feat: Add Google Gemini Data Analysis Examples (Python & TypeScript)

README.md

@@ -112,6 +112,13 @@ Read more about E2B on the [E2B website](https://e2b.dev) and the official [E2B
       <td><a href="https://github.com/e2b-dev/e2b-cookbook/tree/main/examples/watsonx-ai-code-interpreter-python">Python</a></td>
       <td><a href="https://github.com/e2b-dev/e2b-cookbook/tree/main/examples/watsonx-ai-code-interpreter-js">TypeScript</a></td>
     </tr>
+    <tr>
+      <td>Google</td>
+      <td>Gemini 3.0 Flash</td>
+      <td>Data analysis and visualization</td>
+      <td><a href="https://github.com/e2b-dev/e2b-cookbook/tree/main/examples/gemini-data-analysis-python">Python</a></td>
+      <td><a href="https://github.com/e2b-dev/e2b-cookbook/tree/main/examples/gemini-data-analysis-js">TypeScript</a></td>
+    </tr>
   </tbody>
 </table>
 

examples/gemini-data-analysis-js/.env.template

@@ -0,0 +1,5 @@
+# Get your E2B API key here: https://e2b.dev/docs/getting-started/api-key
+E2B_API_KEY=
+
+# Get your Google Gemini API key here: https://ai.google.dev/
+GEMINI_API_KEY=

examples/gemini-data-analysis-js/README.md

@@ -0,0 +1,40 @@
+# Data Analysis and Visualization with Google Gemini
+
+This is an example that tests the capabilities of Google's Gemini models. You can choose from models like gemini-3-flash-preview. We let the LLM write the code to analyze a dataset from Kaggle and generate visualizations. We use the E2B Code Interpreter SDK for running the LLM-generated code tasks in a secure and isolated cloud environment.
+
+## Tech Stack
+- [E2B Code Interpreter SDK]((https://github.com/e2b-dev/code-interpreter)) for running the LLM-generated code
+- [Gemini's 3.0 Flash](https://ai.google.dev/gemini-api/docs/models) as an LLM
+- JavaScript/TypeScript
+
+
+## Setup
+### 1. Set up API keys
+1. Copy the `.env.template` file to `.env`:
+   ```bash
+   cp .env.template .env
+   ```
+
+2. Get your [E2B API key](https://e2b.dev/docs/getting-started/api-key) and [Google AI Studio API key](https://aistudio.google.com/).
+
+3. Add your API keys to the `.env` file.
+
+4. Install the dependencies:
+   ```bash
+   npm install
+   ```
+
+5. Run the example:
+   ```bash
+   npm start
+   ```
+
+After running the program, you should get the results of a chart visualizing the yearly temperature trends in the top 5 hottest cities.
+
+![Example of the output](temperature_analysis.png)
+
+If you encounter any problems, please let us know at our [Discord]((https://discord.com/invite/U7KEcGErtQ)).
+If you want to let the world know about what you're building with E2B, tag [@e2b_dev](https://twitter.com/e2b_dev) on X (Twitter).
+
+### 4. Visit our docs
+Check the documentation to learn more about how to use E2B. [Visit our docs](https://e2b.dev/docs).

