Repository navigation
⚠ Error: 503 Service temporarily unavailable. Please try again shortly. #588
Description
Summary
I've tried switching between multiple models, but the result is always the same. This issue started on Thursday, July 23.
Before that, I had been using DeepSeek v4 Flash for nearly three weeks without any issues. Since Thursday, however, the platform has become extremely slow and laggy. Even simple tasks like reading files, analyzing code, or generating responses are taking much longer than they used to.
One thing I've noticed is that if I restart or reload the session, everything works normally for about 2–3 minutes, and then the lag comes back. Similarly, if I press Ctrl+C to interrupt a request and then submit a new prompt, it will work normally again for a couple of minutes before slowing down once more. This pattern has been consistent since Thursday, July 23, and it's becoming very frustrating.
I've already tried the following:
Restarted my session multiple times.
Refreshed the browser.
Switched between several different models (including GLM-5.2 with Max Effort).
Verified my internet connection and ran speed tests.
My internet connection is 300+ Mbps, I'm located in Southern California, and my ISP is Cox Communications, so I'm confident the bottleneck isn't on my end.
Yesterday, I also started receiving the following error:
⚠ Error: 503 Service temporarily unavailable. Please try again shortly.
Type "continue" to try again. If the issue persists, contact support:
https://commandcode.ai/discord
Trace ID: 1af675ea4313e5f0c1e9b87e822bea01
Is anyone else experiencing the same issue? Has there been a backend outage, infrastructure issue, or capacity problem since Thursday? The slowdown started before I began seeing the 503 errors, so I'm wondering if they're related.
I'm on the $100/month plan, and this is starting to impact my business. I rely on Command Code every day for software development, and up until Thursday everything worked great. Now the performance has degraded to the point where it's affecting my productivity.
If anyone has experienced something similar, found a workaround, or knows whether the team has acknowledged an issue, I'd really appreciate any information.
Thanks
⌘ Tracing… esc to interrupt • 30m 45s • ↓ 2.8k
Expected Behavior
I am expecting that it would take 30 seconds to process 2.8k
Actual Behavior
⌘ Studying… esc to interrupt • 35m 0s • ↓ 2.8k
Steps to reproduce the issue
Just typing a simple prompt like "remember to save session logs." takes an unusually long time to process. It seems like the application gets stuck every few minutes.
After I log in, everything works normally for about 2 minutes, but then it starts hanging again. In some cases, it can remain in a "thinking" or "processing" state for 50+ minutes without returning an error or producing any output. It just sits there indefinitely without making any progress.
This behavior has been consistently occurring, making the application unreliable to use.
Command Code Version
1.4.1
Operating System
Windows
Terminal/IDE
Unknown
Shell
cmd.exe
Additional context
I am using Windows Terminal (admin), nothing fancy.
whoami - aehcp\va_45
login
Microsoft Windows [Version 10.0.26200.8894]
(c) Microsoft Corporation. All rights reserved.
POWERSHELL [cmdc info]
│ ─── System Information ───────────────────────────────────
│ ● Version 1.4.1
└ ... +20 lines [ctrl+o to expand]
⠶ Command Code v1.4.1, Windows, AEHCP, 12 cores / 16 GB, Node v24.16.0.
POWERSHELL [cmdc whoami]
│
│ User Information:
│ ℹ Name: saazeacorphe0i
│ ℹ Email: saazeacorp@gmail.com
│ ℹ Username: saazeacorphe0i
│ #< CLIXML
│ - Fetching user information...
│ ✔ User information loaded
└ System.Management.Automation.PSCustomObjectS
ystem.Object1Preparing modules for
first use.0-1-1Completed-1
2Preparing modules for first use.0-1-1Completed-1
⠶ Logged in as saazeacorphe0i (saazeacorp@gmail.com).
