Videos about using Hermes Agent
Curated by: Tonbi's AI Garage (35 videos)
A fleet of agents that keeps an LLM knowledge base current and clean — scout, judge, research, ingest, lint, commit — built live from my open-source template. Sign up for my FREE weekly newsletter, where I spill my unfiltered thoughts on the latest AI news, cool research, and projects I'm building: https://www.onchainaigarage.com/ 🐦 Follow Tonbi on X for real-time AI x blockchain updates! https://x.com/tonbistudio A follow-up to my multi-agent Kanban workflow video, this time with a use case almost anyone can reuse: automatically maintaining an LLM wiki / knowledge base (Karpathy-style). I build it live from my open-source Hermes Multi-Agent Workflow template, so you see exactly what changes from the clone — a new triage rubric, a deterministic verification linter and a thin git helper, and four sub-agent skills. The fleet: a scout watches for new Hermes releases, transcripts, and docs; the orchestrator scores each find against a rubric (novelty, source confidence, scope, relevance, clarity gain — 65+ to ingest); two researchers verify it in the repo and find affected pages in parallel; an ingestor writes the wiki pages; and a linter checks format, links, and freshness before a deterministic commit. There are two human gates in Telegram — one to approve an ingest, one to approve any prune (deleting knowledge), and a conflict route for when new info contradicts an existing page. I also swap in a mix of models to prove the workflow is model-agnostic: Grok for the scout, NVIDIA Nemotron 3 Ultra (free) for the researcher/linter, MiniMax M3 for the ingestor, GPT-5.5 as the orchestrator — and the open/free models held up with no errors. Along the way the system dedupes a release I'd already added manually, and pauses one ingestor to avoid a concurrent write on a shared page. Everything runs on a branch-per-ingest git repo, so it's fully reversible and the raw sources are never touched. Resources: 🔗 Hermes Multi-Agent Workflow template: https://github.com/tonbistudio/hermes-multi-agent-workflow 🔗 Tonbi Studio LLM Wiki template: https://github.com/tonbistudio/llm-wiki Timestamps: 0:00 - Recap + a more general workflow: maintaining an LLM wiki 1:48 - Why a knowledge base, and why it's painful to maintain 4:15 - The workflow: scouts, researchers, ingestor, linter 7:53 - Scoring rubric, new/update/conflict routing, human gates 9:25 - Demo: building from the template in Claude Code 15:45 - Swapping in open models (Grok, Nemotron, MiniMax M3) 18:08 - Handing the runbook to Hermes to build the profiles 21:52 - Live run: scout sweep, researchers fan out, dedup 30:19 - Ingest, concurrent-write pause, lint, and commit Do you keep an LLM knowledge base / wiki, and what would you point a workflow like this at? Drop it in the comments — and if you've got a better name than "Cyberbrain" for a hosted-knowledge-base service, I'm all ears. I'll be doing a lot more in the multi-agent workflow space, so please like, subscribe, and hit the bell! 🦐✨ #HermesAgent #MultiAgent #AIAgents #KnowledgeBase #LLM #Kanban #AgenticWorkflows #Automation #OpenSource