William Valentin
125 lines
3.9 KiB
TypeScript
125 lines
3.9 KiB
TypeScript
import { execFile } from 'child_process';
|
|
import { randomUUID } from 'crypto';
|
|
import { readFile, unlink } from 'fs/promises';
|
|
import { join } from 'path';
|
|
import { tmpdir } from 'os';
|
|
|
|
import type { Tool, ToolResult } from '../types.js';
|
|
|
|
interface CaptureArgs {
|
|
format?: 'png' | 'jpg';
|
|
}
|
|
|
|
interface CameraCaptureArgs extends CaptureArgs {
|
|
device?: string;
|
|
}
|
|
|
|
const DEFAULT_TIMEOUT_MS = 15000;
|
|
|
|
function tempPath(format: string): string {
|
|
return join(tmpdir(), `flynn-capture-${randomUUID()}.${format}`);
|
|
}
|
|
|
|
function exec(command: string, args: string[], timeoutMs = DEFAULT_TIMEOUT_MS): Promise<void> {
|
|
return new Promise((resolve, reject) => {
|
|
execFile(command, args, { timeout: timeoutMs }, (error, _stdout, stderr) => {
|
|
if (error) {
|
|
reject(new Error(`${command} ${args.join(' ')} failed: ${stderr || error.message}`));
|
|
return;
|
|
}
|
|
resolve();
|
|
});
|
|
});
|
|
}
|
|
|
|
async function readBase64AndCleanup(path: string): Promise<string> {
|
|
try {
|
|
const data = await readFile(path);
|
|
return data.toString('base64');
|
|
} finally {
|
|
await unlink(path).catch(() => {});
|
|
}
|
|
}
|
|
|
|
export const screenCaptureTool: Tool = {
|
|
name: 'screen.capture',
|
|
description: 'Capture the current screen and return a base64-encoded image (png/jpg).',
|
|
inputSchema: {
|
|
type: 'object',
|
|
properties: {
|
|
format: { type: 'string', description: 'Image format: png (default) or jpg' },
|
|
},
|
|
},
|
|
execute: async (rawArgs: unknown): Promise<ToolResult> => {
|
|
try {
|
|
const args = (rawArgs as CaptureArgs | undefined) ?? {};
|
|
const format = args.format === 'jpg' ? 'jpg' : 'png';
|
|
const out = tempPath(format);
|
|
|
|
if (process.platform === 'darwin') {
|
|
await exec('screencapture', ['-x', '-t', format === 'jpg' ? 'jpg' : 'png', out]);
|
|
} else if (process.platform === 'linux') {
|
|
try {
|
|
await exec('grim', [out]);
|
|
} catch {
|
|
await exec('import', ['-window', 'root', out]);
|
|
}
|
|
} else {
|
|
return { success: false, output: '', error: 'screen.capture is only supported on macOS/Linux hosts' };
|
|
}
|
|
|
|
const data = await readBase64AndCleanup(out);
|
|
return {
|
|
success: true,
|
|
output: JSON.stringify({ mimeType: format === 'jpg' ? 'image/jpeg' : 'image/png', data }),
|
|
};
|
|
} catch (error) {
|
|
return {
|
|
success: false,
|
|
output: '',
|
|
error: error instanceof Error ? error.message : String(error),
|
|
};
|
|
}
|
|
},
|
|
};
|
|
|
|
export const cameraCaptureTool: Tool = {
|
|
name: 'camera.capture',
|
|
description: 'Capture a single frame from the default camera and return a base64-encoded image.',
|
|
inputSchema: {
|
|
type: 'object',
|
|
properties: {
|
|
format: { type: 'string', description: 'Image format: png (default) or jpg' },
|
|
device: { type: 'string', description: 'Optional camera device identifier (platform-specific)' },
|
|
},
|
|
},
|
|
execute: async (rawArgs: unknown): Promise<ToolResult> => {
|
|
try {
|
|
const args = (rawArgs as CameraCaptureArgs | undefined) ?? {};
|
|
const format = args.format === 'jpg' ? 'jpg' : 'png';
|
|
const out = tempPath(format);
|
|
|
|
if (process.platform === 'darwin') {
|
|
await exec('imagesnap', ['-q', '-w', '1', out]);
|
|
} else if (process.platform === 'linux') {
|
|
const device = args.device ?? '/dev/video0';
|
|
await exec('ffmpeg', ['-y', '-f', 'video4linux2', '-i', device, '-frames:v', '1', out]);
|
|
} else {
|
|
return { success: false, output: '', error: 'camera.capture is only supported on macOS/Linux hosts' };
|
|
}
|
|
|
|
const data = await readBase64AndCleanup(out);
|
|
return {
|
|
success: true,
|
|
output: JSON.stringify({ mimeType: format === 'jpg' ? 'image/jpeg' : 'image/png', data }),
|
|
};
|
|
} catch (error) {
|
|
return {
|
|
success: false,
|
|
output: '',
|
|
error: error instanceof Error ? error.message : String(error),
|
|
};
|
|
}
|
|
},
|
|
};
|