Three products published by IgnitionAI, three distinct audiences
A RAG platform sold to other consultancies, an Agentic AI course for Node.js, React and Next.js developers, an open-source framework to train reinforcement-learning agents in the browser. Click a product for the detail. A product is only brought into an engagement if your use case genuinely benefits from it.
Commercial platform·Consultancies & AI agencies·Solo 59 € · Team 249 € · Agency-ESN-Dedicated licence
IgnitionRAG
« Ship the document AI your clients expect »
IgnitionRAG is a complete RAG platform, sold as SaaS and under license, to ship document AI without rebuilding the ingestion-search-agent chain every time. It handles multimodal content (PDF, DOCX, images, tables), hybrid vector + BM25 search with reranking, agents with custom tools, and an MCP server. SaaS Solo and Team: platform-managed AI, credits included, no LLM key to provide. Agence, ESN or Dedicated license: BYOK available, no margin on tokens. Hosted on OVH France, GDPR-native.
Solo 59 € / Team 249 € / Agency-ESN-Dedicated licence
SaaS Solo/Team: platform-managed AI, credits included, no LLM key to provide. Agence/ESN/Dedicated license: BYOK available, no margin on tokens
When this product comes into an IgnitionAI engagement
When your project falls within IgnitionRAG's scope (document RAG, agents with tools, multi-tenant for your own clients), production goes from several months to a few weeks. SaaS Solo and Team include credits, with no LLM key to provide. An Agence, ESN or Dedicated license allows BYOK, with no margin on tokens, and then lets you deploy new use cases on your own, without depending on IgnitionAI Agency.
Training program·Node.js, React, Next.js devs·On request · lifetime access
GenAI Labs
« Agentic AI training for Node.js, React and Next.js »
GenAI Labs is a 100% hands-on program for Node.js, React and Next.js developers who want to master building agents in production. The content covers LangGraph (state, transitions, tools, routing, memory), the operational ecosystem (golden-set evaluation, observability, tracing) and production (prompt policies, deployment, monitoring, generative UI). Private Discord, weekly live sessions, code reviews by the experts. Pricing on request, lifetime access.
Key features
Three pillars: LangGraph & Agents, Ecosystem & Ops, Production & Security
Full progression: LLM calls, state & memory, tools & routing, RAG, reasoning
Modules on evaluation (golden sets) and observability (tracing/logs)
Production patterns: prompt policies, deployment, monitoring, generative UI
Private Discord with peer support and code review
Weekly live sessions and guidance from experts
When this product comes into an IgnitionAI engagement
After an IgnitionAI engagement, your JavaScript developers inherit a system in production. GenAI Labs gives them the conceptual framework and patterns to extend it and design new agents without starting over from LangGraph's official documentation. The program explicitly targets Node.js/React/Next.js devs, not Python data scientists: it's the stack most product teams already have in-house.
MIT open-source framework·Creative JS devs · game developers·MIT open source · npm
IgnitionRL
« Train RL agents in your browser »
IgnitionRL is an open-source JavaScript framework to train reinforcement-learning agents directly in the browser. No Python, no server, no GPU cluster: you describe your environment, call env.train(), and deploy via ONNX to Unity (Sentis), Unreal (NNE), Python or edge. Zero-config: input and output dimensions are inferred automatically. Automatic WebGPU > WebGL > WASM > CPU fallback depending on the device. Published on npm as @ignitionai/core, MIT license.
Key features
Three interchangeable algorithms: DQN (value-based), PPO (policy gradient), tabular Q-Learning
Auto-configuration: automatic inference of input and output dimensions
WebGPU acceleration with automatic WebGL > WASM > CPU fallback
Six playable demos: Maze, CartPole 2D/3D, MountainCar, Car Circuit, Drone Navigation
Public npm packages: @ignitionai/core, @ignitionai/backend-tfjs, @ignitionai/environments
When this product comes into an IgnitionAI engagement
Trial-and-error reinforcement learning is relevant for a narrow subset of cases: digital twin of an industrial process, optimising a policy under constraints, simulating agents in a closed environment. When your project falls into that category, the R&D run on IgnitionRL feeds directly into our architecture proposals. For other engagements (RAG, chatbot, agent with tools), the framework remains a public proof of the technical mastery of the team that will audit your systems.
To make your internal documents searchable, connect an agent to SharePoint or your business databases, or deploy augmented search for your clients. When the case falls outside the covered scope, we build it directly rather than forcing the tool.
After delivery, on demand
GenAI Labs
For Node.js, React or Next.js teams that want to extend the delivered system or build other agents on their own. A good complement to the skill transfer contracted at the end of the engagement.
Narrow cases, but then central
IgnitionRL
Digital twin of an industrial process, optimising a policy under constraints, simulating agents in a closed environment. Otherwise the framework remains a public proof of the team's technical mastery.
