@falai/agent 0.8.0 → 0.9.0-alpha-1

package/examples/ai-providers/anthropic-integration.ts

@@ -0,0 +1,377 @@
+/**
+ * Anthropic Provider Integration Example
+ *
+ * This example demonstrates how to use the Anthropic provider with Claude models
+ * for building conversational agents. Shows provider configuration, model selection,
+ * and Claude-specific features.
+ *
+ * Key concepts:
+ * - Anthropic provider setup
+ * - Claude model configuration
+ * - Temperature and other parameters
+ * - Streaming responses
+ * - Tool calling with Claude
+ */
+
+import { Agent, AnthropicProvider, type Tool } from "../../src";
+
+// Define data types for our example
+interface ResearchData {
+  topic: string;
+  depth: "overview" | "detailed" | "comprehensive";
+  sources: number;
+  format: "summary" | "bullet_points" | "structured";
+  researchId?: string;
+}
+
+// Research tool that Claude can use
+const conductResearch: Tool<unknown, unknown[], unknown, ResearchData> = {
+  id: "conduct_research",
+  description: "Conduct comprehensive research on a given topic",
+  parameters: {
+    type: "object",
+    properties: {},
+  },
+  handler: async ({ data }) => {
+    const researchData = data as Partial<ResearchData>;
+    console.log(
+      `Conducting ${researchData?.depth} research on: ${researchData?.topic}`
+    );
+
+    const researchId = `RES-${Date.now()}`;
+
+    // Simulate research process
+    const findings = {
+      overview: `Found ${researchData?.sources || 3} sources about ${
+        researchData?.topic
+      }`,
+      key_points: [
+        "Latest developments in the field",
+        "Current trends and challenges",
+        "Future outlook and predictions",
+      ],
+      sources: ["academic papers", "industry reports", "expert interviews"],
+    };
+
+    return Promise.resolve({
+      data: `Research completed (ID: ${researchId}). ${findings.overview}`,
+      dataUpdate: {
+        researchId,
+      },
+    });
+  },
+};
+
+// Create agent with Anthropic provider
+const agent = new Agent({
+  name: "ClaudeResearcher",
+  description: "A research assistant powered by Claude",
+  provider: new AnthropicProvider({
+    apiKey: process.env.ANTHROPIC_API_KEY!,
+    model: "claude-sonnet-4-5", // Latest Claude model
+    config: {
+      temperature: 0.7, // Balanced creativity vs consistency
+      max_tokens: 4096,
+      top_p: 0.9,
+    },
+  }),
+});
+
+// Create research route
+const researchRoute = agent.createRoute<ResearchData>({
+  title: "Research Assistant",
+  description: "Conduct research using Claude's analytical capabilities",
+  schema: {
+    type: "object",
+    properties: {
+      topic: { type: "string", description: "The research topic" },
+      depth: {
+        type: "string",
+        enum: ["overview", "detailed", "comprehensive"],
+        default: "detailed",
+      },
+      sources: {
+        type: "number",
+        minimum: 1,
+        maximum: 20,
+        default: 5,
+        description: "Number of sources to analyze",
+      },
+      format: {
+        type: "string",
+        enum: ["summary", "bullet_points", "structured"],
+        default: "structured",
+        description: "Output format preference",
+      },
+      researchId: { type: "string" },
+    },
+    required: ["topic"],
+  },
+  identity: `You are Claude, an AI research assistant created by Anthropic.
+               You have access to extensive knowledge and can conduct thorough research.
+               Always be helpful, truthful, and provide well-reasoned analysis.`,
+  initialStep: {
+    prompt: `Hello! I'm Claude, your AI research assistant. I can help you research any topic thoroughly.
