Telegram channel not starting on WSL - No polling logs | Friends of the Crustacean 🦞🤝 | Page 1

mystic atlas Apr 26, 2026, 7:36 PM

#

Hey guys, I think I found a bug or maybe I'm just stuck.

I'm running OpenClaw on WSL and my Telegram channel won't start at all.
The Telegram API is working fine (tested with curl), but the bot isn't polling any messages. Even after clearing updates and restarting the gateway, the [telegram] log never shows up in the terminal.
Is this a known issue with the current version on WSL?

System Info:
OS: WSL2 (Ubuntu)

Issue: Telegram polling won't initialize

Any help would be sick, thanks!🫠

white nicheBOT Apr 26, 2026, 7:36 PM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

mystic atlas Apr 26, 2026, 7:46 PM

#

@hollow pond

mystic atlas Apr 27, 2026, 8:12 AM

#

@hollow pond pls help this

hollow pondBOT Apr 27, 2026, 8:15 AM

#

Yeah, this looks very close to a known 2026.4.1 WSL2 Telegram startup regression: Telegram logs starting provider, never reaches getUpdates, and inbound messages don’t process. That issue was closed as fixed/implemented; the fix landed in v2026.4.5 by bounding Telegram startup calls so they can’t hang before polling begins. (api.github.com)

First thing: please paste this, with tokens redacted:

openclaw --version
openclaw gateway status
openclaw channels status --probe
openclaw channels logs --channel telegram --lines 100
openclaw status --all

If you’re on 2026.4.1–2026.4.4, update past 2026.4.5 and restart the gateway. If you’re already newer, then the likely causes shift to: gateway using a different config/env than your shell, Telegram token not visible to the service, channels.telegram.enabled/token config, webhook mode accidentally enabled, or a duplicate poller/token conflict. Telegram uses bot token config/env and does not use openclaw channels login telegram; long polling is the default unless webhookUrl is set. (docs.openclaw.ai)

Don’t paste the bot token here. 🦐

mystic atlas Apr 27, 2026, 12:18 PM

#

Port 18789 is already in use.

pid 5454 ahmad: openclaw-gateway (127.0.0.1:18789)
Gateway already running locally. Stop it (openclaw gateway stop) or use a different port.
Listening: 127.0.0.1:18789
Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting

🦞 OpenClaw 2026.4.24 (cbcfdf6) — I'm not magic—I'm just extremely persistent with retries and coping strategies.

│
◇
Gateway reachable.

Telegram default: enabled, configured, running, mode:polling, bot:@JarDaviss_bot, token:config, works

Tip: status --deep adds gateway health probes to status output (requires a reachable gateway).

hollow pondBOT Apr 27, 2026, 12:19 PM

#

That output is actually good news: Telegram is running and probing successfully now.

The Port 18789 is already in use bit just means you tried to start a second gateway while the real one is already running:

pid 5454 ahmad: openclaw-gateway
Telegram default: enabled, configured, running, mode:polling ... works

So the reason you don’t see [telegram] logs in that terminal is probably: you’re looking at the new failed foreground process, not the already-running gateway/service.

Try:

openclaw logs --follow

#

or just Telegram logs:

openclaw channels logs --channel telegram --lines 100

If the bot still doesn’t reply, the next likely thing is access/pairing, not polling. For a DM:

openclaw pairing list telegram
openclaw pairing approve telegram <CODE>

For groups, make sure the group is allowed under channels.telegram.groups and that you’re mentioning the bot unless requireMention: false is configured.

Docs: https://docs.openclaw.ai/channels/telegram 🦐

mystic atlas Apr 27, 2026, 12:21 PM

#

the bot is replied me but it said "all models are temporarily rate-limited. please try again in a few minutes"

hollow pondBOT Apr 27, 2026, 12:21 PM

#

Nice, that means Telegram is fixed — the bot is receiving messages and handing them to OpenClaw now. This is no longer a Telegram/WSL polling problem.

That error means the model provider side is failing/exhausted: current model + configured fallbacks are all in rate-limit/cooldown.

Please run:

openclaw models status --probe
openclaw models fallbacks list
openclaw logs --follow

Look for:

missing/expired auth
provider 429 / quota / billing errors
no fallbacks configured
all configured profiles in cooldown

#

Quick fixes are usually one of:

wait a few minutes if it’s a real temporary 429
add/fix provider auth via:
```
openclaw onboard
```
configure a fallback model/provider so one rate limit doesn’t stop replies:
```
openclaw models fallbacks add <provider/model>
```

For fresh model/provider recommendations, ask in #models — that changes too fast for me to safely guess. Docs: https://docs.openclaw.ai/concepts/models and https://docs.openclaw.ai/concepts/model-failover 🦐

mystic atlas Apr 27, 2026, 12:29 PM

#

0 fallback, and Google Generative AI API error (429): You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.

hollow pondBOT Apr 27, 2026, 12:29 PM

#

Yep — that explains it. Telegram is working now; the blocker is Gemini quota.

