🖥️ @readme-gen/cli

The local-first AI documentation engine for your terminal.

📖 Overview

The readmegen CLI (also aliased as devcon) is a powerful tool for analyzing your local codebase and generating README files without ever leaving your terminal. It leverages the shared @readme-gen/analyzer package to extract project grounding evidence and provide state-of-the-art documentation generation.

⚙️ Core Commands

🚦 Quick Start

# Initialize your API keys and default provider
readmegen init

# Generate a basic README with default options
readmegen generate

# Preview the last generated README in your terminal
readmegen preview

🛠️ Config Management

# View your current CLI configuration
readmegen config view

# Update a specific provider's API key
readmegen config set-key <key> --provider groq

# Set the default model for all generations
readmegen config set-model <model-id>

# Reset all CLI settings to default
readmegen config reset

🏗️ Generation Flows

The CLI intelligently chooses between two distinct generation paths:

1. Semantic Local Flow (Default)

The Semantic Local Flow uses the analyzer package to build a structured project understanding locally. It handles:

• Rewrite: Improving an existing README while preserving structure.
• Append: Adding new grounded sections derived from code evidence.
• Dynamic Context: Injecting extra business context into the generation.

2. Backend / Template Flow

The Backend Flow delegates generation to the API for more complex tasks:

• Monorepo Templates: Using pre-defined layouts for complex projects.
• Nested READMEs: Generating and saving multiple README files for sub-directories in a single pass.
• Model Overrides: Forcing specific platform-hosted models.

💡 Advanced Usage

Skipping Interactive Prompts

If you provide a flag, the CLI is smart enough to skip that specific interactive prompt.

# This will skip the "tone" Selection and use "minimal" automatically
readmegen generate --tone minimal

Full Automated Generation

For CI/CD or non-interactive environments, use the --yes flag:

readmegen generate \
  --tone professional \
  --persona "Senior Developer" \
  --sections Installation Usage Features \
  --yes

Pacing LLM Requests

If you are working on a large repository and hitting Rate Limits (TPM), use the --llm-delay-ms flag:

readmegen generate --llm-delay-ms 30000

🛠️ Development

From the workspace root:

pnpm --filter @readme-gen/cli build
pnpm --filter @readme-gen/cli start