DevMemory AI 🧠

A local AI assistant that helps developers remember and search their own codebase.

Developers work on multiple projects and often forget where they wrote a specific function, how a feature was implemented, or where a certain logic exists. DevMemory AI solves this by creating a searchable memory of your code using embeddings, vector search, and a local LLM.

The project runs locally, meaning your code stays on your machine.

---

Features

• 🔍 Search your codebase using natural language
• 🧠 AI-powered answers based on your own code
• 💾 Local vector memory using ChromaDB
• 🤗 Hugging Face embeddings for semantic search
• 🦙 Local Llama model support through Ollama
• ⚡ Fast terminal-based workflow
• 🔒 Privacy-focused (no code upload required)

---

How it works

DevMemory AI follows a simple RAG (Retrieval-Augmented Generation) pipeline.

Your Code
    |
    ↓
File Scanner
    |
    ↓
Code Parser
    |
    ↓
Hugging Face Embeddings
    |
    ↓
ChromaDB Vector Database
    |
    ↓
Semantic Search
    |
    ↓
Llama Local Model
    |
    ↓
AI Answer

Example

Instead of manually searching:

Where did I create authentication?

You can ask:

devmemory ask "Where is authentication implemented?"

DevMemory searches your stored code and explains the relevant files.

---

Project Structure

DevMemory-AI/

│
├── app/
│   │
│   ├── ai.py              # Handles Llama AI responses
│   ├── cli.py             # Terminal commands
│   ├── embeddings.py      # Creates vector embeddings
│   ├── parser.py          # Reads and processes code files
│   ├── scanner.py         # Finds project files
│   └── vector_store.py    # ChromaDB operations
│
├── vector_db/             # Local vector database (generated)
│
├── main.py                # Application entry point
├── requirements.txt       # Python dependencies
├── .env.example           # Environment variable example
├── .gitignore
├── LICENSE
└── README.md

---

Installation

1. Clone the repository

git clone https://github.com/Himalsed/DevMemory-AI.git

Move into the project:

cd DevMemory-AI

---

2. Create virtual environment

python -m venv .venv

Activate it:

macOS / Linux

source .venv/bin/activate

Windows

.venv\Scripts\activate

---

3. Install dependencies

pip install -r requirements.txt

---

Setup Local AI Model

DevMemory uses Ollama to run Llama locally.

Install Ollama:

https://ollama.com

Download the model:

ollama pull llama3.2

Check:

ollama list

---

Environment Setup

Create a .env file:

cp .env.example .env

Add your Hugging Face token:

HF_TOKEN=your_token_here

The token is optional for public models but recommended for better download limits.

---

Usage

Index a project

Go to the project you want DevMemory to remember.

Example:

python main.py index /path/to/project

DevMemory will:

• scan files
• split code into chunks
• create embeddings
• store them in ChromaDB

---

Ask questions

Example:

python main.py ask "Where is database connection handled?"

Example output:

DevMemory AI:

The database connection is handled in:

database.py

The file creates the database client
and manages connection settings.

---

Technologies Used

Technology	Purpose
Python	Main language
Typer	CLI interface
ChromaDB	Vector database
Hugging Face	Text/code embeddings
Sentence Transformers	Embedding generation
Ollama	Local Llama model
RAG	AI retrieval pipeline

---

Why this project?

Most code search tools depend on cloud services or require uploading your source code.

DevMemory AI focuses on:

• keeping code private
• running locally
• creating personal AI memory for developers

The goal is to make working with large codebases easier.

---

Future Improvements

• Support more programming languages
• Git history integration
• Automatic documentation generation
• Code explanation command
• VS Code extension
• Team shared memory

---

License

This project is licensed under the MIT License.

Feel free to use, modify, and improve it.