Enhance logging and error handling across tools; add full tool audit and Playwright tests

- Added model mismatch warnings in colorize, enhance-faces, and upscale routes.
- Improved error handling in colorize, enhance_faces, remove_bg, restore, and upscale scripts with detailed logging.
- Updated Dockerfile to align NCCL versions for compatibility.
- Introduced a new full tool audit script to test all tools for functionality and GPU usage.
- Created Playwright E2E tests for GPU-dependent tools to ensure proper functionality and performance.

packages/ai/src/bridge.ts

@@ -32,11 +32,26 @@ function extractPythonError(error: unknown): string {
           if (trimmed && !trimmed.startsWith("Traceback")) {
             return trimmed;
           }
+          // Extract the last meaningful line from a Python Traceback
+          if (trimmed) {
+            const lines = trimmed
+              .split("\n")
+              .map((l) => l.trim())
+              .filter(Boolean);
+            const lastLine = lines[lines.length - 1];
+            if (lastLine && lastLine !== "Traceback (most recent call last):") {
+              return lastLine;
+            }
+          }
         }
       }
     }
     if (execError.message) return execError.message;
+    // Return empty string for {stdout, stderr} objects with no useful content
+    // so the caller's fallback message (e.g. exit code) kicks in.
+    return "";
   }
+  if (error instanceof Error) return error.message;
   return String(error);
 }
 
@@ -85,6 +100,9 @@ function startDispatcher(): ChildProcess | null {
           if (parsed.ready === true) {
             dispatcherReady = true;
             dispatcherGpuAvailable = parsed.gpu === true;
+            console.log(
+              `[bridge] Python dispatcher ready (GPU: ${parsed.gpu === true})`,
+            );
             continue;
           }
 
@@ -97,7 +115,11 @@ function startDispatcher(): ChildProcess | null {
             }
           }
         } catch {
-          // Not JSON - collect as error output for pending requests
+          // Not JSON - forward diagnostic messages to Node.js logger,
+          // collect the rest as error output for pending requests.
+          if (trimmed.startsWith("[")) {
+            console.log(`[python] ${trimmed}`);
+          }
           for (const req of pendingRequests.values()) {
             req.stderrLines.push(trimmed);
           }
@@ -121,14 +143,17 @@ function startDispatcher(): ChildProcess | null {
           if (pending) {
             pendingRequests.delete(reqId);
             if (response.exitCode !== 0) {
-              pending.reject(
-                new Error(
-                  extractPythonError({
-                    stdout: response.stdout,
-                    stderr: pending.stderrLines.join("\n"),
-                  }) || `Python script exited with code ${response.exitCode}`,
-                ),
-              );
+              const errText =
+                extractPythonError({
+                  stdout: response.stdout,
+                  stderr: pending.stderrLines.join("\n"),
+                }) ||
+                (response.exitCode === 137
+                  ? "Process killed (out of memory) — try a lighter model or smaller image"
+                  : response.exitCode === 139
+                    ? "Process crashed (segmentation fault)"
+                    : `Python script exited with code ${response.exitCode}`);
+              pending.reject(new Error(errText));
             } else {
               pending.resolve({
                 stdout: response.stdout || "",
@@ -143,6 +168,7 @@ function startDispatcher(): ChildProcess | null {
     });
 
     child.on("error", (err: NodeJS.ErrnoException) => {
+      console.error(`[bridge] Dispatcher error: ${err.message} (code: ${err.code})`);
       if (err.code === "ENOENT") {
         // Venv python not found - mark as failed, will fall back to per-request
         dispatcherFailed = true;
@@ -307,7 +333,7 @@ function runPythonPerRequest(
         }
       });
 
-      child.on("close", (code) => {
+      child.on("close", (code, signal) => {
         clearTimeout(timer);
 
         if (stderrBuffer.trim()) {
@@ -322,7 +348,16 @@ function runPythonPerRequest(
         const stderr = stderrLines.join("\n");
 
         if (code !== 0) {
+          // When the process was killed by a signal, use a clear message
+          // instead of surfacing unrelated stderr (e.g. CUDA warnings).
+          const signalMsg =
+            signal === "SIGKILL" || code === 137
+              ? "Process killed (out of memory) — try a lighter model or smaller image"
+              : signal === "SIGSEGV" || code === 139
+                ? "Process crashed (segmentation fault)"
+                : null;
           const errorText =
+            signalMsg ||
             extractPythonError({ stdout: stdout.trim(), stderr }) ||
             `Python script exited with code ${code}`;
           rejectPromise(new Error(errorText));
@@ -361,6 +396,9 @@ export function runPythonWithProgress(
       // Retry in an isolated per-request process which starts clean and has
       // more available memory than the warm dispatcher.
       if (err.message === "Python dispatcher exited unexpectedly") {
+        console.warn(
+          `[bridge] Dispatcher crashed during ${scriptName}, retrying with per-request process`,
+        );
         return runPythonPerRequest(scriptName, args, options);
       }
       throw err;