​

Data Analysis

Use LLMs to generate analysis and plotting code, execute it across AVM nodes, and retrieve rich outputs such as charts or summary statistics.

​

Scenario: Trend Discovery

Analyze token transaction CSVs to extract metrics and visualize patterns without manual scripting.

​

Solution: LLM + AVM

1. Prepare Sample
   Load a CSV subset locally.
2. Model Prompt
   Ask the LLM to write a Python function for metrics and plotting.
3. Sandbox Execution
   Run the code with AVM’s runPython, leveraging pandas and matplotlib.
4. Collect Results
   Extract base64-encoded images or JSON stats for downstream use.

​

Example (TypeScript)

import { runPythonTool } from "@avm-ai/avm-vercel-ai";
import { generateText } from "ai";
import { openai } from "@ai-sdk/openai";
import fs from "fs";

const tool = runPythonTool();
const csvSample = fs.readFileSync("data.csv", "utf-8");

async function analyze() {
  const prompt = \`
Write execute(input) that computes min, max, mean, median, 
and returns a line chart as base64 PNG.\`;
  const { text: code } = await generateText({
    model: openai("gpt-4o"),
    prompt,
    tools: { runPython: tool },
  });
  const result = await tool.exec({ code, input: { data: csvSample } });
  // Use result.output.image
}

​

Next Steps

• Customize visuals (bar, scatter).
• Integrate seaborn or plotly via custom runtimes.
• Embed in agent workflows via MCP.