+
+What would you like me to research today? For best results, include:
+- The specific topic or question
+- How deep you want the research (overview, detailed, or comprehensive)
+- What format you'd prefer (summary, bullet points, or structured)
+
+For example: "Research the impact of artificial intelligence on healthcare, comprehensive analysis, structured format"`,
+    collect: ["topic"],
+  },
+});
+
+// Build research conversation flow
+const askTopic = researchRoute.initialStep.nextStep({
+  prompt: "What topic would you like me to research?",
+  collect: ["topic"],
+  skipIf: (data) => !!data.topic,
+});
+
+const askDepth = askTopic.nextStep({
+  prompt:
+    "How deep should I go with this research? (overview, detailed, or comprehensive)",
+  collect: ["depth"],
+  requires: ["topic"],
+  skipIf: (data) => !!data.depth,
+});
+
+const askSources = askDepth.nextStep({
+  prompt: "How many sources should I analyze? (1-20, default is 5)",
+  collect: ["sources"],
+  requires: ["topic"],
+  skipIf: (data) => data.sources !== undefined,
+});
+
+const askFormat = askSources.nextStep({
+  prompt:
+    "What format would you prefer for the results? (summary, bullet_points, or structured)",
+  collect: ["format"],
+  requires: ["topic"],
+  skipIf: (data) => !!data.format,
+});
+
+askFormat.nextStep({
+  prompt: ({ session }) => {
+    const data = session?.data;
+    return `I'll now conduct ${data?.depth || "detailed"} research on "${
+      data?.topic
+    }" using ${data?.sources || 5} sources. This may take a moment...`;
+  },
+  tools: [conductResearch],
+  requires: ["topic"],
+});
+
+// Demonstrate Claude's research capabilities
+async function demonstrateClaudeResearch() {
+  console.log("=== Claude Research Assistant Demo ===\n");
+
+  // Example 1: Comprehensive research request
+  console.log("Example 1: Comprehensive AI research");
+  console.log(
+    "User: Research the impact of artificial intelligence on healthcare, comprehensive analysis, structured format"
+  );
+
+  const response1 = await agent.respond({
+    history: [
+      {
+        role: "user",
+        content:
+          "Research the impact of artificial intelligence on healthcare, comprehensive analysis, structured format",
+        name: "Researcher",
+      },
+    ],
+  });
+
+  console.log("Claude:", response1.message);
+  console.log(
+    "Collected data:",
+    JSON.stringify(response1.session?.data, null, 2)
+  );
+  console.log();
+
+  // Example 2: Step-by-step research configuration
+  console.log("Example 2: Step-by-step configuration");
+  console.log("User: I want to research renewable energy");
+
+  const response2a = await agent.respond({
+    history: [
+      {
+        role: "user",
+        content: "I want to research renewable energy",
+        name: "Student",
+      },
+    ],
+  });
+
+  console.log("Claude:", response2a.message);
+
+  console.log("User: Make it detailed with 8 sources");
+  const response2b = await agent.respond({
+    history: [
+      {
+        role: "user",
+        content: "I want to research renewable energy",
+        name: "Student",
+      },
+      {
+        role: "assistant",
+        content: response2a.message,
+      },
+      {
+        role: "user",
+        content: "Make it detailed with 8 sources",
+        name: "Student",
+      },
+    ],
+    session: response2a.session,
+  });
+
+  console.log("Claude:", response2b.message);
+
+  console.log("User: Use bullet points format");
+  const response2c = await agent.respond({
+    history: [
+      {
+        role: "user",
+        content: "I want to research renewable energy",