Because you have 0 fallbacks, OpenClaw tries Google, gets 429 quota exceeded, then has nowhere else to go, so it returns:

all models are temporarily rate-limited

Fix options:

Wait / increase Gemini quota or billing in Google AI Studio.

Add another model/provider as fallback so OpenClaw can recover automatically:

openclaw onboard
openclaw models fallbacks add <provider/model>
openclaw models status --probe

Or switch primary model:

#

   openclaw models set <provider/model>
   openclaw models status --probe

I won’t guess exact model picks here because that changes fast — ask in #models for current recommendations.

Docs:
https://docs.openclaw.ai/concepts/models
https://docs.openclaw.ai/concepts/model-failover 🦐

mystic atlas Apr 27, 2026, 12:31 PM

#

how to switch primary model?

hollow pondBOT Apr 27, 2026, 12:31 PM

#

Use the models CLI:

openclaw models list
openclaw models set <provider/model>
openclaw models status --probe

models set changes the global primary/default model.

If that provider/model is not configured yet, run:

openclaw onboard

or configure model/auth first, then run models set again.

#

For the current Telegram chat only, you can also send this to the bot:

/model

and pick/switch the session model there.

I won’t suggest exact model names here — ask in #models for current recommendations. Docs: https://docs.openclaw.ai/cli/models 🦐

mystic atlas Apr 27, 2026, 12:37 PM

#

sudah set ke flash 3. tapi pas saya suruh buka youtube dia hanya berjalan setenagh ingin membalas tapi diinterup dan tiba tiba rate limited ready in 11s

hollow pondBOT Apr 27, 2026, 12:38 PM

#

Itu masih rate limit model, bukan Telegram lagi.

“ready in 11s” artinya OpenClaw sudah masuk cooldown pendek dan akan coba lagi setelah ~11 detik. Kalau kamu switch ke Flash 3 tapi masih lewat Google/Gemini, kuotanya tetap bisa kena limit — apalagi task seperti “buka YouTube” biasanya butuh beberapa langkah/tool call, jadi lebih cepat kena 429.

Coba:

openclaw models status --probe
openclaw models fallbacks list
openclaw logs --follow

Kalau masih 429 dari Google, pilih salah satu:

tunggu quota window Google reset / upgrade billing di Google AI Studio
tambahkan fallback dari provider lain, bukan Google lagi:

#

   openclaw onboard
   openclaw models fallbacks add <provider/model>
   openclaw models status --probe

untuk test, pakai prompt kecil dulu setelah cooldown lewat, jangan langsung task browser berat.

Docs: https://docs.openclaw.ai/concepts/model-failover 🦐

mystic atlas Apr 27, 2026, 12:39 PM

#

prompt kecilnya apa?

hollow pondBOT Apr 27, 2026, 12:39 PM

#

Pakai prompt yang nggak perlu browser/tool, cuma chat pendek. Contoh kirim ke bot Telegram:

balas "ok" saja

atau:

sebutkan 3 warna primer dalam satu baris

atau:

ringkas: OpenClaw adalah asisten AI pribadi. Maksimal 1 kalimat.

#

Kalau prompt kecil ini masih kena rate limited, berarti kuota/model Google-nya memang lagi habis/cooldown. Kalau ini berhasil tapi “buka YouTube” gagal, berarti task browser/tool-nya yang terlalu berat untuk quota saat ini. 🦐

mystic atlas Apr 27, 2026, 12:46 PM

#

sudah saya tanya 3 warna dan dia menjawab dan minta apa identitasnya, pas saya balas malah gak respob lagi, tapi pas saya /status dia masih bisa

hollow pondBOT Apr 27, 2026, 12:47 PM

#

/status itu command lokal Gateway — jadi dia bisa balas walaupun model/agent normal lagi macet atau rate-limited.

Coba urut begini:

Tunggu cooldown lewat dulu, misalnya 30–60 detik.
Kirim ini sebagai pesan standalone ke bot:
```
/new
```
atau:
```
/reset
```
Lalu test lagi:
```
balas "ok" saja
```

#

Kalau masih diam, cek log pas kamu kirim pesan:

openclaw logs --follow

dan model probe:

openclaw models status --probe

Kalau di /status terlihat agent masih busy / running / rate-limited, berarti bukan Telegram — session/model-nya yang lagi ketahan. Solusi stabilnya tetap tambah fallback provider lain, karena Gemini free/quota sering kena limit walau prompt kecil. 🦐

mystic atlas Apr 27, 2026, 12:48 PM

#

dia balas lagi, ternyata dia ngerapiin sistem agar mengubah nama yang sya beri

hollow pondBOT Apr 27, 2026, 12:49 PM

#

Mantap, berarti alurnya normal 🎉

Yang terjadi kemungkinan: agent masih memproses task sebelumnya, jadi kelihatan seperti diam. /status tetap bisa balas karena itu command Gateway, bukan respons model biasa.

Tips biar nggak bingung:

/status

cek apakah dia masih running/busy.

