> **Licensing:** this add-on is a thin wrapper licensed under **Apache 2.0**. It contains no code of its own and simply pulls in the **AI Support AppJar** modules, which are **commercial software** distributed under the [AppJars Software License Agreement](https://www.appjars.com/legal/license/v1/). You can use AI Support for free in **free mode** (up to 50 messages exchanged per day, manual and automatic combined). A license removes all limits. See [Licensing](https://docs.appjars.com/licensing/). **AI Support** embeds AI-powered support assistants in your application, with a chat interface and an administration UI built with Vaadin Flow. Assistants answer from your own knowledge base using retrieval-augmented generation (RAG), administrators can step into any conversation when a person is needed, and every exchange with the model is recorded for inspection. ## Features - **Configurable assistants:** combine a language model (OpenAI-compatible or local Ollama), a prompt and knowledge categories into assistants for different purposes, and choose a default one. - **RAG knowledge base:** upload documents (PDF, TXT, DOC, DOCX, LOG), organize them in categories and let assistants ground their answers through semantic vector search. - **Human handoff and private notes:** users can call in an administrator with a mention, the conversation passes from the assistant to a person and back, and administrators exchange notes the end user never sees. - **Full chat view and floating chat bubble:** users can ask a question from any page without leaving their task, attach files and rate the session when it is done. - **Application tools:** expose selected methods of your code as tools that an assistant can call to answer from live application data, assigned per assistant. - **Messaging channels and LLM Inspector:** connect WhatsApp, Facebook Messenger and Instagram accounts to an assistant, and review every model exchange with its retrieved passages, tool calls and token usage. ## Installation If your project doesn't declare it yet, add the Vaadin add-ons repository: ```xml vaadin-addons https://maven.vaadin.com/vaadin-addons ``` Add the dependency. The wrapper declares the AppJars repository (`https://maven.appjars.com`), so Maven resolves the modules automatically: ```xml org.vaadin.addons.appjars appjars-ai-support 2.1.0 ``` - **Gradle** ignores repositories declared in dependency POMs, so add `maven { url "https://maven.appjars.com" }` to your repositories. - **Nexus/Artifactory mirroring everything (`mirrorOf *`):** your administrator needs to add a proxy for `https://maven.appjars.com`. Then follow the [Getting started guide](https://docs.appjars.com/ai-support/getting-started/). ## Requirements - Java 21 - Vaadin 25.2 - Spring Boot 4.1 - PostgreSQL with the pgvector extension, used to store and query document embeddings ([details](https://docs.appjars.com/ai-support/getting-started/#configure-application-properties)) - An OpenAI-compatible API or a locally hosted Ollama model - JPA / Hibernate ## Included dependencies | Artifact | License | |---|---| | `com.appjars:appjars-ai-support-flow` | AppJars Software License Agreement v1 (commercial) | | `com.appjars:appjars-ai-support-business-impl` | AppJars Software License Agreement v1 (commercial) | | `com.appjars:appjars-ai-support-data-impl` | AppJars Software License Agreement v1 (commercial) | These modules transitively bring in the rest of the AI Support modules and `appjars-*-utils` (same commercial license), plus open-source libraries such as Vaadin add-ons by Flowing Code (Apache 2.0). ## Links - [Product page](https://www.appjars.com/catalog/ai-support/) - [Documentation](https://docs.appjars.com/ai-support/overview/) - [Demo application](https://github.com/AppJars/ai-support-demo) - [Licensing and free mode](https://docs.appjars.com/licensing/) AppJars are built and maintained by [Flowing Code](https://www.flowingcode.com).