Build Influence

build-influence is a Python command-line tool designed to analyze local code repositories, conduct AI-powered interviews about the codebase, and generate social media content (like blog posts or tweets) based on the analysis and interview insights.

Features

• Repository Analysis: Scans local code repositories to understand structure, languages, key files, and potential points of interest.
• AI-Powered Interviews: Uses an LLM to ask targeted questions about the analyzed repository to gather deeper insights and narrative angles.
• Content Generation: Creates content drafts tailored for different platforms (Markdown, Dev.to, Twitter, LinkedIn) based on the analysis and interview data.
• Content Publication: (Experimental/Future) Publishes generated content directly to configured platforms.
• Configurable: Uses a config.yaml file and environment variables for flexible configuration of LLMs, output directories, and platform credentials.

Installation

1. Clone the repository:

   git clone <your-repo-url>
   cd build-influence

2. Create a virtual environment (recommended):

   python3.12 -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install dependencies:

   pip install -r requirements.txt

Configuration

Configuration is handled via config.yaml and environment variables (which take precedence).

1. Create config.yaml: Copy config.example.yaml (if it exists) or create a new config.yaml file in the project root.

2. Review src/build_influence/config/loader.py: This file shows all configurable options and their corresponding environment variables.

3. Set Environment Variables: For sensitive data like API keys, it's recommended to use environment variables (e.g., by creating a .env file). Key variables include:

   • LLM_API_KEY: Your primary LLM provider API key (e.g., Anthropic).
   • LLM_MODEL: The specific LLM model to use (defaults to claude-3.5-sonnet-20240620).
   • DEVTO_API_KEY: API key for Dev.to publication.
   • TWITTER_API_KEY, TWITTER_API_SECRET, TWITTER_ACCESS_TOKEN, TWITTER_ACCESS_TOKEN_SECRET: Credentials for Twitter/X publication.
   • LINKEDIN_CLIENT_ID, LINKEDIN_CLIENT_SECRET: Credentials for LinkedIn publication.
   • LOG_LEVEL, LOG_FILE: Logging configuration.
   • OUTPUT_DIR_BASE: Base directory for output files (defaults to ./output).

   Load environment variables by creating a .env file in the project root:

   OPENAI_API_KEY=your_openai_key
   DEVTO_API_KEY=your_devto_key
   # ... other variables

Usage

The tool is run via the build-influence command (or python -m build_influence.cli if installed differently).

build-influence --help

Common Workflow:

1. Analyze a Repository:

   build-influence analyze path/to/your/repo
   # Example: analyze the current directory
   # build-influence analyze .

   This creates an analysis JSON file in the configured analysis output directory (e.g., output/analysis/your_repo_name_analysis.json).

2. Conduct an Interview:

   build-influence interview your_repo_name

   This uses the analysis file generated in step 1 to ask questions. The interview log is saved (e.g., output/interviews/your_repo_name_interview.json).

3. Generate Content:

   # Generate for all platforms
   build-influence generate your_repo_name
   
   # Generate only for a specific platform (e.g., markdown)
   build-influence generate your_repo_name --platform markdown
   
   # Generate a specific content type (if implemented by the generator)
   # build-influence generate your_repo_name --platform devto announcement

   Content files are saved in the configured content output directory (e.g., output/content/your_repo_name/).

4. Publish Content (Use with caution):

   build-influence publish path/to/generated/content.md --platform devto

   This attempts to publish the specified content file to the target platform using configured credentials.

Other Commands:

• build-influence configure: (Currently placeholder) Intended for interactive configuration.
• build-influence logs: View application logs.

  build-influence logs --lines 50 # Show last 50 lines
  build-influence logs --follow   # Tail logs

Development

• Code Style: Follows Black formatting and isort for imports.
• Testing: Uses pytest. Run tests with pytest.
• Dependencies: Managed in requirements.txt.