Built and tested - An embeddable AI assistant for parenting resources
Making a parenting content library accessible through grounded conversation. Built and tested.
- Client
- Children's Services Council
- Year
- Service
- Knowledge retrieval, Public-facing AI, Search modernization, Web integration
- Engagement
- Built and tested
The Challenge
EveryParent needed an AI Chat retrieval foundation that could use its content library and fit into an existing public website.
The Approach
BeeNex built a Firebase and Vertex AI retrieval foundation for AI Chat, with custom parsing, synchronization and QA. The website experience uses an isolated web component and a lazy-loaded chat iframe, with documented localization and origin-validation controls.
Capabilities
- Firebase/Vertex retrieval architecture
- Custom parser
- Content synchronization
- Retriever testing and debugging
- Five widget design options
- Zero-dependency web component
- Shadow DOM isolation
- Lazy-loaded Flutter iframe
- Origin validation
- English/Spanish/Haitian Creole localization
Workflow
- Ingest and parse content
- Synchronize changes
- Retrieve relevant sources
- Generate chat response
- Embed isolated widget
- Validate website origin
- Present localized experience
- Firebase
- Firestore
- Vertex AI Search
- Cloud Functions
- Flutter Web
- Web Components
- Shadow DOM
- Sitefinity 15.4
Engagement Outcome
The project reached external user testing and selection of a final widget design. The record documents the build and testing stage, with public rollout still to be completed.
Scope in the record
- Documented interface languages
- 3