Dual Ollama setup challenge

[long sonnet]

I got it to work!!!

First I set up the fallback server in the auth section:
"auth": {
"profiles": {
"ollama:default": {
"provider": "ollama",
"mode": "api_key"
},
"ollama-extra-fallback:default": {
"provider": "ollama-extra-fallback",
"mode": "api_key"
},

Then I added it to the models section:
(first the primary)
"models": {
"mode": "merge",
"providers": {
"ollama": {
"baseUrl": "http://192.168.1.46:11434",
"apiKey": "OLLAMA_API_KEY",
"api": "ollama",
"models": [
{
"id": "gemma2:27b",

Then like this:
"ollama-extra-fallback": {
"baseUrl": "http://192.168.1.36:11434/v1",
"api": "openai-completions",
"apiKey": "ollama",
"models": [
{
"id": "gemma4:e2b",
"name": "gemma4:e2b",
"reasoning": false,
"input": [
"text"
],
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 128000,
"maxTokens": 8192
},

And finally under agents:
"agents": {
"defaults": {
"model": {
"primary": "ollama/qwen3.5:latest",
"fallbacks": [
"ollama-extra-fallback/gemma4:e2b",
"google/gemini-3-flash-preview"
]
},

@grave hound

[grave hound]

Oh great news! May I ask how many your agents setup ? I have 2 agents but I need each agent has own Ollama server.

[long sonnet]

I'm still in the n00b state - I wouldn't know how to set up multiple agents. So I'm just working with one. What I needed was a fallback in case of the primary Ollama-server went down.

[grave hound]

Oh you must try since you have more than 1 brain ready for the task 🙂

[long sonnet]

I keep running out of tokens. Even on my local LLM's 🤣 I just found that my OpenClaw had totally forgotten everything overnight, and I see that the token window of 64k - which I had set so it could fit my GPU - had reset.

How to handle things like this?

[grave hound]

for my case I have one run ollama without gpu and vram so which task that not hurry and too complex i would ask an agent that bind that that server to do it

[grave hound]

I'm noob too but did you setup memory for an agent ?

[long sonnet]

LOL I thought I did, by asking it to document everything he did. But I must have failed somehow. I'm just now instructing him how to work. "Him" being my Claw-instance

[grave hound]

I saw some user ask agent to summarize their work or chat as daily diary at the end of the day, maybe cron should help that to automate. must try 🙂