UK

[rain stream]

Northampton lets build!!

[frigid heron]

Bicester, so pretty close 🙂

[abstract needle]

Swansea, but travel to Reading often.

Using openclaw with llama directly on Windows laptop. Newb. 1st/2nd line level tech skills.

Are either of you using open-source models? Ian

[rain stream]

Nice meeting you @frigid heron it was awesome hope we meet on town sometime share some ideas

[rain stream]

Oh we all are newb here mate, i am about to use open-source soon. I need to finish some ongoing projects yet

[abstract needle]

I tried and failed to set OpenClaw up with Ollama (locally hosted). Today I changed to OpenAI, via my Plus subscription and OAuth which worked seamlessly. Got it linked with WhatsApp, but struggling with cronies jobs.

[frigid heron]

I tried GLM 4.7 Flash, but it couldn't orchestrate well. Minimax via OpenRouter is OK, but I decided to pony up for Claude Max ;0

[abstract needle]

Nice, people rave about Claude Max

[frigid heron]

It really does feel different ...without it, it's just not got the soul

I code professionally with GPT 5.2, and also use Big Pickle via OpenCode, which is OK

Get OpenClaw to delegate coding tasks to Big Pickle via OpenCode to save Claude tokens

& I use Minimax for subagents for research

[abstract needle]

That sounds like a great setup.

Did you give your OpenClaw access to email?

I am working on that currently (with the caveat of not being able to 'send' hopefully).

[frigid heron]

Yeah, email works...not my email though 😉

Can get a feel for Opus via 1st month free AntiGravity sub...then cancel...but warning that's what hooked me 😉
Just not enough quota for serious use

[abstract needle]

Addictive I'm sure!

I am unable to get OpenClaw to 'send me a reminder WhatsApp message in 10 minutes'.

Struggling with the resolution. It may be that with OAuth there is no API key.

Issue summary (from OpenClaw)...

Cron jobs stay queued in ~/.openclaw/cron/jobs.json, but cron.runs never shows anything and nextWakeAtMs stays null, so reminders never execute.

• Calling openclaw cron run … --force after the scheduled time still says “not-due,” meaning the scheduler refuses to mark jobs ready even though the clock has passed.

• Restarting the gateway, checking cron.status, and injecting system events manually all succeed, so the gateway is running—but the cron scheduler isn’t waking jobs to fire the WhatsApp reminders.

• Workaround in place: background PowerShell task now sends WhatsApp message directly after 10 minutes instead of relying on cron, so you still get the reminder while the scheduler issue is investigated further.

[crimson hamlet]

same, seems like a bug in the system https://discord.com/channels/1456350064065904867/1467462530225148005

[lunar ibex]

How did you manage to do this? I thought OpenAI integration could only be achieved via an OpenAI API key which is a pay-per-use payment plan and not under the Plus sub.

[abstract needle]

I thought so to, but that's not the case. You can connect utilising ChatGPT subscription or Claude subscription, you just get rate limited more often. If you can afford it, API key is best.

Follow the auto setup with Openclaw and select OpenAI as the model, it then give you option of OAuth or API.

If you get stuck, I can try and help more if you send screen grabs (never include keys though).

[urban tide]

im in london

im trying to program my bot with skills via hostinger but it seems to run me in diffrent directions , my gbt keeps trying to make me dowload it from the vps but ive come to realise the hostinger vps terminal seems to be seperate from the ui it lauched for me when i fist setup

[lunar ibex]

Thx. this is enough info for Claude to take it and implement.

[abstract needle]

Hey all. I am running ChatGPT via OAuth, but having trouble enabling memory (memory search). Has anyone else experienced this?

My openclaw bot advises:

Memory search needs access to an embeddings endpoint, and the ChatGPT (Codex) OAuth token only covers chat/completions — it doesn’t grant embedding access. The docs spell this out: semantic memory search will stay disabled unless you configure a real embeddings key (e.g., OPENAI_API_KEY for OpenAI embeddings, GEMINI_API_KEY, or a local model). OAuth alone can’t unlock that.

If you ever do decide to add an embeddings key, let me know which provider you’d prefer and I’ll walk you through updating the gateway config (memorySearch.provider, models.providers.*.apiKey, or environment variables) and restarting so the vector index can be rebuilt.