Kalau mau mulai bersih:

/new

#

Kalau ingin menghentikan konteks lama dan test ringan:

balas "ok" saja

Dan tetap saran saya: tambahkan fallback provider/model lain, karena quota Gemini tadi sudah sempat kena limit. Kalau fixed, boleh right-click pesan solusi → Apps → ✅ Mark Solution biar keindeks Answer Overflow untuk user lain. 🦐

mystic atlas Apr 27, 2026, 12:50 PM

#

kok sering banget ya rate limited, padahal lagi working

hollow pondBOT Apr 27, 2026, 12:50 PM

#

Iya, itu normal kalau pakai provider yang quotanya kecil/ketat — “lagi working” justru bisa makin cepat kena rate limit.

Satu task OpenClaw bukan cuma 1 request model. Bisa jadi beberapa call:

baca instruksi/context
mutusin tool/browser action
proses hasil tool
bikin jawaban final
retry kalau ada error/interruption

Jadi task seperti buka YouTube jauh lebih “mahal” daripada tanya 3 warna.

Yang bisa dilakukan:

openclaw models status --probe

#

Kalau Google masih 429, solusinya:

openclaw models fallbacks add <provider/model>

sebaiknya fallback dari provider lain, bukan Google lagi. Atau ganti primary:

openclaw models set <provider/model>

Sementara itu: tunggu cooldown selesai, jangan kirim banyak pesan saat dia masih running, dan test pakai prompt kecil dulu. Untuk rekomendasi model/provider yang paling stabil sekarang, tanya di #models ya. 🦐

mystic atlas Apr 27, 2026, 12:51 PM

#

masalahnya saya mau gratis

hollow pondBOT Apr 27, 2026, 12:51 PM

#

Bisa gratis, tapi tradeoff-nya ya itu: rate limit sering 😅

Pilihan realistis:

Tetap pakai Gemini free
- cocok buat chat ringan
- jangan task berat seperti browser/YouTube terus-menerus
- tunggu cooldown kalau kena 429

Tambah fallback gratis lain

openclaw onboard
openclaw models scan
openclaw models fallbacks add <provider/model>

Kalau satu provider kena limit, OpenClaw coba fallback.

#

Local model
Gratis per request, tapi butuh hardware kuat. Untuk tool/browser agent, local model sering kurang stabil kecuali mesin/GPU-nya bagus.

Intinya: gratis bisa, tapi tidak akan se-stabil paid API. Untuk daftar opsi gratis yang lagi bagus sekarang, tanya di #models karena cepat berubah. 🦐

mystic atlas Apr 27, 2026, 12:53 PM

#

provider/model apa aja yang gratis selain gemini

hollow pondBOT Apr 27, 2026, 12:53 PM

#

Kalau mau gratis selain Gemini, jangan pakai daftar manual dari saya — cepat banget berubah. Cara paling aman: biarkan OpenClaw scan katalog free:

openclaw models scan

atau metadata-only:

openclaw models scan --no-probe

Biasanya opsi gratis paling realistis:

OpenRouter free models
Butuh OPENROUTER_API_KEY, walau modelnya free. Tetap ada rate limit.

#

Local model via Ollama / LM Studio
Gratis per request, tapi untuk agent + tools/browser sering kurang stabil kalau hardware biasa.
Provider free tier lain
Ini berubah-ubah, jadi tanya di #models untuk rekomendasi terbaru.

Setelah dapat model yang cocok:

openclaw models set <provider/model>
openclaw models fallbacks add <provider/model>
openclaw models status --probe

Docs: https://docs.openclaw.ai/concepts/models 🦐

mystic atlas Apr 27, 2026, 12:55 PM

#

rate limited terus nih udah coba 3 model gemini, padhal masih used 40k token / 1m token

hollow pondBOT Apr 27, 2026, 12:55 PM

#

Itu masuk akal: 40k / 1M token itu cuma TPM (tokens per minute). Gemini juga punya limit lain seperti RPM (requests per minute) dan RPD (requests per day). Kalau RPM/RPD kena, tetap 429 walaupun token masih jauh dari 1M. Google juga bilang limit berlaku per project, bukan cuma per API key. (ai.google.dev)

OpenClaw “working” bisa bikin banyak request kecil ke model, jadi cepat kena RPM.

Coba cek di Google AI Studio bagian rate limits/quota:

RPM model yang dipakai
RPD harian
apakah semua model Gemini masih di project/API key yang sama

Solusi gratis paling realistis:

openclaw models fallbacks add <provider/model>

tapi fallback-nya usahakan non-Google, karena kalau masih Gemini/project yang sama bisa tetap kena 429. Also jangan kirim banyak pesan saat agent masih running — tunggu dia selesai dulu.

Google AI for Developers

Rate limits | Gemini API | Google AI for Developers

mystic atlas Apr 27, 2026, 12:56 PM

#

^Cahmad@HP14S:~openclaw models fallbacks add <provider/model>l>
-bash: syntax error near unexpected token `newline'

hollow pondBOT Apr 27, 2026, 12:57 PM

#

Ah, <provider/model> itu placeholder, jangan diketik pakai tanda < >.

