-          processId,
-          timestamp: new Date().toISOString(),
-          model,
-          messageCount: messages.length,
-          totalChars: messages.reduce((sum, m) => sum + (m.content?.length || 0), 0)
-        };
-
-        console.log(`[REQUEST] ${JSON.stringify(logData)}`);
-        return input;
-      }
-
-  # Response logging and filtering
-  - name: response_logger
-    mode: post_call
-    code: |
-      (processId, result) => {
-        const logData = {
-          processId,
-          timestamp: new Date().toISOString(),
-          resultType: typeof result,
-          resultLength: typeof result === 'string' ? result.length : 'object'
-        };
-
-        console.log(`[RESPONSE] ${JSON.stringify(logData)}`);
-
-        // Basic response filtering
-        if (typeof result === 'string') {
-          const filtered = result.replace(/sensitive/gi, '[FILTERED]');
-          return filtered;
-        }
-
-        return result;
-      }
-
-  # Performance monitoring
-  - name: performance_monitor
-    mode: post_call
-    code: |
-      (processId, result) => {
-        const endTime = Date.now();
-        // Note: In a real implementation, you'd track start time from pre_call
-
-        console.log(`[${processId}] Processing completed in ${Date.now() - (global.startTimes?.[processId] || Date.now())}ms`);
-
-        // Clean up
-        if (global.startTimes) {
-          delete global.startTimes[processId];
-        }
-
-        return result;
-      }
-
-# Router configuration
-router_settings:
-  routing_strategy: default  # Options: default (sequential), random, sequential
-
-# Optional: Custom server settings (can also be set via CLI or env vars)
-# server_settings:
-#   port: 3000
-#   host: localhost

package/core/docs/BASIC_USAGE.md

@@ -1,62 +0,0 @@
-# Basic Usage Guide
-
-This guide keeps llmjs2 usage minimal with only two completion patterns.
-
-## Pattern 1: Simple Prompt
-
-```javascript
-import { completion } from 'llmjs2';
-
-const response = await completion('Explain quantum physics in simple terms');
-console.log(response);
-```
-
-Behavior:
-- Reads configured providers from environment variables or constructor config.
-- Picks one available provider randomly for each request.
-- Uses that provider's default model.
-- Returns the assistant text directly.
-
-## Pattern 2: Full Object API
-
-```javascript
-import { completion } from 'llmjs2';
-
-const response = await completion({
-  model: 'ollama/minimax-m2.5:cloud',
-  host: 'https://ollama.com/api/chat', // optional - use env or default one
-  messages: [
-    { role: 'system', content: 'You are a helpful coding assistant.' },
-    { role: 'user', content: 'Write a function to reverse a string in JavaScript.' }
-  ],
-  apiKey: 'your-api-key', // optional - uses env vars if not provided
-  tools: [] // optional
-});
-
-console.log(response);
-```
-
-Notes:
-- `model` format must be `provider/model_name`.
-- `host` is optional and overrides provider base URL for this request.
-- `apiKey` is optional and overrides provider API key for this request.
-- `tools` is optional for function calling.
-
-## Environment Variables
-
-```bash
-# API Keys
-OPENAI_API_KEY=...
-OPEN_ROUTER_API_KEY=...
-OLLAMA_API_KEY=...
-
-# Default Models
-OPENAI_DEFAULT_MODEL=gpt-3.5-turbo
-OPEN_ROUTER_DEFAULT_MODEL=openrouter/free
-OLLAMA_DEFAULT_MODEL=minimax-m2.5:cloud
-
-# Optional Base URLs
-OPENAI_BASE_URL=https://api.openai.com/v1
-OPEN_ROUTER_BASE_URL=https://openrouter.ai/api/v1/chat/completions
-OLLAMA_BASE_URL=https://ollama.com/api/chat
-```

