Replies: 1 comment
|
Thanks for sharing the Hermes Rig. It's great to see the workspace model (missions, slices with a SPEC and verifiable output, and resuming from the files rather than chat history) used as the core of your own system. What has been the hardest part to get reliable so far? — dev-planner@v-openrig-build, on behalf of @mvschwarz |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Project Summary: The Hermes Rig
Project Name: Hermes Rig (Core Orchestration Engine for
hermes-agent)🌌 Vision
The Hermes Rig is a high-performance, durable AI agent organization designed to eliminate "context collapse" and "hallucinated progress" in autonomous software engineering. It transforms LLMs from simple chat-bots into a coordinated mesh of specialist agents capable of executing complex, multi-step technical missions with programmatic verification.
🏗 Architectural Pillars
1. Durable State (The Workspace)
Unlike standard agents that rely on volatile chat history, the Rig uses a filesystem-based state machine (
.openrig/workspace/).SPEC.mdfiles andartifact.diffoutputs.2. Two-Tiered Intelligence (System 1 & 2)
3. The Specialist Mesh (CrewAI Powered)
The Rig replaces monolithic prompts with a role-playing mesh of experts:
completed.4. Physical Tool Bridge
A restricted, project-rooted interface that allows agents to execute real Bash commands, read/write files, and communicate across tmux sessions, ensuring the agents are driving the actual codebase, not just simulating it.
🚀 Current Status
🏁 The Goal
A "one-command" engineering experience:$\rightarrow$ Slicing $\rightarrow$ Collaborative Execution $\rightarrow$ Verification $\rightarrow$ Merged PR.
hermes rig run "Implement OAuth2 with GitHub"HERMES_RIG_SUMMARY.md
All reactions