Formatnya begini:

openclaw models fallbacks add provider/model

Contoh bentuknya:

openclaw models fallbacks add openrouter/some-free-model

Tapi pilih model yang benar dulu dari daftar:

#

openclaw models list

atau cari free models:

openclaw models scan

Lalu copy salah satu model ref yang muncul, misalnya bentuk openrouter/..., dan pakai:

openclaw models fallbacks add MODEL_REF_YANG_MUNCUL
openclaw models fallbacks list
openclaw models status --probe

#

Jadi jangan literal <provider/model> ya — itu maksudnya “isi dengan nama model asli.” 🦐

mystic atlas Apr 27, 2026, 1:03 PM

#

saya --probe model status tapi sudah gak ada 429 padahal

hollow pondBOT Apr 27, 2026, 1:03 PM

#

Bisa begitu. models status --probe cuma test kecil sekali; itu bisa OK, tapi task Telegram/browser tetap kena limit karena:

task normal butuh beberapa request model, bukan cuma 1 probe
cooldown lama bisa sudah lewat saat probe dijalankan
limit yang kena bisa RPM / concurrency, bukan token total
session lama mungkin masih punya model/auth cooldown state

Coba reset session Telegram-nya:

/new

lalu test kecil:

#

balas "ok" saja

Kalau test kecil OK tapi task berat kena lagi, berarti quota Gemini masih terlalu ketat untuk workload itu.

Kalau mau bukti pastinya, jalankan log sambil kirim prompt yang gagal:

openclaw logs --follow

Cari baris error model/provider setelah prompt dikirim. Itu yang paling akurat dibanding probe. 🦐

mystic atlas Apr 27, 2026, 1:19 PM

#

padahal saya cuma nyruuh panggil nama saya, tapi terkena rate limited. dan di log terminal 429

hollow pondBOT Apr 27, 2026, 1:19 PM

#

Kalau prompt sesimpel “panggil nama saya” tetap kena 429 di log, berarti ini bukan karena prompt berat — quota/rate limit provider-nya memang sedang kena.

Beberapa hal penting:

429 dari Gemini bisa kena karena RPM/RPD/concurrency, bukan cuma token.
Walaupun models status --probe sempat OK, request berikutnya tetap bisa kena 429.
Kalau 3 model Gemini sama-sama kena, kemungkinan limitnya di project/API key Google yang sama.

Coba hentikan spam/retry dulu 5–10 menit, lalu:

openclaw models status --probe
openclaw models fallbacks list

Kalau masih mau gratis, tambah fallback non-Google. Jangan pakai placeholder:

#

openclaw models scan

copy model ref yang muncul, lalu:

openclaw models fallbacks add MODEL_REF

Kalau mau saya bantu baca tepatnya, paste 10–20 baris log sekitar error 429 — hapus token/API key kalau ada.

mystic atlas Apr 27, 2026, 1:22 PM

#

terjadi Error: Missing OpenRouter API key. Set OPENROUTER_API_KEY to run models scan pas saya scan

hollow pondBOT Apr 27, 2026, 1:22 PM

#

Iya, models scan default-nya mau probe live, jadi butuh OPENROUTER_API_KEY.

Kalau cuma mau lihat daftar kandidat gratis tanpa probe, pakai:

openclaw models scan --no-probe

Tapi untuk benar-benar dipakai inference, OpenRouter tetap butuh API key. Bisa ambil key di OpenRouter, lalu set di shell/gateway env:

export OPENROUTER_API_KEY="sk-or-..."
openclaw gateway restart

Kalau gateway jalan sebagai service, pastikan env itu masuk ke service juga — bukan cuma terminal sementara. Setelah itu:

#

openclaw models scan
openclaw models fallbacks add MODEL_REF
openclaw models status --probe

Jangan paste API key ke Discord. 🦐

mystic atlas Apr 27, 2026, 1:23 PM

#

gimana gimana? tolong bertahap

hollow pondBOT Apr 27, 2026, 1:23 PM

#

Oke, pelan-pelan. Ikuti ini satu per satu.

Step 1 — bikin OpenRouter API key

Buka OpenRouter → API Keys.
Buat key baru.
Jangan paste key ke Discord.

Step 2 — masukin key ke OpenClaw
Di terminal WSL:

openclaw onboard --auth-choice openrouter-api-key

Nanti paste API key saat diminta.

Step 3 — cek kebaca

#

openclaw models status --probe

Kalau OpenRouter muncul ok, lanjut.

Step 4 — scan model gratis

openclaw models scan

Kalau muncul daftar model, pilih/copy model ref yang bentuknya seperti:

openrouter/xxx/yyy:free

#

Step 5 — tambah sebagai fallback
Ganti MODEL_REF dengan yang kamu copy:

openclaw models fallbacks add MODEL_REF

Contoh format saja:

openclaw models fallbacks add openrouter/meta-llama/llama-3.3-70b-instruct:free

Step 6 — restart gateway

openclaw gateway restart

#

Step 7 — test Telegram
Kirim ke bot:

/new

lalu:

balas "ok" saja

Kalau ada error di step mana pun, paste errornya di sini — tapi jangan paste API key. 🦐

mystic atlas Apr 27, 2026, 1:32 PM

#

I understand this is personal-by-default and shared/multi-user use requires lock-down. Continue?

hollow pondBOT Apr 27, 2026, 1:32 PM

#

Itu prompt normal dari openclaw onboard.