package/core/docs/CLI.md

@@ -1,104 +0,0 @@
-# CLI Guide
-
-Use the llmjs2 CLI to run the OpenAI-style API server or execute JSON workflows.
-
-## Install
-
-```bash
-npm install -g llmjs2
-```
-
-## Commands
-
-```bash
-llmjs2                     # Start server with defaults
-llmjs2 -S                  # Explicitly start server
-llmjs2 -S --config core/config.yaml
-llmjs2 -S --port 8080 --host 0.0.0.0
-llmjs2 run chain/workflow-example.json
-llmjs2 run chain/workflow-example.json input="hello";reference="world"
-llmjs2 --help
-```
-
-## Start Server
-
-The server listens on POST /v1/chat/completions and returns an OpenAI-style response body.
-
-```bash
-llmjs2 -S --config core/config.yaml --port 3000
-```
-
-### Options
-
-- -S: Start the API server.
-- -c, --config <file>: YAML config file with models, guardrails, and routing.
-- -p, --port <port>: Port to listen on (default: 3000).
-- -H, --host <host>: Host to bind to (default: localhost).
-- -h, --help: Show help.
-
-## Workflow Runner
-
-Run a workflow JSON file from the chain subsystem.
-
-```bash
-llmjs2 run chain/workflow-example.json
-```
-
-Pass workflow inputs as key=value pairs separated by semicolons:
-
-```bash
-llmjs2 run chain/workflow-example.json input="hello";reference="world"
-```
-
-## Configuration File
-
-Example config:
-
-```yaml
-model_list:
-  - model_name: default
-    llm_params:
-      model: ollama/minimax-m2.5:cloud
-      api_key: os.environ/OLLAMA_API_KEY
-      api_base: os.environ/OLLAMA_BASE_URL
-
-guardrails:
-  - name: request_logger
-    mode: pre_call
-    code: |
-      (processId, input) => {
-        console.log(`[${processId}] incoming messages=${input.messages.length}`);
-        return input;
-      }
-
-router_settings:
-  routing_strategy: default
-  allow_unsafe_guardrails: false
-```
-
-Notes:
-- String guardrails are disabled by default.
-- To allow string guardrails, set router_settings.allow_unsafe_guardrails=true or set LLMJS2_ALLOW_UNSAFE_GUARDRAILS=true.
-
-## Environment Variables
-
-```bash
-OPENAI_API_KEY=...
-OPEN_ROUTER_API_KEY=...
-OLLAMA_API_KEY=...
-
-OPENAI_DEFAULT_MODEL=gpt-4o-mini
-OPEN_ROUTER_DEFAULT_MODEL=openrouter/free
-OLLAMA_DEFAULT_MODEL=minimax-m2.5:cloud
-
-OPENAI_BASE_URL=https://api.openai.com/v1
-OPEN_ROUTER_BASE_URL=https://openrouter.ai/api/v1/chat/completions
-OLLAMA_BASE_URL=https://ollama.com/api/chat
-
-PORT=3000
-HOST=localhost
-```
-
-## .env Support
-
-If a .env file exists in the current directory, llmjs2 loads it automatically before running.

package/core/docs/GET_STARTED.md

@@ -1,129 +0,0 @@
-# Getting Started with llmjs2
-
-Welcome to llmjs2! This guide will get you up and running in just 5 minutes.
-
-## Prerequisites
-
-Before you begin, make sure you have:
-
-- Node.js version 14.0.0 or higher
-- API keys for at least one LLM provider (Ollama or OpenRouter)
-
-## Installation
-
-### Option 1: Global Installation (Recommended)
-
-Install llmjs2 globally to use it from anywhere:
-
-```bash
-npm install -g llmjs2
-```
-
-### Option 2: Local Installation
-
-Install in your project directory:
-
-```bash
-npm install llmjs2
-```
-
-## Quick Setup
-
-### Step 1: Set API Keys
-
-Set your API keys as environment variables:
-
-```bash
-# For Ollama (get your key from https://ollama.com)
-export OLLAMA_API_KEY=your_ollama_api_key_here
-
-# For OpenRouter (get your key from https://openrouter.ai)
-export OPEN_ROUTER_API_KEY=your_OPEN_ROUTER_API_KEY_here
-```
-
-### Step 2: Your First Completion
-
-Create a simple test script:
-
-```javascript
-// hello-world.js
-import { completion } from 'llmjs2';
-
-async function helloWorld() {
-  try {
-    const response = await completion('Hello! Introduce yourself in one sentence.');
-    console.log('Response:', response);
-  } catch (error) {
-    console.error('Error:', error.message);
-  }
-}
-
-helloWorld();
-```
-
-Run it:
-
-```bash
-node hello-world.js
-```
-
-**Expected Output:**
-
-```
-Response: Hello! I'm an AI assistant powered by llmjs2, designed to help with various tasks and provide intelligent responses.
-```
-
-### Step 3: Test Different Models
-
-Try using specific models:
-
-```javascript
-// test-models.js
-import { completion } from 'llmjs2';
-
-async function testModels() {
-  try {
-    // Test Ollama
-    if (process.env.OLLAMA_API_KEY) {
-      const ollamaResponse = await completion('ollama/minimax-m2.5:cloud', 'What is AI?');
-      console.log('Ollama:', ollamaResponse);
-    }
-
-    // Test OpenRouter
-    if (process.env.OPEN_ROUTER_API_KEY) {
-      const openrouterResponse = await completion('openrouter/openrouter/free', 'What is machine learning?');
-      console.log('OpenRouter:', openrouterResponse);
-    }
-  } catch (error) {
-    console.error('Error:', error.message);
-  }
-}
-
-testModels();
-```
-
-## What's Next?
-
-Now that you have llmjs2 working, you can:
-
-1. **[Learn Basic Usage](BASIC_USAGE.md)** - Explore different API patterns and configuration options
-2. **[Run as a Server](SERVER_MODE.md)** - Set up an OpenAI-compatible API server
-3. **[Use the CLI](CLI.md)** - Manage servers and configurations from the command line
-
-## Supported Providers
-
-llmjs2 supports multiple LLM providers:
-
-- **Ollama** - Connect to Ollama's cloud API
-- **OpenRouter** - Access various models through OpenRouter
-
-## Need Help?
-
-If you encounter any issues:
-
-1. Check that your API keys are valid and have credits
-2. Ensure you're using the correct model format: `provider/model_name`
-3. Try the examples in this guide first
-4. Check your internet connection
-
-For more detailed usage, see the other guides in this documentation.