+        name: "Student",
+      },
+      {
+        role: "assistant",
+        content: response2a.message,
+      },
+      {
+        role: "user",
+        content: "Make it detailed with 8 sources",
+        name: "Student",
+      },
+      {
+        role: "assistant",
+        content: response2b.message,
+      },
+      {
+        role: "user",
+        content: "Use bullet points format",
+        name: "Student",
+      },
+    ],
+    session: response2b.session,
+  });
+
+  console.log("Claude:", response2c.message);
+  console.log(
+    "Final collected data:",
+    JSON.stringify(response2c.session?.data, null, 2)
+  );
+  console.log("Research complete:", response2c.isRouteComplete);
+}
+
+// Demonstrate streaming with Claude
+async function demonstrateClaudeStreaming() {
+  console.log("\n=== Claude Streaming Demo ===\n");
+
+  console.log("User: Give me a comprehensive overview of quantum computing");
+
+  let accumulatedResponse = "";
+  for await (const chunk of agent.respondStream({
+    history: [
+      {
+        role: "user",
+        content: "Give me a comprehensive overview of quantum computing",
+        name: "Scientist",
+      },
+    ],
+  })) {
+    // Show streaming in real-time
+    process.stdout.write(chunk.delta);
+    accumulatedResponse += chunk.delta;
+
+    if (chunk.done) {
+      console.log("\n\nStreaming complete!");
+      console.log("Total length:", accumulatedResponse.length, "characters");
+      console.log(
+        "Session data:",
+        JSON.stringify(chunk.session?.data, null, 2)
+      );
+      break;
+    }
+  }
+}
+
+// Show different Claude model configurations
+function demonstrateModelConfigurations() {
+  console.log("\n=== Claude Model Configurations ===\n");
+
+  const configurations = [
+    {
+      name: "Creative Writing Assistant",
+      model: "claude-sonnet-4.5",
+      temperature: 0.9,
+      maxTokens: 8192,
+      useCase: "High creativity, long-form content",
+    },
+    {
+      name: "Code Review Bot",
+      model: "claude-haiku-4.5",
+      temperature: 0.3,
+      maxTokens: 4096,
+      useCase: "Fast, consistent code analysis",
+    },
+    {
+      name: "Data Analyst",
+      model: "claude-sonnet-4.5",
+      temperature: 0.1,
+      maxTokens: 16384,
+      useCase: "Precise analysis, complex reasoning",
+    },
+  ];
+
+  configurations.forEach((config, index) => {
+    console.log(`${index + 1}. ${config.name}`);
+    console.log(`   Model: ${config.model}`);
+    console.log(`   Temperature: ${config.temperature}`);
+    console.log(`   Max Tokens: ${config.maxTokens}`);
+    console.log(`   Use Case: ${config.useCase}`);
+    console.log();
+  });
+
+  console.log("Example configuration code:");
+  console.log(
+    `
+const creativeWriter = new AnthropicProvider({
+  apiKey: process.env.ANTHROPIC_API_KEY!,
+  model: "claude-sonnet-4.5",
+  temperature: 0.9,  // High creativity
+  maxTokens: 8192,   // Long responses
+});
+
+const codeReviewer = new AnthropicProvider({
+  apiKey: process.env.ANTHROPIC_API_KEY!,
+  model: "claude-haiku-4.5",
+  temperature: 0.3,  // Low creativity, high consistency
+  maxTokens: 4096,   // Shorter responses
+});
+  `.trim()
+  );
+}
+
+// Run demonstrations
+async function main() {
+  try {
+    demonstrateModelConfigurations();
+    await demonstrateClaudeResearch();
+    await demonstrateClaudeStreaming();
+  } catch (error) {
+    console.error("Error:", error);
+    console.log(
+      "\nNote: Make sure to set ANTHROPIC_API_KEY environment variable"
+    );
+  }
+}
+
+if (import.meta.url === `file://${process.argv[1]}`) {
+  main().catch(console.error);
+}