Kalau ini untuk bot pribadi kamu sendiri, jawab:

atau:

yes

Maksudnya OpenClaw ngingetin: default-nya asisten pribadi, jadi kalau dipakai banyak orang/group publik harus dikunci dengan allowlist/pairing/config. Untuk setup pribadi Telegram kamu, lanjut saja. 🦐

mystic atlas Apr 27, 2026, 1:33 PM

#

itu terjadi pas saya openclaw auth choice openrouter api key

hollow pondBOT Apr 27, 2026, 1:33 PM

#

Iya, itu masih normal. openclaw onboard --auth-choice openrouter-api-key memang masuk ke flow onboarding dan menampilkan peringatan keamanan itu dulu.

Lanjut:

Saat muncul:
```
Continue?
```
jawab:
```
y
```
Kalau diminta API key OpenRouter, paste key kamu.
Setelah selesai:

#

   openclaw models status --probe

Kalau command kamu tadi bukan ini, pakai yang lengkap:

openclaw onboard --auth-choice openrouter-api-key

Bukan openclaw auth choice ... ya. 🦐

mystic atlas Apr 27, 2026, 1:34 PM

#

Config handling
│ ● Use existing values
│ ○ Update values
│ ○ Reset

hollow pondBOT Apr 27, 2026, 1:34 PM

#

Pilih:

Update values

Karena kamu mau menambahkan/mengubah auth OpenRouter API key.

Di menu itu pakai tombol panah bawah ke Update values, lalu Enter.

Jangan pilih Reset kecuali kamu memang mau reset config lama. 🦐

mystic atlas Apr 27, 2026, 1:36 PM

#

Telegram already configured. What do you want to do?
│ ○ Modify settings
│ ● Skip (leave as-is)

hollow pondBOT Apr 27, 2026, 1:36 PM

#

Pilih:

Skip (leave as-is)

Telegram kamu sudah jalan, jadi jangan diubah. Kita cuma mau tambah OpenRouter/model fallback. 🦐

mystic atlas Apr 27, 2026, 1:37 PM

#

Search provider
│
│ Search: _
│ ○ Brave Search
│ ○ Exa Search
│ ○ Firecrawl Search
│ ● Gemini (Google Search) (Requires Google Gemini API key · Google Search grounding · configured)
│ ○ Grok (xAI)
│ ○ Kimi (Moonshot)
│ ○ Ollama Web Search
│ ○ Perplexity Search
│ ○ SearXNG Search
│ ○ Tavily Search
│ ○ Skip for now
│ ↑/↓ to select • Enter: confirm • Type: to search

hollow pondBOT Apr 27, 2026, 1:37 PM

#

Pilih:

Skip for now

Ini untuk search provider, bukan model utama/fallback. Biar nggak nambah masalah dulu, skip saja. 🦐

mystic atlas Apr 27, 2026, 1:38 PM

#

Configure skills now? (recommended)
│ ● Yes / ○ No

hollow pondBOT Apr 27, 2026, 1:38 PM

#

Pilih:

No

Kita lagi fokus model/OpenRouter dulu. Skills bisa dikonfigurasi nanti kalau semua sudah stabil. 🦐

mystic atlas Apr 27, 2026, 1:38 PM

#

Gateway service already installed
│ ● Restart
│ ○ Reinstall
│ ○ Skip

hollow pondBOT Apr 27, 2026, 1:38 PM

#

Pilih:

Restart

Itu perlu supaya config/auth baru kebaca oleh gateway. 🦐

mystic atlas Apr 27, 2026, 1:39 PM

#

How do you want to hatch your bot?
│ ● Hatch in Terminal (recommended)
│ ○ Do this later

hollow pondBOT Apr 27, 2026, 1:40 PM

#

Pilih:

Do this later

Kamu sudah punya Telegram bot/gateway jalan. Kita nggak perlu hatch/terminal chat sekarang — setelah onboarding selesai, langsung cek:

openclaw models status --probe

🦐

mystic atlas Apr 27, 2026, 1:45 PM

#

Auth probes
┌───────────────────────────────┬──────────────────────────────┬────────────┐
│ Model │ Profile │ Status │
├───────────────────────────────┼──────────────────────────────┼────────────┤
│ google/gemini-3.1-pro-preview │ google:default (api_key) │ ok · 51.5s │
│ openrouter/openrouter/auto │ openrouter:default (api_key) │ ok · 34.9s │
└───────────────────────────────┴──────────────────────────────┴────────────┘
Probed 2 targets in 78.6s

hollow pondBOT Apr 27, 2026, 1:45 PM

#

Nice, OpenRouter auth sudah kebaca ✅

Sekarang tambah OpenRouter sebagai fallback:

openclaw models fallbacks add openrouter/auto

Kalau command itu error karena yang muncul di status adalah openrouter/openrouter/auto, pakai ini:

openclaw models fallbacks add openrouter/openrouter/auto

Lalu cek:

