Concepts¶
Understanding the core ideas behind rp1 helps you get the most out of its workflows. These concept guides explain the "why" behind rp1's design decisions.
-
Constitutional Prompting
Learn how rp1 encodes expert knowledge into AI-executable workflows for consistent, single-pass execution.
-
Knowledge-Aware Agents
Discover how agents understand your codebase before writing code, respecting your patterns and architecture.
-
Command-Agent Pattern
Understand the architecture where thin wrapper commands delegate to autonomous agents.
-
Map-Reduce Workflows
See how rp1 parallelizes work for knowledge base generation and PR reviews.
-
Stateless Agents
Learn how resumable interview workflows use file-based state for robustness.
-
Parallel Worktrees
Run multiple tasks simultaneously using git worktrees for isolated execution.
-
Builder-Reviewer Agents
Learn how adversarial cooperation between builder and reviewer agents ensures reliable code implementation.
-
Eval System
Understand how content-addressable attestation ensures prompt changes are tested and traceable.
Quick Overview¶
| Concept | What It Is | Why It Matters |
|---|---|---|
| Constitutional Prompting | Expert knowledge encoded in prompts | No iteration loops, consistent output |
| Knowledge-Aware Agents | Agents that understand your codebase | Output fits your architecture |
| Command-Agent Pattern | Commands delegate to agents | Clean, reusable workflows |
| Map-Reduce Workflows | Parallel processing pattern | Fast KB and PR analysis |
| Stateless Agents | Resumable interview workflows | Robust, transparent state |
| Parallel Worktrees | Git worktree isolation | Run multiple tasks at once |
| Builder-Reviewer Agents | Adversarial agent cooperation | Higher quality, reliable code |
| Eval System | Content-addressable attestation | Deterministic, traceable releases |
Where to Start¶
New to rp1? Start with Constitutional Prompting to understand rp1's core philosophy.
Want context-aware output? Read Knowledge-Aware Agents to learn about the knowledge base.
Curious about architecture? See Command-Agent Pattern for the technical design.
Need performance? Check Map-Reduce Workflows for parallelization patterns.
Building resumable workflows? See Stateless Agents for interrupt-safe interview patterns.
Running tasks in parallel? Check Parallel Worktrees for isolated task execution.
Want reliable implementations? See Builder-Reviewer Agents for adversarial cooperation.
Building evals for prompts? See Eval System for content-addressable attestation.