usecomputer 0.0.2 → 0.0.3

package/CHANGELOG.md

@@ -4,6 +4,19 @@
 
 All notable changes to `usecomputer` will be documented in this file.
 
+## 0.0.3
+
+- Implement real screenshot capture + PNG file writing on macOS.
+- Screenshot path handling now uses the requested output path reliably.
+- Unimplemented commands now return explicit `TODO not implemented: ...` errors.
+- Clarify `--display` index behavior as 0-based in help/docs.
+
+## 0.0.2
+
+- Publish macOS native binaries for both `darwin-arm64` and `darwin-x64`.
+- Add package metadata/docs for npm distribution.
+- Improve CLI coordinate input with `-x` / `-y` flags.
+
 ## 0.0.1
 
 - Initial npm package release for macOS.

package/README.md

@@ -35,3 +35,17 @@ Commands that target coordinates accept `-x` and `-y` flags:
 - `usecomputer mouse move -x <n> -y <n>`
 
 Legacy coordinate forms are also accepted where available.
+
+## Display index options
+
+For commands that accept `--display`, the index is 0-based:
+
+- `0` = first display
+- `1` = second display
+- `2` = third display
+
+Example:
+
+```bash
+usecomputer screenshot ./shot.png --display 0 --json
+```

package/build.zig

@@ -29,6 +29,7 @@ pub fn build(b: *std.Build) void {
     if (target.result.os.tag == .macos) {
         lib.root_module.linkFramework("CoreGraphics", .{});
         lib.root_module.linkFramework("CoreFoundation", .{});
+        lib.root_module.linkFramework("ImageIO", .{});
     }
 
     napigen.setup(lib);

package/dist/cli.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"cli.d.ts","sourceRoot":"","sources":["../src/cli.ts"],"names":[],"mappings":"AASA,OAAO,KAAK,EAAsB,iBAAiB,EAAE,MAAM,YAAY,CAAA;AAuDvE,wBAAgB,SAAS,CAAC,EAAE,MAAuB,EAAE,GAAE;IAAE,MAAM,CAAC,EAAE,iBAAiB,CAAA;CAAO,2BA4PzF;AAED,wBAAgB,MAAM,IAAI,IAAI,CAG7B"}
+{"version":3,"file":"cli.d.ts","sourceRoot":"","sources":["../src/cli.ts"],"names":[],"mappings":"AAWA,OAAO,KAAK,EAAsB,iBAAiB,EAAE,MAAM,YAAY,CAAA;AAuDvE,wBAAgB,SAAS,CAAC,EAAE,MAAuB,EAAE,GAAE;IAAE,MAAM,CAAC,EAAE,iBAAiB,CAAA;CAAO,2BAyQzF;AAED,wBAAgB,MAAM,IAAI,IAAI,CAG7B"}

package/dist/cli.js

@@ -3,6 +3,8 @@ import { goke } from 'goke';
 import { z } from 'zod';
 import dedent from 'string-dedent';
 import { createRequire } from 'node:module';
+import fs from 'node:fs';
+import pathModule from 'node:path';
 import url from 'node:url';
 import { createBridge } from './bridge.js';
 import { parseDirection, parseModifiers, parsePoint, parseRegion } from './command-parsers.js';
@@ -40,7 +42,7 @@ function parseButton(input) {
     return 'left';
 }
 function notImplemented({ command }) {
-    throw new Error(`Command \"${command}\" is not implemented yet`);
+    throw new Error(`TODO not implemented: ${command}`);
 }
 export function createCli({ bridge = createBridge() } = {}) {
     const cli = goke('usecomputer');
@@ -51,16 +53,25 @@ export function createCli({ bridge = createBridge() } = {}) {
         This command uses a native Zig backend over macOS APIs.
       `)
         .option('-r, --region [region]', z.string().describe('Capture region as x,y,width,height'))
-        .option('--display [display]', z.number().describe('Display index for multi-monitor setups'))
+        .option('--display [display]', z.number().describe('Display index for multi-monitor setups (0-based: first display is index 0)'))
         .option('--annotate', 'Annotate screenshot with labels')
         .option('--json', 'Output as JSON')
         .action(async (path, options) => {
+        const outputPath = path
+            ? path.startsWith('/')
+                ? path
+                : `${process.cwd()}/${path}`
+            : undefined;
+        if (path) {
+            const parentDirectory = pathModule.dirname(outputPath);
+            fs.mkdirSync(parentDirectory, { recursive: true });
+        }
         const region = options.region ? parseRegion(options.region) : undefined;
         if (region instanceof Error) {
             throw region;
         }
         const result = await bridge.screenshot({
-            path,
+            path: outputPath,
             region,
             display: options.display,
             annotate: options.annotate,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "usecomputer",
-  "version": "0.0.2",
+  "version": "0.0.3",
   "type": "module",
   "description": "Fast computer automation CLI for AI agents. Control any desktop with accessibility snapshots, clicks, typing, scrolling, and more.",
   "bin": "./bin.js",

