🌟 Charisma Studio 🌟

Ask questions about your local folders and internet sources using local LLMs with RAG-powered insights.

Charisma Studio is a desktop client that transforms how you interact with local directories (codebases, research papers, projects) and web sources by combining LLMs like Ollama with Retrieval-Augmented Generation (RAG) and a vector database search. Select any folder, and/or any sources from the web and get AI-powered context about its contents through an intuitive chat interface.

---

🖼️ Screenshots

• Charisma explaining it's own code. --

🚀 Features

Core Capabilities

• Local LLM Integration - Direct connection to Ollama instances
• RAG-Powered Insights - Database vector search for precise context
• Multi-Format Support - Process code, text, Markdown, and PDFs
• Real-Time Streaming - Typewriter-style response delivery

Developer Experience

• Electron Packaging - Cross-platform desktop app
• TypeSafe Architecture - Built with React + TypeScript

---

🛠 Technology Stack

Component	Choice	Why?
Frontend	React + TypeScript
Styling	Tailwind CSS + Headless UI
LLM Core	Ollama + HNSWLib	LangChain
Packaging	Electron + Vite	Cross-platform binaries

---

⚙️ Development

Prerequisites Checklist

• Node.js 18+ (download)

# 1. Clone repository
git clone https://github.com/JoshDiDuca/charisma-ai.git
cd charisma-ai

# 2. Install dependencies
pnpm install

# 3. Start dev application
pnpm dev

📦 Distribution

For all platforms

Note: Check Electron Builder docs for more knowledge

pnpm build

For a specific one

pnpm build --mac
# OR
pnpm build --win
# OR
pnpm build --linux

The built apps will be available on the release folder.

🛣 Roadmap

Updates

• Multiple conversations with storage
• TTS
• Improved RAG data sources
• AI response sources
• Web RAG queries
• Switch to hnswlib from ChromaDB (licensing and more flexible)
• Automatically download Ollama and dependencies
• Full download for models instead of Ollama pulling
• Message attachments
• Flexible UI
• Improved handling of Ollama being installed and exe files not present in release.
• Fix large pdf files/text files not embedding
• Web RAG query directly reading file urls like .pdf etc
• Next prompt suggestions
• Reasoning model "thinking" support
• Hugging face
• Ability to add custom models
• Relational database queries
• Revisit TTS and fix that
• Translate results
• LlamaIndex full integration
• JSON "Tools/Agents"
• Settings
• Dark theme
• Ignore Paths, Files, Source Settings
• Coding mode setting
• TTS voice model selection
• Source viewing/editing along with the ability to read using TTS
• CSV file support
• Agents/Pipelines/Custom Reasoning
• Agent tasks such as creating, editing, writing files

Future Vision

• Easy access to run any AI locally
• AI Pair Programmer - Refactor code via chat
• Multi Source Data Analysis - Images/PDFs/Web Sources/Database support
• AI Collaboration - Shared session histories

---

🤝 Contributing

We welcome contributions! Please follow our contribution guidelines:

1. Fork the repository
2. Create feature branch (git checkout -b feat/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push branch (git push origin feat/amazing-feature)
5. Open Pull Request

---

📜 License

*Ollama (MIT), Piper (MIT) and hnswlib/hnswlib-node (Apache) have their own licenses - please review separately.