examples/gemini-data-analysis-js/app.ts

@@ -0,0 +1,164 @@
+import fs from 'node:fs'
+import { GoogleGenAI, Type } from '@google/genai'
+import { Sandbox, Result } from '@e2b/code-interpreter'
+import { OutputMessage } from '@e2b/code-interpreter'
+import * as dotenv from 'dotenv'
+
+dotenv.config()
+
+// 1. Initialize Gemini Client
+const client = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });
+const MODEL_NAME = 'gemini-3-flash-preview'
+
+const SYSTEM_PROMPT = `
+You are an expert Data Analyst and Python programmer.
+You have access to a remote Python environment (E2B sandbox).
+Your task is to analyze data, write Python code to perform the analysis and generate visualizations.
+
+When asked to analyze data:
+1.  Verify what files are available.
+2.  Write Python code to load the data (usually pandas).
+3.  Perform the request analysis.
+4.  Generate charts using matplotlib or seaborn.
+5.  ALWAYS save the chart as a file (e.g., 'output.png', 'chart.png').
+6.  Print relevant insights to stdout.
+
+IMPORTANT CODE OUTPUT FORMAT:
+You must strictly output your Python code in the following format:
+\`\`\`python
+# ... your code here ...
+\`\`\`
+Do not include any other markdown code blocks or shell commands unless explicitly asked.
+`
+
+// 2. Define the tool
+const executePythonTool = {
+    name: 'execute_python',
+    description: 'Execute python code in a Jupyter notebook cell and returns any result, stdout, stderr, display_data, and error.',
+    parameters: {
+        type: Type.OBJECT,
+        properties: {
+            code: {
+                type: Type.STRING,
+                description: 'The python code to execute in a single cell.',
+            }
+        },
+        required: ['code']
+    }
+}
+
+// 3. E2B Code Interpreter Helper
+async function codeInterpret(codeInterpreter: Sandbox, code: string) {
+    console.log('Running code interpreter...')
+
+    const exec = await codeInterpreter.runCode(code, {
+        onStderr: (msg: OutputMessage) => console.log(`[Stderr] ${msg.line}`),
+        onStdout: (stdout: OutputMessage) => console.log(`[Stdout] ${stdout.line}`),
+    })
+
+    if (exec.error) {
+        console.log('[Code Interpreter ERROR]', exec.error)
+        throw new Error(exec.error.value)
+    }
+    return exec
+}
+
+async function uploadDataset(codeInterpreter: Sandbox): Promise<string> {
+    console.log('Uploading dataset to Code Interpreter sandbox...')
+    const datasetPath = './city_temperature.csv'
+
+    if (!fs.existsSync(datasetPath)) {
+        throw new Error('Dataset file not found')
+    }
+
+    const fileBuffer = fs.readFileSync(datasetPath)
+
+    try {
+        const remotePath = await codeInterpreter.files.write('city_temperature.csv', fileBuffer)
+        if (!remotePath) {
+            throw new Error('Failed to upload dataset')
+        }
+        console.log('Uploaded at', remotePath)
+        return remotePath.path
+    } catch (error) {
+        console.error('Error during file upload:', error)
+        throw error
+    }
+}
+
+// 4. Main Chat Logic
+async function run() {
+    console.log('Starting E2B Code Interpreter...');
+    const codeInterpreter = await Sandbox.create({ apiKey: process.env.E2B_API_KEY });
+
+    try {
+        await uploadDataset(codeInterpreter);
+
+        // Initialize Chat Session
+        const chat = client.chats.create({
+            model: MODEL_NAME,
+            config: {
+                systemInstruction: SYSTEM_PROMPT,
+                tools: [{ functionDeclarations: [executePythonTool] }],
+            },
+        });
+
+        // First Message
+        let response = await chat.sendMessage({ message: "Analyze the temperature data for the top 5 hottest cities globally. Create a visualization showing their average temperatures over the years." });
+
+        while (true) {
+            const parts = response.candidates[0].content.parts;
+            const functionCalls = parts.filter(p => p.functionCall);
+
+            if (functionCalls.length > 0) {
+                const toolResponseParts = [];
+
+                for (const call of functionCalls) {
+                    const fc = call.functionCall;
+                    console.log(`\nTool Call: ${fc.name}`);
+
+                    const exec = await codeInterpret(codeInterpreter, fc.args.code);
+
+                    let outputStr = "";
+
+                    // Handle Results (Images, etc)
+                    for (const res of exec.results) {
+                        if (res.png) {
+                            // Decode base64 PNG from E2B and save
+                            fs.writeFileSync('temperature_analysis.png', Buffer.from(res.png, 'base64'));
+                            console.log("Chart saved to temperature_analysis.png");
+                        }
+                    }
+
+                    // Collect logs from the Execution object
+                    if (exec.logs.stdout.length) outputStr += exec.logs.stdout.join("\n") + "\n";
+                    if (exec.logs.stderr.length) outputStr += exec.logs.stderr.join("\n") + "\n";
+
+                    if (!outputStr) outputStr = "Code executed successfully.";
+
+                    toolResponseParts.push({
+                        functionResponse: {
+                            name: fc.name,
+                            response: { result: outputStr }
+                        }
+                    });
+                }
+
+                console.log("Sending tool outputs back to Gemini...");
+                response = await chat.sendMessage({ message: toolResponseParts });
+
+            } else {
+                console.log(`\nFinal Response: ${response.text}`);
+                break;
+            }
+        }
+
+    } catch (e) {
+        console.error("Error in execution loop:", e);
+    } finally {
+        await codeInterpreter.kill();
+        console.log("Sandbox closed.");
+    }
+}
+
+run();