Setting Up a Project

This guide covers how to set up a RaiSE project — whether you’re starting from scratch (greenfield) or adding RaiSE to an existing codebase (brownfield).

Greenfield: New Project

Initialize

mkdir my-project && cd my-project
git init
rai init --name my-project

This creates the .raise/ directory:

.raise/
├── manifest.yaml          # Project metadata
└── rai/
    ├── memory/
    │   ├── patterns.jsonl  # Learned patterns
    │   └── index.json      # Unified memory index
    ├── session-state.yaml  # Current session state
    ├── identity/
    │   ├── core.md         # Rai's identity
    │   └── perspective.md  # Rai's perspective
    └── personal/           # Your private data (gitignored)
        ├── sessions/
        └── telemetry/

Set Up Governance

RaiSE creates a governance/ directory with templates:

governance/
├── constitution.md     # Your principles
├── prd.md              # Your requirements
├── guardrails.md       # Your rules
├── backlog.md          # Your work items
└── architecture/
    ├── system-context.md
    └── system-design.md

Fill these in with your project’s specifics:

1. Constitution — What are your non-negotiable principles? “Type annotations on all code.” “Tests before implementation.” Write 5-10 of these.

2. PRD — What are you building? Define 3-5 requirements at the feature level.

3. Guardrails — What rules must your code follow? Use MUST for non-negotiable, SHOULD for recommended. Link each guardrail to a requirement.

You don’t need to fill everything at once. Start with a few principles and guardrails. Add more as you learn what matters for your project.

Set Up Skills

Skills live in .claude/skills/. RaiSE comes with a standard set of lifecycle skills. To see what’s available:

rai skill list

You can create project-specific skills:

rai skill scaffold my-custom-skill --lifecycle utility

Build Memory

After filling governance files, build the memory index:

rai graph build

This creates the unified knowledge graph from all your governance documents, making them queryable and loadable into your AI’s context.

First Session

Open your AI assistant (Claude Code) in the project directory and run:

/rai-session-start

This loads your project’s governance, memory, and skills into context. You’re ready to work. From here, you interact through skills (/rai-story-start, /rai-story-plan, etc.) — they orchestrate the CLI for you.

Brownfield: Existing Project

Initialize with Detection

For existing codebases, use --detect to analyze conventions:

cd existing-project
rai init --detect

This does everything rai init does, plus:

• Scans your source code for patterns
• Identifies coding conventions (naming, formatting, imports)
• Detects testing patterns and frameworks
• Generates guardrails from detected conventions

Review the generated governance/guardrails.md — it’s a starting point, not gospel. Adjust to match your team’s actual standards.

Discovery Scan

For deeper analysis, run the discovery pipeline:

# Scan source code
rai discover scan src/ --language python

# Analyze with confidence scoring
rai discover scan src/ -l python -o json | rai discover analyze

# Check for architectural drift later
rai discover drift

Discovery extracts your codebase’s structure — classes, functions, modules — and builds a component map. This feeds into the knowledge graph, giving your AI partner architectural awareness.

Integrate with Existing Workflow

RaiSE adds structure alongside your existing tools:

• Git: RaiSE uses standard Git branching. Story branches are created from your development branch (dev). Epics are logical containers (directories), not branches.
• CI/CD: RaiSE doesn’t touch your pipeline. Guardrails are enforced at the AI level, not the CI level.
• Editor: Skills are invoked through your AI assistant (e.g., /rai-story-start in Claude Code). No editor plugin needed.

Project Structure Reference

A fully set up RaiSE project looks like:

my-project/
├── .raise/                    # RaiSE runtime
│   ├── manifest.yaml
│   └── rai/
│       ├── memory/            # Shared memory (committed)
│       ├── personal/          # Private data (gitignored)
│       ├── session-state.yaml
│       └── identity/
├── .claude/
│   └── skills/                # Skill definitions
│       ├── session-start/
│       ├── story-start/
│       └── ...
├── governance/                # Project governance
│   ├── constitution.md
│   ├── prd.md
│   ├── guardrails.md
│   └── architecture/
├── work/                      # Work tracking
│   └── epics/
│       └── e01-my-epic/
│           ├── SCOPE.md
│           └── stories/
├── src/                       # Your code
└── tests/                     # Your tests

What to Commit

Directory	Commit?	Why
.raise/rai/memory/	Yes	Shared patterns and calibration
.raise/rai/personal/	No	Developer-specific, gitignored
.raise/manifest.yaml	Yes	Project metadata
governance/	Yes	Shared governance documents
.claude/skills/	Yes	Shared skill definitions
work/epics/	Yes	Work tracking and retrospectives