• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Hermes - Your new best friend

M

mashie

Mawd Gawd
Joined
Oct 25, 2000
Messages
1,761
I have been using Hermes for a month and a half now and it doesn't cease to impress me.

At home I have it hosted on my main Linux PC while the LLM lives on a DGX Spark cluster. Either running 2 or 4 nodes at the moment depending on what I'm doing. It will be permanently 4 nodes down the road.

Current LLM is Deepseek v4 Flash 0731 running FP8 KV cache with a 1M context length.

At work I have a similar setup in our test lab, a cluster of 2 DGX Sparks running DSv4. I have just had Hermes migrate itself from a VM to a bare metal host that I originally had as a worker node.

The worker node used to have one RTX 5090, as of last week it has two of them. I had Hermes sort out the fix to enable P2P between the cards and then evaluate what backend was the faster for serving Qwen 3.8 27B FP8 with TP=2. SGLang turned out to be a lot faster than vLLM that originally was used.

The 5090s are used for delegations, need to see if I also can make them work as a fallback LLM next in case I take down the Sparks for maintenance.

Fun times ahead and so far I have only scratched the surface of what these can do.
 
Some fun today, wanted to get a web interface for the setup at work. Had Hermes install https://github.com/nesquena/hermes-webui and promptly realised it was designed a single user experience.

Well, nothing that a good little harness can't fix. I asked Hermes to fork it, investigate if it could become multi-tenant, including admin dashboard for users management and stuff. Well here we are, I have a multi-tenant Hermes Web GUI deployed. Took about 35 minutes to code my initial ask and then another hour of me asking for enhancements, like a central push button to get all profiles updated to whatever models were set on the admin account.
 
Not sure if this thread is intended to be a Hermes catch all, but more huge things.

https://x.com/NousResearch/status/2094515104670715940
https://github.com/NousResearch/hermes-agent/releases/tag/v2026.8.31

Another massive Hermes release - Pantheon

The Pantheon Release. v0.20.0 made Hermes the herald — he spoke, and he carried word to other agents. In v0.21.0 the gods assemble. Bot Mode ships built into the desktop app: a society of named agents with their own faces and group chats, where your bots talk to each other — and to you — like a team, not a toolbox. Around that spine: cron jobs gained memory and continuity so scheduled agents actually learn between runs, subagents can be steered live mid-flight, the MCP surface became a real command center, and the agent can now drive the desktop's own browser. This release rolls up everything from the v0.20.1–v0.20.6 infrastructure patch tags — those windows are fully documented here
 
4saken said:
The Pantheon Release. v0.20.0 made Hermes the herald — he spoke, and he carried word to other agents. In v0.21.0 the gods assemble. Bot Mode ships built into the desktop app: a society of named agents with their own faces and group chats, where your bots talk to each other — and to you — like a team, not a toolbox. Around that spine: cron jobs gained memory and continuity so scheduled agents actually learn between runs, subagents can be steered live mid-flight, the MCP surface became a real command center, and the agent can now drive the desktop's own browser. This release rolls up everything from the v0.20.1–v0.20.6 infrastructure patch tags — those windows are fully documented here
Click to expand...
oof, come on Nous who is doing your marketing, I literally cannot tell what the new feature is from this post. It seems like they've added the ability for the main session to chat with the sub-sessions? That's a good feature - it prevents main session context from getting poisoned with failed attempts - but is kind of meh for self hosted deployments since you need a $50K machine to be able to handle more than 8x concurrency.

in any case most of the impressive behavior you are seeing is from the model, not the harness. Hermes is an OK harness for technical work but it is really built for office automation type stuff, where it can aggressively crawl the web, bypass anti-bot defenses, and assemble reports.
 
bwang said:
oof, come on Nous who is doing your marketing, I literally cannot tell what the new feature is from this post. It seems like they've added the ability for the main session to chat with the sub-sessions? That's a good feature - it prevents main session context from getting poisoned with failed attempts - but is kind of meh for self hosted deployments since you need a $50K machine to be able to handle more than 8x concurrency.

in any case most of the impressive behavior you are seeing is from the model, not the harness. Hermes is an OK harness for technical work but it is really built for office automation type stuff, where it can aggressively crawl the web, bypass anti-bot defenses, and assemble reports.
Click to expand...
It's in the full readme, guess I could have pasted all of it instead of the summary header.

Hermes is OK at technical stuff. From scratch Hermes driven - One node was vanilla linux install and it brought it all the way up to parity with another identical(hw) host prior to setting up the kubernetes cluster etc.

1788218289080.png
 
Last edited:
You must log in or register to reply here.
Back
Top