package/src/cli.ts

@@ -4,6 +4,8 @@ import { goke } from 'goke'
 import { z } from 'zod'
 import dedent from 'string-dedent'
 import { createRequire } from 'node:module'
+import fs from 'node:fs'
+import pathModule from 'node:path'
 import url from 'node:url'
 import { createBridge } from './bridge.js'
 import { parseDirection, parseModifiers, parsePoint, parseRegion } from './command-parsers.js'
@@ -59,7 +61,7 @@ function parseButton(input?: string): MouseButton {
 }
 
 function notImplemented({ command }: { command: string }): never {
-  throw new Error(`Command \"${command}\" is not implemented yet`)
+  throw new Error(`TODO not implemented: ${command}`)
 }
 
 export function createCli({ bridge = createBridge() }: { bridge?: UseComputerBridge } = {}) {
@@ -75,16 +77,29 @@ export function createCli({ bridge = createBridge() }: { bridge?: UseComputerBri
       `,
     )
     .option('-r, --region [region]', z.string().describe('Capture region as x,y,width,height'))
-    .option('--display [display]', z.number().describe('Display index for multi-monitor setups'))
+    .option(
+      '--display [display]',
+      z.number().describe('Display index for multi-monitor setups (0-based: first display is index 0)'),
+    )
     .option('--annotate', 'Annotate screenshot with labels')
     .option('--json', 'Output as JSON')
     .action(async (path, options) => {
+      const outputPath = path
+        ? path.startsWith('/')
+          ? path
+          : `${process.cwd()}/${path}`
+        : undefined
+
+      if (path) {
+        const parentDirectory = pathModule.dirname(outputPath)
+        fs.mkdirSync(parentDirectory, { recursive: true })
+      }
       const region = options.region ? parseRegion(options.region) : undefined
       if (region instanceof Error) {
         throw region
       }
       const result = await bridge.screenshot({
-        path,
+        path: outputPath,
         region,
         display: options.display,
         annotate: options.annotate,

package/zig/src/lib.zig

@@ -8,6 +8,7 @@ const napigen = if (builtin.is_test) undefined else @import("napigen");
 const c = if (builtin.target.os.tag == .macos) @cImport({
     @cInclude("CoreGraphics/CoreGraphics.h");
     @cInclude("CoreFoundation/CoreFoundation.h");
+    @cInclude("ImageIO/ImageIO.h");
 }) else struct {};
 
 pub const std_options: std.Options = .{
@@ -54,7 +55,7 @@ fn execute(command: []const u8, payload_json: []const u8) ![]const u8 {
         return makeOkJson(allocator, "{\"text\":\"\"}");
     }
     if (std.mem.eql(u8, command, "screenshot")) {
-        return makeOkJson(allocator, "{\"path\":\"./screenshot.png\"}");
+        return executeScreenshotCommand(allocator, payload_json);
     }
 
     if (
@@ -63,7 +64,8 @@ fn execute(command: []const u8, payload_json: []const u8) ![]const u8 {
         std.mem.eql(u8, command, "scroll") or
         std.mem.eql(u8, command, "clipboard-set")
     ) {
-        return makeOkJson(allocator, "null");
+        const message = try std.fmt.allocPrint(allocator, "TODO not implemented: {s}", .{command});
+        return makeErrorJson(allocator, message);
     }
 
     return makeErrorJson(allocator, "unknown command");
@@ -140,6 +142,130 @@ const DragPayload = struct {
     button: ?[]const u8 = null,
 };
 