Activity
saazeacorp-planter commentedon Jul 25, 2026
Trace ID: 3e58d7b8-772a-4b21-ba75-ebaa72cbdeec
Account: saazeacorphe0i · Version: 1.4.1 · Model: deepseek/deepseek-v4-pro (effort: high) ·
Provider: command-code
21 stalls > 2 min this session (2026-07-25), times in PDT (UTC−7):
┌───────────┬───────────────────────┬───────────┐
│ Start PDT │ Duration │ UTC start │
├───────────┼───────────────────────┼───────────┤
│ 11:31:28 │ 29m50s │ 18:31:28 │
├───────────┼───────────────────────┼───────────┤
│ 12:06:17 │ 15m30s │ 19:06:17 │
├───────────┼───────────────────────┼───────────┤
│ 12:21:47 │ 8m47s │ 19:21:47 │
├───────────┼───────────────────────┼───────────┤
│ 12:31:26 │ 5m51s │ 19:31:26 │
├───────────┼───────────────────────┼───────────┤
│ 12:42:39 │ 3m35s │ 19:42:39 │
├───────────┼───────────────────────┼───────────┤
│ 12:49:50 │ 59m19s ← ~1 hour hang │ 19:49:50 │
├───────────┼───────────────────────┼───────────┤
│ 13:51:37 │ 21m4s │ 20:51:37 │
├───────────┼───────────────────────┼───────────┤
│ 14:13:43 │ 3m8s │ 21:13:43 │
├───────────┼───────────────────────┼───────────┤
│ 14:17:38 │ 2m27s │ 21:17:38 │
├───────────┼───────────────────────┼───────────┤
│ 14:20:57 │ 10m23s │ 21:20:57 │
├───────────┼───────────────────────┼───────────┤
│ 14:46:14 │ 3m32s │ 21:46:14 │
├───────────┼───────────────────────┼───────────┤
│ 14:50:43 │ 2m36s │ 21:50:43 │
├───────────┼───────────────────────┼───────────┤
│ 14:59:15 │ 2m22s │ 21:59:15 │
├───────────┼───────────────────────┼───────────┤
│ 15:15:52 │ 3m4s │ 22:15:52 │
├───────────┼───────────────────────┼───────────┤
│ 15:21:17 │ 2m45s │ 22:21:17 │
├───────────┼───────────────────────┼───────────┤
│ 15:28:52 │ 2m17s │ 22:28:52 │
├───────────┼───────────────────────┼───────────┤
│ 15:32:40 │ 4m5s │ 22:32:40 │
├───────────┼───────────────────────┼───────────┤
│ 15:37:20 │ 3m52s │ 22:37:20 │
├───────────┼───────────────────────┼───────────┤
│ 15:44:33 │ 2m53s │ 22:44:33 │
├───────────┼───────────────────────┼───────────┤
│ 15:49:38 │ 2m11s │ 22:49:38 │
└───────────┴───────────────────────┴───────────┘
Total stalled time: ~3h 14m across 21 events.
Worst: 59m19s (12:49:50→13:49:09 PDT) — full hour with no output. 29m50s, 21m4s, 15m30s,
10m23s also multi-minute hangs.
Ruled out locally: CPU 16%, no proxy, valid auth, dev server unaffected (only the agent
runtime stalls). RAM pressure noted (1.6 GB free, node processes paged to disk with
negative working sets).
Question for support: Were these stalled turns reaching the deepseek-v4-pro endpoint and
hanging server-side, or never leaving the client? The hour-long 0-token hang suggests a
streaming connection opening but receiving nothing.
