Error whule talking on Telegram

[keen gull]

I have an error while talkin to my bot on Telegram even though I'm on OAuth : 21:07:04 [agent/embedded] embedded run agent end: runId=51b7b8cf-c7bd-4ac7-aca3-576bdd2e77cb isError=true error=⚠️ API rate limit reached. Please try again later.
21:07:04 [diagnostic] lane task error: lane=main durationMs=3724 error="FailoverError: ⚠️ API rate limit reached. Please try again later."
21:07:04 [diagnostic] lane task error: lane=session:agent:main:telegram:group:-1003673798985:topic:1 durationMs=3728 error="FailoverError: ⚠️ API rate limit reached. Please try again later."
21:07:08 [agent/embedded] embedded run agent end: runId=51b7b8cf-c7bd-4ac7-aca3-576bdd2e77cb isError=true error=⚠️ API rate limit reached. Please try again later.
21:07:08 [diagnostic] lane task error: lane=main durationMs=4185 error="FailoverError: ⚠️ API rate limit reached. Please try again later."
21:07:08 [diagnostic] lane task error: lane=session:agent:main:telegram:group:-1003673798985:topic:1 durationMs=4189 error="FailoverError: ⚠️ API rate limit reached. Please try again later."

[austere vector]

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

[untold pumice]

That error isn’t Telegram auth-related — it’s your LLM provider saying “429 / rate limit” (OpenClaw surfaces it as ⚠️ API rate limit reached), and then the Gateway wraps it as a FailoverError.

OAuth (or Claude setup-token) doesn’t mean unlimited; it’s still subject to whatever rate limits/quota your provider/account has.

What to do:

1. Confirm which provider is rate-limiting

• In the logs a few lines above that error there’s usually an HTTP 429 + provider name (Anthropic/OpenAI/etc).
• If you paste ~20 lines before the error (redact tokens), we can tell you exactly which one.

2. Quick fix

• Just wait a bit (rate limits are window-based), and avoid sending a bunch of messages rapidly in the Telegram group.

3. Longer-term fix

• If you’re on Anthropic subscription auth (setup-token): you may need to wait for the window to reset or upgrade/enable extra usage, or switch to an Anthropic API key for more predictable limits. (FAQ section: https://docs.openclaw.ai/start/faq → “Why am I seeing HTTP 429: rate_limit_error from Anthropic?”)
• Configure fallback models so the bot can keep replying when one provider is rate-limited: https://docs.openclaw.ai/concepts/model-failover
  (If you already have fallbacks configured but it still keeps failing, try starting a fresh session and/or restart the gateway — some versions/sessions can “stick” to the current model until a new session.)

If you can share:

• openclaw --version
• openclaw models status
• plus the log lines right before the 429/rate-limit error
  …I can give you the exact next step for your setup.

[keen gull]

[gateway] agent model: openai-codex/gpt-5.2

[untold pumice]

Good, that line tells us what’s actually happening:

[gateway] agent model: openai-codex/...

So the “⚠️ API rate limit reached” is OpenAI (Codex OAuth) rate-limiting you — not Telegram, and not an OpenClaw auth bug.

What this usually means / what to do:

• OAuth still has quotas + rate limits. If you send a few messages quickly (or you’re in a Telegram group and multiple people trigger the bot), you can hit the provider’s per‑minute limits and get 429’d for a bit.
  Fix: wait for the window to reset, then retry (and avoid rapid-fire messages).

• If you want reliability: set up a fallback (a second provider/model backed by an API key) so OpenClaw can keep replying when Codex OAuth is cooling down. Docs: https://docs.openclaw.ai/concepts/model-failover

• If you already have fallbacks configured but it still won’t switch: try a fresh session (e.g. /new in the chat) and/or restart the gateway after changing the model config. Some setups don’t move an existing session off the currently-selected model until a new session starts.

If you paste the few log lines just before the rate-limit message (they often include an HTTP 429 payload), we can confirm whether it’s a straight per-minute limit vs a temporary provider-side throttling.

[keen gull]

I can't seem to be able to publish a skill on clawhub : Mais la publication bloque encore : licence MIT‑0 non acceptée côté ClawHub (c’est un flag compte, pas token).

👉 Va sur clawhub.ai → Account/Settings → Accept MIT‑0.
Dès que c’est validé, je relance et je te donne le lien public.

[untold pumice]

Yep — for ClawHub publishing, that message is expected behavior.

ClawHub currently treats every published skill as MIT‑0 (it’s not “whatever license you put in your repo”), and it will block publishing until your account has explicitly accepted MIT‑0. The CLI docs even spell it out: “Publishing a skill means it is released under MIT‑0 on ClawHub.” (from docs/cli.md in the openclaw/clawhub repo)

What to do:

1. Go to https://clawhub.ai → Account / Settings
2. Find the MIT‑0 acceptance checkbox/banner and accept it (this is an account flag, not a token scope)
3. Then retry:
   • clawhub whoami (to confirm you’re logged in as the expected user)
   • clawhub publish ...

If it still says MIT‑0 not accepted after you tick it, the two usual culprits are:

• you’re logged into a different GitHub account in the browser vs the CLI token
• the CLI token is old → re-run clawhub login and retry publish

If you paste the exact clawhub publish ... output (just the error text, redact nothing sensitive), I can tell which one it is.

(Separate note: your Telegram “API rate limit reached” issue is unrelated — that one is upstream provider rate limiting, as discussed above.)

[keen gull]

Ok, i don't have this setting on clawhub though...

can't accept MIT-0 acceptance as there's no option to do that

[untold pumice]

You’re right — there’s no “MIT‑0 accept” toggle in the Account/Settings page. I misdirected you there.

How it works on ClawHub today is:

• The backend blocks publishing unless the publish request includes acceptLicenseTerms: true
• In the web UI, that’s a checkbox on the Publish/Upload page (not in settings):
  • go to https://clawhub.ai/upload
  • scroll to the License section
  • tick “I have the rights … and agree to publish it under MIT‑0”
  • then publish

If you’re publishing via the CLI (clawhub publish ...), it already sends acceptLicenseTerms: true automatically — so if you’re still getting the MIT‑0 error, you’re probably:

• using a different tool/script to publish (not the official clawhub CLI), or
• on an old CLI / wrong registry URL.

Can you paste the exact command you run to publish (and the exact error line)? That’ll tell us which path you’re on and what to fix.

Yeah — there’s no account-level setting for this.

“Accept MIT‑0” is per publish request:

If you’re using the website

You only get the checkbox on the skills upload page:

• Go to https://clawhub.ai/upload (make sure it’s clawhub.ai, not onlycrabs.ai)
• Sign in
• In the form there’s a License card with a checkbox like “I agree to publish under MIT‑0”
• You must tick that to publish a skill

If you’re on onlycrabs.ai (souls mode) or uploading a soul, that license checkbox won’t show.

If you’re using the CLI

clawhub publish is supposed to send acceptLicenseTerms: true automatically.
So if you still see the MIT‑0 error from the CLI, it usually means you’re on an older clawhub CLI (or using a different publisher script).