Building a web research agent

import { MCPClient } from '@agenite/mcp';

// Initialize MCP client with fetch server
const mcpClient = new MCPClient({
  name: 'web-research-agent',
  mcpServers: {
    // Option 1: Using the hosted MCP fetch server via SSE
    fetch: {
      url: 'https://router.mcp.so/sse/rrrcnqm8s6mf5l',
    },
    
    // Option 2: Running a local fetch server via stdio
    // fetch: {
    //   command: 'npx',
    //   args: ['-y', '@modelcontextprotocol/server-fetch'],
    // }
  },
});

// Get all tools from the MCP servers
const tools = await mcpClient.getAllTools();

console.log(`Found ${tools.length} tools from MCP servers`);

import { Agent } from '@agenite/agent';
import { OllamaProvider } from '@agenite/ollama';
import { prettyLogger } from '@agenite/pretty-logger';

// Create the agent with MCP tools
const agent = new Agent({
  name: 'web-research-agent',
  provider: new OllamaProvider({
    model: 'llama3.2',
    baseURL: 'http://localhost:11434', // Update with your Ollama endpoint
  }),
  tools: tools, // MCP tools we obtained earlier
  instructions: `
    You are a web research assistant that can search and retrieve information from the web.
    When asked about current events, facts, or online content, use the fetch tools to get accurate information.
    Always cite your sources by providing the URLs you retrieved information from.
    For complex queries, break down your research into multiple steps and explain your approach.
  `,
  middlewares: [prettyLogger()],
});

// Example research question
const result = await agent.execute({
  messages: [
    {
      role: 'user',
      content: [
        { 
          type: 'text', 
          text: 'What is Agenite and what are its main features? Please research and provide a summary.' 
        },
      ],
    },
  ],
});

// Print the result
if (result.messages.length > 0) {
  const lastMessage = result.messages[result.messages.length - 1];
  if (lastMessage) {
    console.log('Research results:', lastMessage.content);
  }
}

import { Agent } from '@agenite/agent';
import { OllamaProvider } from '@agenite/ollama';
import { MCPClient } from '@agenite/mcp';
import { prettyLogger } from '@agenite/pretty-logger';

async function main() {
  // Initialize MCP client
  const mcpClient = new MCPClient({
    name: 'web-research-agent',
    mcpServers: {
      fetch: {
        url: 'https://router.mcp.so/sse/rrrcnqm8s6mf5l',
      },
    },
  });

  try {
    // Get all tools from MCP servers
    console.log('Fetching tools from MCP servers...');
    const tools = await mcpClient.getAllTools();
    console.log(`Found ${tools.length} tools`);

    // Create the agent
    const agent = new Agent({
      name: 'web-research-agent',
      provider: new OllamaProvider({
        model: 'llama3.2',
        baseURL: 'http://localhost:11434',
      }),
      tools: tools,
      instructions: `
        You are a web research assistant that can search and retrieve information from the web.
        When asked about current events, facts, or online content, use the fetch tools to get accurate information.
        Always cite your sources by providing the URLs you retrieved information from.
        For complex queries, break down your research into multiple steps and explain your approach.
      `,
      middlewares: [prettyLogger()],
    });

    // Example research question
    const result = await agent.execute({
      messages: [
        {
          role: 'user',
          content: [
            { 
              type: 'text', 
              text: 'What is Agenite and what are its main features? Please research and provide a summary.' 
            },
          ],
        },
      ],
    });

    // Print the result
    if (result.messages.length > 0) {
      const lastMessage = result.messages[result.messages.length - 1];
      if (lastMessage) {
        console.log('Research results:', lastMessage.content);
      }
    }
  } catch (error) {
    console.error('Error during research:', error);
  }
}

main().catch((error) => {
  console.error('Error:', error);
  process.exit(1);
});

// Create an iterator for streaming responses
const iterator = agent.iterate({
  messages: [
    {
      role: 'user',
      content: [
        { 
          type: 'text', 
          text: 'What are the latest developments in artificial intelligence?' 
        },
      ],
    },
  ],
});

// Process streaming responses
for await (const chunk of iterator) {
  // Handle different types of chunks (message, tool calls, etc.)
  if (chunk.type === 'message') {
    process.stdout.write(chunk.content || '');
  } else if (chunk.type === 'tool_call') {
    console.log(`\n[Using tool: ${chunk.name}]`);
  } else if (chunk.type === 'tool_result') {
    console.log(`\n[Tool result received]`);
  }
}