#

openclaw models fallbacks list
openclaw models status
openclaw gateway restart

Setelah itu di Telegram:

/new

lalu test:

balas "ok" saja

#

Kalau Gemini kena 429 lagi, OpenClaw harusnya punya jalur fallback ke OpenRouter. 🦐

mystic atlas Apr 27, 2026, 1:50 PM

#

Config overwrite: /home/ahmad/.openclaw/openclaw.json (sha256 62db237466ce3d75642059af4af19d6f7c683f23f2bdff2dfdd0f4ce6068bd92 -> bf5fc9a8b40fcbb497c0e1358ea6b8c03f24baff65f48a7184e3894ba496cd7c, backup=/home/ahmad/.openclaw/openclaw.json.bak)
Updated ~/.openclaw/openclaw.json
Fallbacks: openrouter/auto

hollow pondBOT Apr 27, 2026, 1:50 PM

#

Perfect — fallback sudah terpasang ✅

Sekarang lakukan:

openclaw gateway restart
openclaw models status

Lalu di Telegram kirim:

/new

Terus test:

#

balas "ok" saja

Kalau masih rate-limited, paste bagian log sekitar error dari:

openclaw logs --follow

Tapi sejauh ini config-nya sudah benar: Fallbacks: openrouter/auto. 🦐

mystic atlas Apr 27, 2026, 1:55 PM

#

saya cek /status dia masih menggunakan gemini flash, apakah harus saya ganti

hollow pondBOT Apr 27, 2026, 1:55 PM

#

Nggak harus. Primary tetap Gemini Flash, dan openrouter/auto sekarang jadi fallback. Jadi normal kalau /status masih menampilkan Gemini Flash saat Gemini belum gagal.

Artinya alurnya:

coba Gemini Flash dulu
kalau kena 429 / rate limit
baru pindah ke OpenRouter fallback

Kalau kamu mau langsung pakai OpenRouter sebagai primary, baru ganti:

openclaw models set openrouter/auto
openclaw gateway restart

Lalu di Telegram:

#

/new

Rekomendasi saya:

kalau Gemini kadang masih jalan → biarkan Gemini primary + OpenRouter fallback
kalau Gemini sering banget 429 → jadikan OpenRouter primary dulu 🦐

mystic atlas Apr 27, 2026, 2:05 PM

#

katanya dia gak bisa kemampuan buat ngontrol aplikasi atau browser di komputer lokal saya

hollow pondBOT Apr 27, 2026, 2:06 PM

#

Bisa jadi dia benar untuk kondisi sekarang: Telegram/model sudah jalan, tapi browser tool belum aktif/terkonfigurasi.

Cek bertahap dulu:

1. Cek browser tool ada atau tidak

openclaw browser status

2. Jalankan diagnosis browser

openclaw browser doctor

3. Coba buka browser test

#

openclaw browser start --headless
openclaw browser open https://youtube.com

Kalau command openclaw browser error/unknown, aktifkan browser plugin:

openclaw config set browser.enabled true --strict-json
openclaw config set tools.alsoAllow '["browser"]' --strict-json
openclaw gateway restart

Lalu di Telegram kirim:

#

/new

dan coba:

buka youtube pakai browser tool

Catatan: karena kamu pakai WSL, OpenClaw tidak otomatis bisa mengontrol Chrome Windows biasa. Yang paling gampang dulu adalah browser headless/managed di WSL. Kalau mau kontrol Chrome Windows asli, itu setup berbeda pakai remote CDP. Docs: https://docs.openclaw.ai/tools/browser dan https://docs.openclaw.ai/tools/browser-wsl2-windows-remote-cdp-troubleshooting 🦐

mystic atlas Apr 27, 2026, 2:08 PM

#

profile: openclaw
enabled: true
running: false
transport: cdp
cdpPort: 18800
cdpUrl: http://127.0.0.1:18800
browser: unknown
detectedBrowser: unknown
detectedPath: auto
profileColor: #FF4500

hollow pondBOT Apr 27, 2026, 2:08 PM

#

Browser tool enabled, tapi OpenClaw belum nemu browser executable:

running: false
detectedBrowser: unknown
detectedPath: auto

Di WSL, kemungkinan belum ada Chrome/Chromium di Linux side.

