simba
An open-source customer service assistant with built-in RAG evaluation and an npm chat widget.
Build, customize, and evaluate your own production-ready customer service assistant.
Simba is an open-source customer service assistant designed for teams who need full control over their RAG pipeline and evaluation metrics. It suits developers and support teams who want to avoid black-box AI vendors and measure retrieval accuracy directly.
What it is
It combines a FastAPI backend, vector store ingestion via Celery, and a modern dashboard with a drop-in npm chat widget. The built-in evaluation framework tracks retrieval precision, generation faithfulness, and latency.
Who it's for
It suits engineering teams building custom AI support agents who want to swap embedding models and vector stores freely. It does not suit teams looking for a zero-setup hosted SaaS product with no technical management.
Key features
Built-in Evaluation
Tracks retrieval accuracy, generation quality, and latency out of the box.
Drop-in npm Widget
Add a customer service chat widget to your website with a single npm install.
Modular RAG Pipeline
Swap out embedding models, LLMs, vector stores, chunking strategies, and rerankers.
Modern Dashboard
Manage documents, monitor conversations, and analyze performance from a clean UI.
Up and running in 3 steps
Our honest verdict
What's great
- Includes built-in evaluation metrics for retrieval and generation quality
- Easy drop-in integration via the simba-chat-widget npm package
- Modular architecture lets you swap vector stores, LLMs, and embedding models
- Supports rapid deployment via Docker for CPU or NVIDIA GPU
Where it falls short
- Requires managing self-hosted infrastructure including Redis and a vector store
- Needs API keys and ongoing paid calls to LLM providers like OpenAI
- Setup requires familiarity with Docker, environment variables, and backend services
Pricing & license
Simba is free and open-source under its license. The only cost is running it yourself and paying for underlying LLM and embedding API providers.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
What LLMs and vector stores does Simba support?
Simba supports OpenAI, Anthropic, and local models for LLMs, and Qdrant, FAISS, or Chroma for vector stores.
How do I add the chat widget to my website?
You can install the simba-chat-widget package via npm and drop the React component directly into your application.
Does Simba evaluate retrieval performance?
Yes, it includes built-in evaluation metrics covering retrieval precision, recall, relevance scores, and generation faithfulness.
