/* eslint-disable no-console */
// Initialize Phoenix tracing before any AI SDK calls
import {
getInputAttributes,
getOutputAttributes,
withSpan,
} from "@arizeai/openinference-core";
import { flush, SESSION_ID } from "./instrumentation.js";
import { anthropic } from "@ai-sdk/anthropic";
import { stepCountIs, tool, ToolLoopAgent } from "ai";
import * as readline from "node:readline";
import { z } from "zod";
// Define tools for the agent
const calculatorTool = tool({
description: "Perform mathematical calculations",
inputSchema: z.object({
expression: z.string().describe("The mathematical expression to evaluate"),
}),
execute: async ({ expression }: { expression: string }) => {
try {
// Simple eval for demo purposes - in production, use a safe math parser
const result = eval(expression);
return { result, expression };
} catch (error) {
return { error: String(error), expression };
}
},
});
const getDateTimeTool = tool({
description: "Get the current date and time",
inputSchema: z.object({}),
execute: async () => {
const now = new Date();
return {
date: now.toLocaleDateString(),
time: now.toLocaleTimeString(),
iso: now.toISOString(),
};
},
});
function printWelcome() {
console.log("╔═══════════════════════════════════════════════════════════╗");
console.log("║ CLI Agent Starter Kit - Interactive Mode ║");
console.log("╚═══════════════════════════════════════════════════════════╝");
console.log("\nAvailable tools:");
console.log(" • Calculator - Perform mathematical calculations");
console.log(" • Date/Time - Get current date and time");
console.log("\nCommands:");
console.log(" /exit or /quit - Exit the agent");
console.log(" /help - Show this help message");
console.log(" /clear - Clear conversation history");
console.log("\n");
}
async function main() {
const apiKey = process.env.ANTHROPIC_API_KEY;
if (!apiKey) {
console.error("Error: ANTHROPIC_API_KEY environment variable is not set");
await flush();
process.exit(1);
}
printWelcome();
// Display session ID for tracking
console.log(`\x1b[90mSession ID: ${SESSION_ID}\x1b[0m\n`);
// Create a ToolLoopAgent with tools
const agent = new ToolLoopAgent({
model: anthropic("claude-sonnet-4-20250514"),
instructions:
"You are a helpful CLI agent. Use the available tools to answer questions accurately. Be concise and friendly.",
tools: {
calculator: calculatorTool,
getDateTime: getDateTimeTool,
},
// Stop after 10 steps max
stopWhen: stepCountIs(10),
// Enable telemetry for Phoenix tracing
experimental_telemetry: { isEnabled: true },
});
// Create readline interface for user input
const rl = readline.createInterface({
input: process.stdin,
output: process.stdout,
prompt: "\n\x1b[36mYou:\x1b[0m ",
});
let conversationHistory: Array<{
role: "user" | "assistant";
content: string;
}> = [];
rl.prompt();
rl.on("line", async (input) => {
const userInput = input.trim();
// Handle commands
if (userInput === "/exit" || userInput === "/quit") {
console.log("\n\x1b[33mGoodbye!\x1b[0m\n");
rl.close();
return;
}
if (userInput === "/help") {
printWelcome();
rl.prompt();
return;
}
if (userInput === "/clear") {
conversationHistory = [];
console.log("\n\x1b[33m✓ Conversation history cleared\x1b[0m");
rl.prompt();
return;
}
if (!userInput) {
rl.prompt();
return;
}
// Add user message to history
conversationHistory.push({ role: "user", content: userInput });
try {
let stepNumber = 0;
const verbose = process.env.VERBOSE === "true";
console.log("\x1b[35mAgent:\x1b[0m");
// Use withSpan directly to create a parent CHAIN span with session.id
// This groups all child spans (LLM, TOOL, etc.) under one trace
const handleInteraction = withSpan(
async (_input: string) => {
return await agent.generate({
prompt: conversationHistory
.map(
(msg) =>
`${msg.role === "user" ? "User" : "Assistant"}: ${msg.content}`
)
.join("\n\n"),
// Log each step if verbose mode
onStepFinish: async ({
usage: _usage,
finishReason,
toolCalls,
}) => {
if (verbose) {
stepNumber++;
const tools = toolCalls
? toolCalls.map((tc) => tc.toolName).join(", ")
: "no tools";
console.log(
`\x1b[90m[Step ${stepNumber}] ${finishReason} - ${tools}\x1b[0m`
);
}
},
});
},
{
name: "cli.interaction",
kind: "CHAIN",
attributes: { "session.id": SESSION_ID },
// Capture input - automatically sets input.value and input.mime_type
processInput: (input: string) => getInputAttributes(input),
// Capture output - automatically sets output.value and output.mime_type
processOutput: (result) => getOutputAttributes(result.text),
}
);
const result = await handleInteraction(userInput);
// Add assistant response to history
conversationHistory.push({ role: "assistant", content: result.text });
console.log(result.text);
if (verbose) {
console.log(`\x1b[90m(${result.steps.length} steps)\x1b[0m`);
}
} catch (error) {
console.error("\n\x1b[31mError:\x1b[0m", error);
}
rl.prompt();
});
rl.on("close", async () => {
console.log("\nFlushing traces...");
await flush();
process.exit(0);
});
}
main()
.then(() => {
// Allow time for spans to be flushed before exit
// The beforeExit handler will ensure proper shutdown
})
.catch(async (error) => {
console.error("Fatal error:", error);
await flush();
process.exit(1);
});