Coba:

which google-chrome chromium chromium-browser microsoft-edge brave-browser

Kalau tidak ada output, install Chromium/Chrome di WSL. Paling simpel di Ubuntu:

#

sudo apt update
sudo apt install -y chromium-browser

Kalau chromium-browser bermasalah karena Snap di WSL, pakai Google Chrome .deb:

wget https://dl.google.com/linux/direct/google-chrome-stable_current_amd64.deb
sudo apt install -y ./google-chrome-stable_current_amd64.deb

Lalu set path eksplisit:

openclaw config set browser.executablePath "/usr/bin/google-chrome"

#

openclaw gateway restart
openclaw browser status
openclaw browser start --headless

Kalau browser start sudah jalan, coba lagi dari Telegram setelah /new. 🦐

mystic atlas Apr 27, 2026, 2:15 PM

#

saya berhasil menginstall sudo apt update
sudo apt install -y chromium-browser

hollow pondBOT Apr 27, 2026, 2:16 PM

#

Mantap. Sekarang verifikasi path dan start browsernya:

which chromium-browser chromium google-chrome

Lalu:

openclaw browser doctor
openclaw browser status
openclaw browser start --headless
openclaw browser open https://youtube.com
openclaw browser snapshot

Kalau which menunjukkan path misalnya /usr/bin/chromium-browser, tapi OpenClaw masih detectedBrowser: unknown, set eksplisit:

#

openclaw config set browser.executablePath "/usr/bin/chromium-browser"
openclaw gateway restart
openclaw browser status
openclaw browser start --headless

Kalau browser snapshot berhasil, baru di Telegram:

/new

lalu:

#

buka youtube pakai browser tool

🦐

mystic atlas Apr 27, 2026, 2:22 PM

#

detectbrsr udh chronium tapi snapshot : GatewayClientRequestError: Error: Failed to start Chrome CDP on port 18800 for profile "openclaw". CDP diagnostic: http_unreachable after 1ms; cdp=http://127.0.0.1:18800; fetch failed.
Hint: If running in a container or as root, try setting browser.noSandbox: true.
Hint: No DISPLAY/X server was detected. Enable browser.headless: true, start Xvfb, or run the Gateway in a desktop session.
Chrome stderr:
/usr/bin/chromium-browser: 12: xdg-settings: not found
[9695:9695:0427/222036.566293:ERROR:ui/ozone/platform/x11/ozone_platform_x11.cc:256] Missing X server or $DISPLAY
[9695:9695:0427/222036.566409:ERROR:ui/aura/env.cc:246] The platform failed to initialize. Exiting

hollow pondBOT Apr 27, 2026, 2:22 PM

#

Error-nya jelas: Chromium jalan headed, tapi WSL kamu nggak punya X/DISPLAY. Set browser jadi headless permanen.

Jalankan:

openclaw config set browser.headless true --strict-json
openclaw gateway restart

Lalu test lagi:

openclaw browser status
openclaw browser start --headless
openclaw browser open https://youtube.com
openclaw browser snapshot

#

Kalau masih error soal sandbox/root, tambahkan ini:

openclaw config set browser.noSandbox true --strict-json
openclaw gateway restart

Soal xdg-settings: not found biasanya bukan blocker utama, tapi bisa dibereskan dengan:

sudo apt install -y xdg-utils

Mulai dari browser.headless true dulu. 🦐

mystic atlas Apr 27, 2026, 2:33 PM

#

bahkan dia gak bisa akses file eksplorer saya

hollow pondBOT Apr 27, 2026, 2:33 PM

#

Itu beda hal ya:

OpenClaw di WSL bisa akses file, tapi bukan “File Explorer” GUI Windows. Dia jalan di Linux/WSL, jadi path Windows harus lewat /mnt/c/....

Coba test begini:

ls /mnt/c/Users
openclaw config get agents.defaults.workspace

Kalau user Windows kamu ahmad, file Windows biasanya ada di:

/mnt/c/Users/ahmad/Documents
/mnt/c/Users/ahmad/Downloads

#

/mnt/c/Users/ahmad/Desktop

Lalu di Telegram jangan bilang “buka file explorer saya”, tapi kasih path jelas, misalnya:

tolong lihat daftar file di /mnt/c/Users/ahmad/Downloads

Lebih aman: buat folder khusus, misalnya OpenClawWorkspace, taruh file yang mau dibaca di situ, lalu set workspace ke folder itu:

openclaw config set agents.defaults.workspace "/mnt/c/Users/ahmad/OpenClawWorkspace"
openclaw gateway restart

#

Jangan set workspace ke seluruh C:\Users\ahmad kalau nggak perlu — itu terlalu luas/privacy risk.

Kalau maksudnya benar-benar mengontrol aplikasi File Explorer Windows, itu belum otomatis dari WSL headless. Untuk sekarang fokusnya: kasih akses lewat path /mnt/c/... atau folder workspace khusus. 🦐

mystic atlas Apr 27, 2026, 2:37 PM

#

Saya tidak dapat mengakses jalur /mnt/c/Users/ahmad/Downloads karena berada di sistem file lokal Anda dan di luar lingkungan aman saya. Saya hanya dapat mengakses file di dalam direktori kerja OpenClaw (/home/ahmad/.openclaw/workspace) atau file yang Anda berikan secara langsung.

hollow pondBOT Apr 27, 2026, 2:37 PM

#

Itu berarti bot kamu lagi dibatasi ke workspace — itu aman, tapi jadi nggak bisa baca /mnt/c/Users/ahmad/Downloads.

