Most of our infrastructure has been built by an agent, inferencebot, running on our tokens but hosted on Nathan's Raspberry Pi. We should convert this key resource to a commons of the co-op by shifting it to our infrastructure.
Why does it matter?
To make the co-op more truly a co-op, it should not rely on a single member's personal tool.
How might it work?
The trick is we need to avoid two risks:
If our infrastructure breaks, the agent should be able to keep running to help fix it.
The agent should not have root access to its own environment, in case any rogue behavior emerges.
As a result, it is probably ideal to keep the agent(s) off our main Cloudron server, and with a fallback inference provider in case our systems fail.
Anything else?
Current setup:
Hermes agent profile on Raspberry Pi
Main gateway is a Matrix room
Main inference: co-op flash models
Fallback inference: Nathan's Ollama API key
API keys to: Cloudron, Gitea
## What's the idea?
Most of our infrastructure has been built by an agent, inferencebot, running on our tokens but hosted on Nathan's Raspberry Pi. We should convert this key resource to a commons of the co-op by shifting it to our infrastructure.
## Why does it matter?
To make the co-op more truly a co-op, it should not rely on a single member's personal tool.
## How might it work?
The trick is we need to avoid two risks:
* If our infrastructure breaks, the agent should be able to keep running to help fix it.
* The agent should not have root access to its own environment, in case any rogue behavior emerges.
As a result, it is probably ideal to keep the agent(s) off our main Cloudron server, and with a fallback inference provider in case our systems fail.
## Anything else?
Current setup:
* Hermes agent profile on Raspberry Pi
* Main gateway is a Matrix room
* Main inference: co-op flash models
* Fallback inference: Nathan's Ollama API key
* API keys to: Cloudron, Gitea
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
What's the idea?
Most of our infrastructure has been built by an agent, inferencebot, running on our tokens but hosted on Nathan's Raspberry Pi. We should convert this key resource to a commons of the co-op by shifting it to our infrastructure.
Why does it matter?
To make the co-op more truly a co-op, it should not rely on a single member's personal tool.
How might it work?
The trick is we need to avoid two risks:
As a result, it is probably ideal to keep the agent(s) off our main Cloudron server, and with a fallback inference provider in case our systems fail.
Anything else?
Current setup:
Duplicative with #17