Skip to main content

Getting Started

This guide walks you through installing LibriScribe, configuring model defaults, and launching the CLI.

Prerequisites​

  • Python 3.10 or later: Check with python --version.
  • pip: Usually included with Python.
  • At least one LLM API key from one of these providers:

Installation Steps​

  1. Clone the repository

    git clone https://github.com/guerra2fernando/libriscribe.git
    cd libriscribe

  2. Install LibriScribe

    python -m pip install "libriscribe[openai]"

Use the anthropic, google, or bedrock extra for those providers, or all-providers to install every provider SDK. OpenAI, OpenRouter, and Bedrock Mantle use the openai extra. To work from a cloned source checkout, use python -m pip install -e ".[openai]" instead. Provider SDKs are optional; the base package supports local keyword retrieval and MCP access without them.

Optional local semantic search can be installed with python -m pip install "libriscribe[semantic]" (combine extras with a comma, for example libriscribe[openai,semantic]). It runs the sentence-transformers model on your device; no embedding API key or hosted vector service is used. Keyword search does not need this extra. Semantic mode uses local-only model loading by default, so explicitly download or place the model in the local Hugging Face cache or provide a local model path before rebuilding a semantic index.

  1. Create your .env file

    Start from the provided example:

    cp .env.example .env

  2. Add API keys and optional model defaults

    Example:

    OPENAI_API_KEY=your_openai_key_here
    OPENAI_MODEL=gpt-4o-mini

    GOOGLE_AI_STUDIO_API_KEY=your_google_key_here
    GOOGLE_AI_STUDIO_MODEL=gemini-2.5-flash

    CLAUDE_API_KEY=your_claude_key_here
    CLAUDE_MODEL=claude-3-opus-20240229

    DEEPSEEK_API_KEY=your_deepseek_key_here
    DEEPSEEK_MODEL=deepseek-coder-6.7b-instruct

    MISTRAL_API_KEY=your_mistral_key_here
    MISTRAL_MODEL=mistral-medium-latest

    OPENROUTER_API_KEY=your_openrouter_key_here
    OPENROUTER_BASE_URL=https://openrouter.ai/api/v1
    OPENROUTER_MODEL=anthropic/claude-3-haiku

    # AWS Bedrock (explicit keys optional — falls back to instance profile / env AWS_* vars)
    BEDROCK_REGION=us-east-1
    BEDROCK_MODEL=us.anthropic.claude-sonnet-4-6
    BEDROCK_ACCESS_KEY_ID=your_access_key_id
    BEDROCK_SECRET_ACCESS_KEY=your_secret_access_key

    # Bedrock Mantle
    BEDROCK_MANTLE_API_KEY=your_api_key_here
    BEDROCK_MANTLE_BASE_URL=https://bedrock-mantle.us-east-1.api.aws/v1
    BEDROCK_MANTLE_MODEL=qwen.qwen3-coder-next

    LibriScribe uses these model values as provider defaults:

    • Simple Guided Setup uses the selected provider's default model automatically.
    • Advanced Guided Setup lets you keep the .env default or enter a single custom model ID for the project.
    • Expert mode can override both the provider default and the project default through config files.

Launching LibriScribe​

Start the interactive CLI with:

libriscribe start

LibriScribe currently offers three setup flows:

  • Simple Guided Setup
  • Advanced Guided Setup
  • Expert: Configuration File

You can also jump directly into Expert mode:

libriscribe start --config examples/expert-config.yaml

Verifying the Installation​

After installation, you can verify the CLI is available with:

libriscribe start --help