simba

High-Efficiency Customer Service Assistant

## See how easy it is to integrate Simba into your website

## What is Simba? Simba is an open-source customer service assistant built for teams who need **full control** over their AI. Unlike black-box solutions, Simba is designed from the ground up around **evaluation** and **customization**, so you can measure performance, iterate fast, and tailor the assistant to your exact needs. ## Why Simba? | Problem | Simba's Solution | |---------|------------------| | Can't measure AI quality | Built-in evaluation framework with retrieval and generation metrics | | Generic responses | Fully customizable RAG pipeline with your own data | | Hard to integrate | Drop-in npm package for instant website integration | | Vendor lock-in | Open-source, self-hosted, swap any component | ## Key Features - **Evaluation-First Design** - Track retrieval accuracy, generation quality, and latency out of the box. Know exactly how your assistant performs. - **Fully Customizable** - Swap embedding models, LLMs, vector stores, chunking strategies, and rerankers. Your pipeline, your rules. - **npm Package for Easy Integration** - Add a customer service chat widget to your website with a single npm install. - **Modern Dashboard** - Manage documents, monitor conversations, and analyze performance from a clean UI. - **Production-Ready** - Streaming responses, async processing, and scalable architecture. ## Quick Start ### Docker (Recommended) The fastest way to get Simba running: ```bash git clone https://github.com/GitHamza0206/simba.git cd simba ``` Create a `.env` file: ```bash OPENAI_API_KEY=your_openai_api_key ``` Run with Docker: ```bash # CPU DEVICE=cpu make build && make up # NVIDIA GPU DEVICE=cuda make build && make up ``` Visit `http://localhost:3000` to access the dashboard. ### Manual Installation If you prefer installing without Docker: ```bash pip install simba-core ``` ```bash simba server simba front ``` ### Development Setup with Claude Code If you're using [Claude Code](https://claude.ai/code), you can set up the project with a single command: ```bash /setup --all ``` This will automatically install all dependencies (Python, frontend, npm package) and start the infrastructure services. Other options: ```bash /setup --backend # Python dependencies only /setup --frontend # Next.js + simba-chat only /setup --services # Start Docker infrastructure only ``` ## Website Integration Add Simba to your website with the npm package: ```bash npm install simba-chat-widget ``` ```jsx import { SimbaChat } from 'simba-chat-widget'; function App() { return ( ); } ``` That's it. Your customers now have an AI assistant powered by your knowledge base. ## Evaluation & Metrics Simba tracks what matters: - **Retrieval Metrics** - Precision, recall, relevance scores - **Generation Metrics** - Faithfulness, answer relevancy, latency - **Conversation Analytics** - User satisfaction, resolution rates Use these metrics to continuously improve your assistant's performance. ## Architecture ``` … ``` ## Docker Deployment ```bash # CPU DEVICE=cpu make build && make up # NVIDIA GPU DEVICE=cuda make build && make up ``` ## Customization Options | Component | Options | |-----------|---------| | Vector Store | Qdrant, FAISS, Chroma | | Embeddings | OpenAI, HuggingFace, Cohere | | LLM | OpenAI, Anthropic, Local models | | Reranker | Cohere, ColBERT, Cross-encoder | | Parser | Docling, Unstructured, PyMuPDF | ## Roadmap - [x] Core evaluation framework - [x] npm chat widget - [x] Streaming responses - [ ] Multi-tenant support - [ ] Advanced analytics dashboard - [ ] Webhook integrations - [ ] Fine-tuning pipeline ## Contributing We welcome contributions! Fork the repo, create a branch, and submit a PR. ## Support - Open an issue on [GitHub](https://github.com/GitHamza0206/simba/issues) - Contact: [[email protected]](mailto:[email protected])