TypeScript Guides

Build with Agents

This example shows how to use the General Agent with tools for research, browsing, and multi-step conversations. The General Agent is the default agent in Spaces.

Key features:

  • Use prebuilt tools like web browsing and search
  • Full TypeScript support with type safety
  • Support for multi-step conversations
  • Custom system prompts for specialized behavior
1

Install Package

$pnpm add athena-intelligence
2

Set Up Client

1import type { AthenaIntelligence } from 'athena-intelligence';
2import { AthenaIntelligenceClient } from 'athena-intelligence';
3
4const client = new AthenaIntelligenceClient({
5  apiKey: process.env.ATHENA_API_KEY,
6});
3

Basic Agent Usage

Start with a simple search request to understand the basic pattern:

1const request: AthenaIntelligence.GeneralAgentRequest = {
2  config: {
3    enabled_tools: ['search'],
4    model: 'gpt-4-turbo-preview',
5  },
6  messages: [
7    {
8      content: 'Search for information about Athena Intelligence and summarize what you find.',
9      role: 'user',
10      type: 'user',
11    },
12  ],
13};
14
15const response = await client.agents.general.invoke(request);
16console.log('Response received:', JSON.stringify(response, null, 2));
4

Multi-Tool Workflows

Combine multiple tools with custom system prompts for more sophisticated tasks:

1const multiToolRequest: AthenaIntelligence.GeneralAgentRequest = {
2  config: {
3    enabled_tools: ['search', 'browse'],
4    system_prompt: 'You are a helpful research assistant. Use search and browse tools to find comprehensive information.',
5    model: 'gpt-4-turbo-preview',
6  },
7  messages: [
8    {
9      content: 'Research the latest developments in AI and provide a detailed summary with sources.',
10      role: 'user',
11      type: 'user',
12    },
13  ],
14};
15
16const multiToolResponse = await client.agents.general.invoke(multiToolRequest);
17console.log('Multi-tool response:', JSON.stringify(multiToolResponse, null, 2));
5

Maintain Conversation Context

Build multi-turn conversations by maintaining the message history:

1// Start with system message and initial user question
2const messages: AthenaIntelligence.InputMessage[] = [
3  {
4    content: 'You are a knowledgeable assistant specializing in technology and business analysis.',
5    role: 'system',
6    type: 'system',
7  },
8  {
9    content: 'What are the key trends in artificial intelligence for 2024?',
10    role: 'user',
11    type: 'user',
12  },
13];
14
15const config: AthenaIntelligence.GeneralAgentConfig = {
16  enabled_tools: ['search'],
17  model: 'gpt-4-turbo-preview',
18};
19
20// First request
21let conversationRequest: AthenaIntelligence.GeneralAgentRequest = {
22  config,
23  messages,
24};
25
26const firstResponse = await client.agents.general.invoke(conversationRequest);
27console.log('First response received');
28
29// Add the assistant's response to conversation history
30if (firstResponse.messages && firstResponse.messages.length > 0) {
31  const lastMessage = firstResponse.messages[firstResponse.messages.length - 1];
32  messages.push({
33    content: lastMessage.kwargs?.content || 'No content',
34    role: 'assistant',
35    type: 'assistant',
36  });
37}
38
39// Add follow-up question
40messages.push({
41  content: 'Can you provide more specific examples of companies implementing these trends?',
42  role: 'user',
43  type: 'user',
44});
45
46// Continue conversation with full context
47conversationRequest = {
48  config,
49  messages,
50};
51
52const secondResponse = await client.agents.general.invoke(conversationRequest);
53console.log('Conversation completed:', JSON.stringify(secondResponse, null, 2));
6

Error Handling

Always include proper error handling for production applications:

1import { AthenaIntelligenceError } from 'athena-intelligence';
2
3try {
4  const response = await client.agents.general.invoke(request);
5  console.log('Response received:', response);
6} catch (error) {
7  if (error instanceof AthenaIntelligenceError) {
8    console.error(`Athena API error (${error.statusCode}): ${error.message}`);
9  } else {
10    console.error('Unexpected error:', error);
11  }
12}
7

Complete Example

Here’s a complete working example that demonstrates all the concepts:

1import type { AthenaIntelligence } from 'athena-intelligence';
2import { AthenaIntelligenceClient, AthenaIntelligenceError } from 'athena-intelligence';
3
4async function runAgentExample() {
5  try {
6    const client = new AthenaIntelligenceClient({
7      apiKey: process.env.ATHENA_API_KEY,
8    });
9
10    // Multi-tool research request
11    const request: AthenaIntelligence.GeneralAgentRequest = {
12      config: {
13        enabled_tools: ['search', 'browse'],
14        system_prompt: 'You are a helpful research assistant. Use search and browse tools to find comprehensive information.',
15        model: 'gpt-4-turbo-preview',
16      },
17      messages: [
18        {
19          content: 'Research the latest developments in AI and provide a detailed summary with sources.',
20          role: 'user',
21          type: 'user',
22        },
23      ],
24    };
25
26    console.log('Sending request to General Agent...');
27    const response = await client.agents.general.invoke(request);
28    
29    console.log('Response received:');
30    console.log(JSON.stringify(response, null, 2));
31  } catch (error) {
32    if (error instanceof AthenaIntelligenceError) {
33      console.error(`Athena API error (${error.statusCode}): ${error.message}`);
34    } else {
35      console.error('Unexpected error:', error);
36    }
37  }
38}
39
40runAgentExample();