DevLog - Personal Code Review Assistant

DevLog is a powerful local-first CLI and TUI tool designed to track your coding activity, analyze your commits using local LLMs (via Ollama), and provide deep code reviews by comparing your work against web-based best practices.

It acts as an automated developer journal that not only records what you changed but helps you understand how to improve it.

🚀 Key Features

• Repository Tracking: Automatically captures every commit you make in tracked repositories using Git hooks.
• AI-Powered Analysis: Uses local LLMs (Ollama) to analyze commits for bugs, code quality, and design patterns.
• Web-Enhanced Reviews: devlog review searches the web for best practices on a specific topic and compares your recent code against them.
• Semantic Search: Search your commit history not just by keyword, but by meaning (e.g., "commits where I fixed memory leaks").
• Interactive TUI: A full-featured terminal user interface for exploring your history, viewing diffs, and running analyses without remembering CLI commands.
• Statistics: Visualize your coding habits, languages used, and activity trends.
• Privacy Focused: All data is stored locally in SQLite. All AI analysis is performed locally using Ollama.

🛠 Prerequisites

1. Python 3.10+
2. Ollama: DevLog relies on Ollama for local AI analysis.
   • Install Ollama.
   • Pull a model (e.g., llama3, mistral, or codellama):

     ollama pull llama3

   • Start the server:

     ollama serve

📦 Installation

1. Clone the repository:

   git clone <your-repo-url>
   cd devlog

2. Set up a virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate

3. Install dependencies:

   pip install -e .

4. Run the setup script (Optional): This script helps create a system-wide devlog alias or symlink.

   ./setup_devlog.sh

📖 Usage Guide

1. Start Tracking

Tell DevLog to watch a repository. This installs a post-commit hook.

devlog track /path/to/your/project

2. Work as Usual

Make changes and commit them using git. DevLog captures the data automatically in the background.

git add .
git commit -m "Refactored login logic"
# DevLog captures this instantly!

3. Analyze Commits

Ask the AI to analyze a specific commit.

# Get the hash from `git log` or `devlog commits`
devlog analyze <commit-hash>

4. Interactive Mode (TUI)

Launch the interactive UI to browse repos, commits, and run analyses visually.

devlog tui

• Navigation: Use mouse or arrow keys.
• Actions: Press a to analyze the selected commit.

5. Deep Code Review

Compare your implementation of a feature against top search results from the web.

devlog review "Python async context managers" --language python

This command:

1. Searches the web for "Python async context managers best practices".
2. Scrapes the content and extracts code examples.
3. Finds your relevant local commits.
4. Uses the LLM to generate a report comparing your code to the industry standards.

6. Search

# Keyword search
devlog find "authentication"

# Semantic search (requires embedding first)
devlog embed  # Run once to generate embeddings
devlog semantic "commits related to database performance"

🧩 Commands Reference

Command	Description
devlog track <path>	Start tracking a repository
devlog untrack <path>	Stop tracking a repository
devlog repos	List tracked repositories
devlog commits	Show recent commits across all repos
devlog analyze <hash>	Analyze a commit with AI
devlog review <topic>	Generate a best-practices review
devlog stats	Show coding statistics
devlog tui	Open the Terminal User Interface
devlog find <query>	Keyword search in commits
devlog semantic <query>	Semantic search (concept-based)
devlog help	Show full help message

⚠️ Current Status & Notes

• Beta: The project is functional but evolving.
• Search Backends: Web search relies on DuckDuckGo (default) or Brave Search. Rate limits may apply.
• Chat Interface: There is an experimental chat interface accessible via devlog-chat (or python -m devlog.cli.chat_tui), but it is currently separate from the main TUI.
• Ollama Dependency: Analysis commands will fail gracefully if Ollama is not running.