There are a few lists like this one: https://github.com/alvinreal/awesome-opensource-ai
Not sure what your needs are. There is a large variety of agentic frameworks listed. From things like LangChain to build your own to HermesAgent or Openclaw variants that are supposed to be more “secure”.
Personally I’m testing Agent Zero which runs in a docker container and is meant to be an orchestrator of multiple subagents.
I use Traefik as reverse proxy for local only services with let’s encrypt certificates. Just needed to a) register the subdomains and b) expose port 80 for the challenges without anything being served on that port.
Wireguard into my network and local DNS via Pihole to ensure proper local IPs. Works like a charm.
The system can only feed into your house line if there is an external 230V 50Hz power net working. The moment the net goes down there will be no power from your micro inverter to your house. It requires power to work and a stable 50 Hz similar to most bigger inverters.
If you combine a power station (battery) with the inverter, this battery might give you a 230V outlet with stable 50 Hz. But this is not the one you should plug back into your house (would require an unsafe cable).
Overall, lots of fear and wrong information in this thread about safety. There is a DIN norm for all of this and the current 800W limit exists to keep your power line from melting in the walls in the most extreme cases.
Totally reletable, I actually want the same functions. I will stick to something more conservative at the moment to test and also because it is relevant for my work.
AgentZero has builtin connectors to Email, Telegram and WhatsApp. You can install plugins for Matrix or Discord. Not sure about the security model as I do not use these currently.
However, it also has a secrets store that replaces secrets in tool calls locally. I use it for SSH access to another machine.
The a0 connector can be used to link it to your PC locally for projects that require file acces, etc.