+const ScreenshotRegion = struct {
+    x: f64,
+    y: f64,
+    width: f64,
+    height: f64,
+};
+
+const ScreenshotPayload = struct {
+    path: ?[]const u8 = null,
+    display: ?usize = null,
+    region: ?ScreenshotRegion = null,
+};
+
+fn executeScreenshotCommand(allocator: std.mem.Allocator, payload_json: []const u8) ![]const u8 {
+    if (builtin.target.os.tag != .macos) {
+        return makeErrorJson(allocator, "screenshot is only supported on macOS");
+    }
+
+    var parsed = std.json.parseFromSlice(ScreenshotPayload, allocator, payload_json, .{
+        .ignore_unknown_fields = true,
+    }) catch {
+        return makeErrorJson(allocator, "invalid screenshot payload json");
+    };
+    defer parsed.deinit();
+
+    const screenshot_payload = parsed.value;
+    const output_path = screenshot_payload.path orelse "./screenshot.png";
+
+    const image = createScreenshotImage(.{
+        .display_index = screenshot_payload.display,
+        .region = screenshot_payload.region,
+    }) catch {
+        return makeErrorJson(allocator, "failed to capture screenshot image");
+    };
+    defer c.CFRelease(image);
+
+    writeScreenshotPng(.{
+        .image = image,
+        .output_path = output_path,
+    }) catch {
+        return makeErrorJson(allocator, "failed to write screenshot file");
+    };
+
+    const path_json = try std.fmt.allocPrint(allocator, "\"{s}\"", .{output_path});
+    const payload_json_response = try std.fmt.allocPrint(allocator, "{{\"path\":{s}}}", .{path_json});
+    return makeOkJson(allocator, payload_json_response);
+}
+
+fn createScreenshotImage(input: struct {
+    display_index: ?usize,
+    region: ?ScreenshotRegion,
+}) !c.CGImageRef {
+    const display_id = resolveDisplayId(input.display_index) catch {
+        return error.DisplayResolutionFailed;
+    };
+
+    if (input.region) |region| {
+        const rect: c.CGRect = .{
+            .origin = .{ .x = region.x, .y = region.y },
+            .size = .{ .width = region.width, .height = region.height },
+        };
+        const region_image = c.CGDisplayCreateImageForRect(display_id, rect);
+        if (region_image == null) {
+            return error.CaptureFailed;
+        }
+        return region_image;
+    }
+
+    const full_image = c.CGDisplayCreateImage(display_id);
+    if (full_image == null) {
+        return error.CaptureFailed;
+    }
+    return full_image;
+}
+
+fn resolveDisplayId(display_index: ?usize) !c.CGDirectDisplayID {
+    const selected_index = display_index orelse 0;
+    var display_ids: [16]c.CGDirectDisplayID = undefined;
+    var display_count: u32 = 0;
+    const list_result = c.CGGetActiveDisplayList(display_ids.len, &display_ids, &display_count);
+    if (list_result != c.kCGErrorSuccess) {
+        return error.DisplayQueryFailed;
+    }
+    if (selected_index >= display_count) {
+        return error.InvalidDisplayIndex;
+    }
+    return display_ids[selected_index];
+}
+
+fn writeScreenshotPng(input: struct {
+    image: c.CGImageRef,
+    output_path: []const u8,
+}) !void {
+    const path_as_u8: [*]const u8 = @ptrCast(input.output_path.ptr);
+    const file_url = c.CFURLCreateFromFileSystemRepresentation(
+        null,
+        path_as_u8,
+        @as(c_long, @intCast(input.output_path.len)),
+        0,
+    );
+    if (file_url == null) {
+        return error.FileUrlCreateFailed;
+    }
+    defer c.CFRelease(file_url);
+
+    const png_type = c.CFStringCreateWithCString(null, "public.png", c.kCFStringEncodingUTF8);
+    if (png_type == null) {
+        return error.PngTypeCreateFailed;
+    }
+    defer c.CFRelease(png_type);
+
+    const destination = c.CGImageDestinationCreateWithURL(file_url, png_type, 1, null);
+    if (destination == null) {
+        return error.ImageDestinationCreateFailed;
+    }
+    defer c.CFRelease(destination);
+
+    c.CGImageDestinationAddImage(destination, input.image, null);
+    const did_finalize = c.CGImageDestinationFinalize(destination);
+    if (!did_finalize) {
+        return error.ImageDestinationFinalizeFailed;
+    }
+}
+
 fn executeMouseMoveCommand(allocator: std.mem.Allocator, payload_json: []const u8) ![]const u8 {
     if (builtin.target.os.tag != .macos) {
         return makeErrorJson(allocator, "mouse-move is only supported on macOS");