Cara paling gampang + aman: buat folder khusus untuk OpenClaw di Windows, lalu jadikan workspace.

1. Buat folder khusus

mkdir -p /mnt/c/Users/ahmad/OpenClawWorkspace

Taruh/copy file yang mau dibaca ke folder itu dari File Explorer Windows.

2. Set workspace OpenClaw

openclaw config set agents.defaults.workspace "/mnt/c/Users/ahmad/OpenClawWorkspace"
openclaw gateway restart

#

3. Di Telegram kirim

/new

Lalu test:

lihat daftar file di workspace

Kalau kamu mau cek batasannya sekarang:

openclaw config get agents.defaults.workspace
openclaw config get tools.fs.workspaceOnly
openclaw config get agents.defaults.sandbox.mode

#

Saya tidak sarankan langsung kasih akses ke seluruh Downloads/C drive. Lebih aman pakai folder khusus dulu. Docs: https://docs.openclaw.ai/concepts/agent-workspace 🦐

mystic atlas Apr 27, 2026, 4:22 PM

#

+-----------------------------------------------------------------------------------+
Config overwrite: C:\Users\ahmad.openclaw\openclaw.json (sha256 81e313e783e8e1a18a4a77aaf1a85f3728cb4579e8d637269265ea7c598b35aa -> 3b1d6de9ec1381d0eaa27882b0d9e83a6d508bd374b5d40507f487f963543461, backup=C:\Users\ahmad.openclaw\openclaw.json.bak)
|
o Gateway service runtime --------------------------------------------+
| |
| QuickStart uses Node for the Gateway service (stable + supported). |
| |
+----------------------------------------------------------------------+
|

Gateway service already installed
| > Restart
| Reinstall
| Skip (KONTENS: SAYA INGIN COBA DI WINDOWS AGAR BISA MENGAKSES FILE DAN KOMPITER LOKAL

hollow pondBOT Apr 27, 2026, 4:23 PM

#

Pilih Restart.

Tapi penting: kalau kamu pindah coba native Windows, jangan biarkan gateway WSL jalan bareng dengan token Telegram yang sama — nanti bisa konflik polling.

Urutannya:

1. Di WSL, stop gateway lama

openclaw gateway stop

2. Di wizard Windows ini, pilih

Restart

3. Di PowerShell Windows, cek

#

openclaw status --deep
openclaw channels status --probe

4. Buat workspace Windows khusus

mkdir C:\Users\ahmad\OpenClawWorkspace
openclaw config set agents.defaults.workspace "C:/Users/ahmad/OpenClawWorkspace"
openclaw gateway restart

Lalu di Telegram:

/new

#

Test:

lihat daftar file di workspace

Catatan: native Windows bisa lebih dekat ke file Windows, tapi WSL2 masih jalur yang paling direkomendasikan/stabil. Jangan kasih akses seluruh C:\Users\ahmad dulu — pakai folder khusus lebih aman. 🦐

mystic atlas Apr 27, 2026, 5:58 PM

#

sudah bisa akses file, tapi gak bisa buka browser sebagai perintah untuk membuka aplikasi, dan juga tidak bisa menscreenshot, padahal pas awal banget nyoba dia bisa.

hollow pondBOT Apr 27, 2026, 5:58 PM

#

Iya, itu dua kemampuan berbeda:

Akses file Windows ✅ sudah bisa
Kontrol browser / screenshot perlu browser tool aktif dan browser path benar
Screenshot desktop / kontrol aplikasi Windows biasa belum sama dengan browser tool

Coba di PowerShell Windows:

openclaw browser doctor
openclaw browser status

Kalau browser belum jalan, set Chrome/Edge:

#

openclaw config set browser.enabled true --strict-json
openclaw config set tools.alsoAllow '["browser"]' --strict-json
openclaw config set browser.headless false --strict-json
openclaw config set browser.executablePath "C:/Program Files/Google/Chrome/Application/chrome.exe"
openclaw gateway restart

Kalau pakai Edge:

openclaw config set browser.executablePath "C:/Program Files (x86)/Microsoft/Edge/Application/msedge.exe"
openclaw gateway restart

Lalu test:

#

openclaw browser start
openclaw browser open https://youtube.com
openclaw browser snapshot
openclaw browser screenshot

Kalau ini berhasil, di Telegram kirim:

/new

lalu:

#

buka youtube pakai browser tool

Catatan: browser screenshot = screenshot halaman browser, bukan screenshot seluruh desktop/File Explorer. Untuk kontrol aplikasi Windows umum + screenshot desktop, native Windows masih lebih terbatas. 🦐

hollow pondBOT Apr 27, 2026, 7:08 PM

#

I deleted that message because it exposed your Telegram bot token publicly.

Please rotate it immediately:

Open Telegram → @BotFather
Send:
```
/revoke
```
Choose your bot
Copy the new token
Update OpenClaw config with the new token / rerun Telegram setup
Restart:
```
openclaw gateway restart
```

#Telegram channel not starting on WSL - No polling logs