Skip to content

Latest commit

 

History

History
121 lines (88 loc) · 2.19 KB

File metadata and controls

121 lines (88 loc) · 2.19 KB

Getting Started

Prerequisites

  • Claude Code CLI installed
  • Plugin support available

Installation

# Clone the repo
git clone https://github.com/code60-AI/repo-knowledge.git
cd repo-knowledge

# Launch Claude Code with plugin
claude --plugin-dir .

The SessionStart hook auto-creates the data directory on first launch:

mkdir -p ~/.repo-knowledge/_repos
touch -a ~/.repo-knowledge/_registry.md

Step 1: Index Your First Repo

/repo-knowledge:rk-create https://github.com/org/my-project.git

Expected output:

Cloning into 'my-project'...
[Indexer Agent] Scanning 42 files...
[Indexer Agent] Generated 128 docs
Knowledge base created: my-project (128 docs)

SSH format also works:

/repo-knowledge:rk-create git@github.com:org/my-project.git

Note: SSH format requires an SSH key configured on this machine and added to your GitHub account.


Step 2: Your First Search

/repo-knowledge:rk-search how does authentication work

Or in Chinese:

/repo-knowledge:rk-search 认证逻辑在哪里

Expected output:

Searching my-project...
[Layer 2 semantic match] → authenticate-user.md

## authenticateUser
**File:** src/auth/authenticate.ts
...

Step 3: Update After New Commits

When the repo has new commits:

/repo-knowledge:rk-update my-project

Expected output:

Pulling latest changes...
Changed files: 3
  Modified: src/auth/authenticate.ts (2 docs regenerated)
  Added: src/utils/hash.ts (1 new doc)
  Deleted: src/auth/legacy.ts (2 docs removed)
Updated: my-project (129 docs, last commit: a1b2c3d)

Step 4: Manage Knowledge Bases

List all:

/repo-knowledge:rk-list

Expected:

| Project    | Git URL                              | Last Update | Docs |
|------------|--------------------------------------|-------------|------|
| my-project | https://github.com/org/my-project... | 2026-04-14  | 129  |

Delete:

/repo-knowledge:rk-delete my-project

System asks for confirmation before deleting.


Tips

  • Search supports any language, including mixed CN+EN
  • After first search, similar queries find it faster
  • Large repos (1000+ files) may take a few minutes to index