Cookbook: Financial Analyst

Build an agent that can fetch real-time stock data, analyze trends, and generate PDF reports.

Overview#

This recipe demonstrates how to chain multiple tools together: an API fetcher, a data analyzer, and a file generator.

Input

"Analyze AAPL and MSFT performance over the last month and summarize the key differences."

➔

Output

"Here is the comparison table and a downloadable PDF report..."

Ingredients#

Akios SDK for the agent loop.
Yahoo Finance API (or mock) for data.
PDFKit for report generation.
Zod for schema validation.

Create the Stock Tool

tools/stock.ts

import { Tool } from '@akios/sdk'
import { z } from 'zod'

export const stockTool = new Tool({
  name: 'get_stock_history',
  description: 'Fetch historical daily prices for a ticker symbol.',
  schema: z.object({
    symbol: z.string().describe('The stock ticker (e.g. AAPL)'),
    days: z.number().default(30)
  }),
  execute: async ({ symbol, days }) => {
    // In production, call a real API like AlphaVantage or Yahoo
    // Here we mock it for the demo
    const mockData = Array.from({ length: days }).map((_, i) => ({
      date: new Date(Date.now() - i * 86400000).toISOString().split('T')[0],
      price: 150 + Math.random() * 20
    }))
    return JSON.stringify({ symbol, history: mockData })
  }
})

Create the Analysis Tool

tools/analyze.ts

import { Tool } from '@akios/sdk'
import { z } from 'zod'

export const analysisTool = new Tool({
  name: 'calculate_volatility',
  description: 'Calculate the standard deviation of a price array.',
  schema: z.object({
    prices: z.array(z.number())
  }),
  execute: async ({ prices }) => {
    const mean = prices.reduce((a, b) => a + b) / prices.length
    const variance = prices.reduce((a, b) => a + Math.pow(b - mean, 2), 0) / prices.length
    return JSON.stringify({ volatility: Math.sqrt(variance) })
  }
})

Assemble the Agent

agent.ts

import { Agent } from '@akios/sdk'
import { stockTool, analysisTool } from './tools'

const analyst = new Agent({
  name: 'WallStreetBot',
  model: 'gpt-4o',
  systemPrompt: `You are a financial analyst.
  1. Fetch stock data for the requested companies.
  2. Calculate volatility for each.
  3. Provide a recommendation based on risk (high volatility = high risk).`,
  tools: [stockTool, analysisTool]
})

// Run it
const result = await analyst.run("Compare risk for AAPL vs TSLA")
console.log(result.output)

Going Further

Add a `generate_chart` tool using a library like `chart.js` or `matplotlib` (if using python backend) to return image URLs that the agent can display.