storyboard-bridge 0.3.1 → 0.3.3

package/index.mjs

@@ -26,13 +26,13 @@
 //   --cdp <url>       CDP_HTTP            the debug-Chrome endpoint (default http://localhost:9222)
 //   --                CLAUDE_BIN/GEMINI_BIN   path to the CLI binary if not on PATH
 import { spawn, execFile } from 'node:child_process';
-import { writeFile, readFile, unlink, stat, mkdir } from 'node:fs/promises';
+import { writeFile, readFile, unlink, stat, mkdir, mkdtemp, rm } from 'node:fs/promises';
 import { join, resolve, dirname, sep } from 'node:path';
 import { fileURLToPath } from 'node:url';
 import { tmpdir } from 'node:os';
 import { WebSocket } from 'ws';
 
-const VERSION = '0.3.1';
+const VERSION = '0.3.3';
 
 // ---- config ----
 const arg = (name, fallback) => {
@@ -71,11 +71,11 @@ function wsUrl(httpUrl) {
 }
 
 // ---- the same CLI contract the backend uses: `<bin> -p <prompt> --output-format json [--model m]` ----
-function runCli(bin, args, stdinInput) {
+function runCli(bin, args, stdinInput, cwd) {
   return new Promise((resolve, reject) => {
     // On Windows the CLI is usually a `.cmd` shim (claude.cmd) that spawn() can't launch without a
     // shell; shell:true uses PATHEXT to resolve it. POSIX (Linux/WSL/macOS) spawns the binary directly.
-    const child = spawn(bin, args, { timeout: JOB_TIMEOUT_MS, shell: IS_WIN });
+    const child = spawn(bin, args, { timeout: JOB_TIMEOUT_MS, shell: IS_WIN, cwd });
     let out = '';
     let err = '';
     child.stdout.on('data', (d) => (out += d));
@@ -98,7 +98,7 @@ function extractText(stdout) {
   }
   return stdout;
 }
-async function runJob(prompt) {
+async function runJob(prompt, images) {
   // Pass the prompt via STDIN, not argv. On Windows the bridge spawns through cmd.exe (shell:true, to
   // resolve claude.cmd), and a big multi-line prompt as an arg gets mangled — cmd splits it on newlines,
   // breaking the command ("'claude' is not recognized"). stdin sidesteps shell parsing entirely; `claude
@@ -106,6 +106,25 @@ async function runJob(prompt) {
   const args = ['-p'];
   if (USE_JSON) args.push('--output-format', 'json'); // claude wraps as JSON; gemini prints raw text
   if (MODEL) args.push('--model', MODEL);
+  if (images?.length) {
+    // vision: write the attached image(s) into a temp dir, then point the prompt's `@name` refs at each
+    // file's ABSOLUTE path. Claude/Gemini resolve a bare `@name` against their own workspace/home (e.g.
+    // C:\Users\<you>), NOT our spawn cwd — so a relative ref silently misses and the model can't see the
+    // image (it goes looking for "image.png" and reports it can't find it). Absolute paths are cwd-independent.
+    const dir = await mkdtemp(join(tmpdir(), 'sbvis-'));
+    try {
+      let p = prompt;
+      for (const im of images) {
+        const name = String(im.name);
+        const abs = join(dir, name);
+        await writeFile(abs, Buffer.from(String(im.data), 'base64'));
+        p = p.split('@' + name).join('@' + abs);
+      }
+      return extractText(await runCli(BIN, args, p, dir));
+    } finally {
+      await rm(dir, { recursive: true, force: true });
+    }
+  }
   return extractText(await runCli(BIN, args, prompt));
 }
 
@@ -230,7 +249,7 @@ function connect() {
     } else if (msg.type === 'job' && msg.id) {
       const t0 = Date.now();
       try {
-        const text = await runJob(String(msg.prompt ?? ''));
+        const text = await runJob(String(msg.prompt ?? ''), Array.isArray(msg.images) ? msg.images : null);
         ws.send(JSON.stringify({ type: 'result', id: msg.id, text }));
         log(`job ${msg.id} ✓ (${Date.now() - t0}ms)`);
       } catch (e) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "storyboard-bridge",
-  "version": "0.3.1",
+  "version": "0.3.3",
   "description": "Desktop bridge that powers a hosted Storyboard AI webapp with your own local Claude Code / Gemini CLI login (no API key).",
   "type": "module",
   "bin": {