saazeacorp-planter commentedon Jul 26, 2026
Session for Support Investigation
┌──────────────────────────┬─────────────────────────────────────────────────────────────┐
│ Detail │ Value │
├──────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Session start │ July 25, 2026 ~10:30 AM PDT │
├──────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Current time │ July 25, 2026 ~6:00 PM PDT (still active) │
├──────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Session directory │ C:\Users\va_45\AppData\Local\Temp\commandcode\C--Users-va-… │
├──────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Project │ C:\Users\va_45\OneDrive\ESCRAP │
├──────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Project session logs │ C:\Users\va_45\OneDrive\ESCRAP\sessions\SESSION-041.md │
└──────────────────────────┴─────────────────────────────────────────────────────────────┘
Model Configuration
┌────────────────────────┬──────────────────────────────────┐
│ Setting │ Value │
├────────────────────────┼──────────────────────────────────┤
│ Main model │ deepseek/deepseek-v4-pro │
├────────────────────────┼──────────────────────────────────┤
│ Permissions mode │ auto-accept │
├────────────────────────┼──────────────────────────────────┤
│ Taste learning │ Enabled (user + project inherit) │
├────────────────────────┼──────────────────────────────────┤
│ Compaction model │ deepseek/deepseek-v4-pro │
├────────────────────────┼──────────────────────────────────┤
│ Title generation model │ deepseek/deepseek-v4-flash │
└────────────────────────┴──────────────────────────────────┘
Throughput / Performance Issues Observed
┌──────────┬───────────┬─────────────┬──────────────────┐
│ Incident │ Tokens │ Time │ Rate │
├──────────┼───────────┼─────────────┼──────────────────┤
│ Stall #1 │ ~2.7k │ 21m 49s │ ~2 tokens/min │
├──────────┼───────────┼─────────────┼──────────────────┤
│ Stall #2 │ ~14.2k │ 7m 49s │ ~30 tokens/min │
├──────────┼───────────┼─────────────┼──────────────────┤
│ Stall #3 │ ~42.5k │ 19m 23s │ ~36 tokens/min │
├──────────┼───────────┼─────────────┼──────────────────┤
│ Stall #4 │ ~2.7k │ 22+ minutes │ effectively dead │
├──────────┼───────────┼─────────────┼──────────────────┤
│ Stall #5 │ ~9.5k │ 4m 27s │ ~35 tokens/min │
├──────────┼───────────┼─────────────┼──────────────────┤
│ Stall #6 │ ~13.3k │ 301 minutes │ ~0.7 tokens/min │
├──────────┼───────────┼─────────────┼──────────────────┤
│ Stall #7 │ ~1.1k │ 4m 59s │ ~3.6 tokens/min │
├──────────┼───────────┼─────────────┼──────────────────┤
│ Stall #8 │ ~0 (dead) │ 9m 8s │ 0 tokens/min │
└──────────┴───────────┴─────────────┴──────────────────┘
Pattern: Repeatedly getting stuck processing small token counts (2-14k) while showing
spinner statuses like "Abracadabraing…", "Philosophizing…", "Contemplificating…" with
throughput dropping to zero. Not the project app — the project backend responds instantly
to API calls.
Platform Details
┌──────────────┬───────────────────────────────────────────────────┐
│ Detail │ Value │
├──────────────┼───────────────────────────────────────────────────┤
│ OS │ Windows (win32) │
├──────────────┼───────────────────────────────────────────────────┤
│ Shell │ PowerShell 5.1 (Windows) + cmd.exe │
├──────────────┼───────────────────────────────────────────────────┤
│ Node.js │ v24.16.0 │
├──────────────┼───────────────────────────────────────────────────┤
│ pnpm │ Workspace monorepo │
├──────────────┼───────────────────────────────────────────────────┤
│ Project type │ TypeScript monorepo (backend + frontend + shared) │
└──────────────┴───────────────────────────────────────────────────┘
Context Size
The session has accumulated ~40+ session logs read, 5+ full code files edited per bug fix,
and multiple long outputs from web_fetch and test runs. Context window may be heavily
loaded with tool outputs.
What Was Working
Tool calls execute correctly when they actually get invoked — TypeScript compiles, tests
pass (178/178), PostgreSQL responds, API endpoints return data. The bottleneck is
exclusively in the agent's own token processing/generation throughput, not in any project
service.
DXXS commentedon Sep 2, 2026
I've been getting "503 Service temporarily unavailable... Type "continue" to try again." regularly lately with Laguna-s-2.1
ahmadawais commentedon Sep 2, 2026
you're using super old version. update to latest 1.40.1.
npm i -g command-code@latestLaguna is a free model, and it's availability depends on capacity.
DXXS commentedon Sep 2, 2026
Not me; it says 1.40.1 (apparently auto-updated, which I'd prefer to turn off). This was even after 1am in the morning. Why don't you ~'try writing your code again' until you may be able to figure out how to send your models 'continue' automatically every so often ;-?