AI Chatbot Faster Than Custom Builds, More Flexible Than SaaS

About Chatguru

Chatguru is Netguru’s open source AI chatbot platform, built by the internal R&D team and released under the MIT license. Designed as a white label solution, it gives teams a ready made foundation for building their own AI chatbot experiences instead of starting from scratch.

Chatguru provides the backend, conversational logic, and AI infrastructure, while Silk, Netguru’s free design system for commerce products, delivers the frontend interface components. Together, they work like modular building blocks that teams can quickly adapt to their own brand, workflows, and business needs.

Explore GitHub

The challenge

Netguru's teams and clients kept running into the same wall. The AI chatbot market offered two options, and neither fit the real needs of companies trying to build meaningful conversational experiences grounded in real business data.

SaaS tools were fast, but limiting

SaaS chatbot tools deploy quickly, but hit a ceiling fast. Teams could launch in days, but customizing flows, integrating product catalogs or inventory systems, controlling brand experience, or connecting to internal tools often meant workarounds or dead ends. Vendor lock in was the default outcome.

Fully custom builds required major investment

Fully custom AI builds gave full control, but at significant cost. Building a production grade chatbot with RAG capabilities, reliable architecture, observability, and real world testing infrastructure could take months and require sustained engineering investment before teams could focus on actual business value.

Reliability became a real business risk

The reliability problem became especially visible in real client scenarios.

• In healthcare related use cases, responses needed to stay compliant and grounded in verified information rather than hallucinated outputs.
• In commerce and inventory driven environments, chatbot answers had to reflect current product availability and real time business data, not generic model assumptions.

When AI responses influence purchasing decisions, operational workflows, or customer trust, accuracy becomes a business requirement, not a nice to have.

The approach

Netguru's R&D team built Chatguru as a production ready foundation, not a prototype. Every architectural decision was designed to solve the reliability, flexibility, and implementation challenges teams faced with existing chatbot solutions.

RAG as the reliability layer

The biggest challenge in deploying AI chatbots is keeping responses grounded in real and up to date business data. Instead of relying on generic model knowledge, Chatguru retrieves information directly from the client’s own product catalogs, policy documents, or knowledge bases before generating a response. This makes the platform suitable for environments where accuracy matters.

Open source and self hosted by design

Because client data stays within the client’s own infrastructure, Chatguru can be used in regulated environments and industries handling sensitive information. Released under the MIT license, the platform removes vendor dependency and gives teams full ownership and flexibility over the implementation.

Chatguru + Silk as modular building blocks

Rather than delivering only the AI backend, Netguru paired Chatguru with Silk, the company’s open source design system for commerce and marketplace products. Chatguru handles the conversational logic and AI infrastructure, while Silk provides reusable frontend components and interface patterns.

Together, they work like modular building blocks that allow teams to launch branded conversational experiences much faster instead of building both the backend and frontend from scratch.

Built in observability and testing

The platform includes Langfuse for monitoring and tracing, pytest for testing, Promptfoo for LLM evaluation, and RAGAS for RAG quality evaluation. These tools are part of the default setup, helping teams continuously measure and improve chatbot performance over time.

Technology stack built for production

Chatguru uses FastAPI and Uvicorn for the backend, LangChain with Azure OpenAI as the default LLM layer, sqlite-vec for vector search, React 19 and Vite for the frontend, and Docker for containerized deployment. The stack was selected for long term maintainability and real production environments rather than demo use cases.

What Chatguru delivers

Chatguru is a starting point, not a finished product. Its value is what it makes possible for the teams that build on it.

Chatbot for commerce teams: Chatguru supports product discovery through natural language instead of filter-based search, guides customers through buying decisions, automates pre-sales support questions about availability and compatibility, and handles post-purchase service interactions. It works across the full shopping journey, from first browse to repeat purchase.

Chatbot for healthcare organizations: RAG grounds every answer in verified clinical or administrative content. Patient navigation, appointment guidance, FAQ automation, and post-visit support can all run on real documentation rather than general AI knowledge, which is the only basis on which accuracy-critical healthcare interactions should operate.

Chatbot for insurance businesses: Policy discovery, coverage explanation, claims support, and renewal conversations all benefit from the same RAG foundation. When a customer asks what a specific clause covers, the answer comes from the actual policy document, not a language model's approximation of what insurance policies tend to say.

Key results

• 1–2 months from kickoff to production, depending on integration and branding complexity.

• 4–5 people involved in a typical implementation, including an AI engineer, DevOps engineer, QA engineer, and PM, with design support added for custom branded experiences.
• Production ready capabilities out of the box, including RAG, real time WebSocket streaming, observability, rate limiting, server side session persistence, and a full testing setup.
• Open source availability on GitHub, giving teams direct access to the codebase without licensing costs or vendor lock in.
• Reduced frontend development effort through Silk integration and reusable UI components, allowing teams to focus on product specific customization instead of building interfaces from scratch.
• One command deployment through Docker Compose, making it possible to launch the full stack, including the product database, quickly in production environments.