package/examples/{openai-agent.ts → ai-providers/openai-integration.ts}

@@ -6,12 +6,10 @@
 import {
   Agent,
   OpenAIProvider,
-  defineTool,
-  createMessageEvent,
-  EventSource,
-  createSession,
+  type Tool,
+  userMessage,
   END_ROUTE,
-} from "../src/index";
+} from "../../src";
 
 // Custom context type
 interface CustomerContext {
@@ -28,32 +26,31 @@ interface WeatherData {
 }
 
 // Define a tool that can access collected data
-const getWeather = defineTool<
-  CustomerContext,
-  [{ location: string }],
-  { location: string; temperature: number; condition: string }
->(
-  "get_weather",
-  async ({ context, data }, args) => {
+const getWeather: Tool<CustomerContext, unknown[], unknown, WeatherData> = {
+  id: "get_weather",
+  description: "Get current weather for a location",
+  parameters: {
+    type: "object",
+    properties: {
+      location: { type: "string", description: "City name" },
+    },
+    required: ["location"],
+  },
+  handler: ({ data }, location) => {
     // Use data location if available, otherwise use args
-    const location = (data as Partial<WeatherData>)?.location || args.location;
+    const weatherData = data as Partial<WeatherData>;
+    const finalLocation = weatherData?.location || location;
 
     // Simulate API call
     return {
       data: {
-        location,
+        location: finalLocation,
         temperature: 72,
         condition: "Sunny",
       },
     };
   },
-  {
-    description: "Get current weather for a location",
-    parameters: {
-      location: { type: "string", description: "City name" },
-    },
-  }
-);
+};
 
 async function main() {
   // Initialize OpenAI provider
@@ -79,6 +76,8 @@ async function main() {
     description:
       "A helpful AI assistant that can check weather and answer questions",
     goal: "Assist users with information and weather queries",
+    identity:
+      "I am the Assistant, an intelligent AI powered by OpenAI's advanced language models. With access to real-time information and deep reasoning capabilities, I'm here to help you with weather updates, questions, and any information you need.",
     context: {
       customerId: "user123",
       name: "Alice",
@@ -133,20 +132,20 @@ async function main() {
 
   // Step 1: Collect location
   const askLocation = weatherRoute.initialStep.nextStep({
-    instructions: "Ask which city they want weather for",
+    prompt: "Ask which city they want weather for",
     collect: ["location"],
     skipIf: (data) => !!data.location,
   });
 
   // Step 2: Get weather data
   const fetchWeather = askLocation.nextStep({
-    tool: getWeather,
+    tools: [getWeather],
     requires: ["location"],
   });
 
   // Step 3: Present weather information
   const showWeather = fetchWeather.nextStep({
-    instructions:
+    prompt:
       "Present the weather information in a friendly way with temperature and condition",
   });
 
@@ -155,22 +154,18 @@ async function main() {
   // Example conversation with session step management
   console.log("🤖 Starting OpenAI Agent Example\n");
 
-  // Initialize session step for multi-turn conversation
-  let session = createSession<WeatherData>();
-
-  // Build history
-  const history = [
-    createMessageEvent(
-      EventSource.CUSTOMER,
-      "Alice",
-      "What's the weather like in San Francisco?"
-    ),
-  ];
+  // Session is automatically managed by the agent
+  console.log("✨ Session ready:", agent.session.id);
 
   try {
-    // Turn 1: Process weather query with session step
-    console.log("📤 Processing with session step...");
-    const response = await agent.respond({ history, session });
+    // Turn 1: Process weather query with session management
+    console.log("📤 Processing with session management...");
+    
+    await agent.session.addMessage("user", "What's the weather like in San Francisco?", "Alice");
+    
+    const response = await agent.respond({ 
+      history: agent.session.getHistory() 
+    });
 
     console.log("\n✅ Agent Configuration:");
     console.log(`   AI Provider: ${openaiProvider.name}`);
@@ -182,18 +177,17 @@ async function main() {
     );
 
     console.log("\n💬 Conversation:");
-    console.log(`   Customer: ${history[0].data.message}`);
+    console.log(`   Customer: What's the weather like in San Francisco?`);
     console.log(`   Agent: ${response.message}`);
     console.log(`   Route: ${response.session?.currentRoute?.title}`);
-    console.log(`   Data:`, response.session?.data);
+    console.log(`   Data:`, agent.session.getData<WeatherData>());
 
-    // Update session with progress
-    session = response.session!;
+    await agent.session.addMessage("assistant", response.message);
 
     // Check for route completion
     if (response.isRouteComplete) {
       console.log("\n✅ Weather route complete!");
-      await logWeatherRequest(agent.getData(session.id) as WeatherData);
+      await logWeatherRequest(agent.session.getData<WeatherData>());
     }
 
     console.log("\n✨ Session step benefits:");