Digital Me
A text and voice interface grounded in my published professional profile.
- Next.js
- Deepgram
- Conversation
Under the hood
Technical architecture
Answers are grounded in a maintained knowledge base. It cannot access private HR systems or perform actions on a visitor’s behalf.
Conversation
Next.js · browser audio
Collects typed questions or voice input and presents responses.
Grounding
Published profile knowledge base
Supplies experience, skills and contact facts to the server-side prompt.
Response
/api/digital-me · AI adapter
Validates conversation turns and requests a grounded text or short spoken response from configured providers.
Audio
Deepgram speech services
Handles transcription and speech synthesis through server routes, keeping credentials server-side.
Failure and fallback path
If live generation fails, supported questions receive a response assembled from published profile facts and labelled as limited mode. Unknown questions return an availability error and contact route.
One request, end to end
- Submit a typed question or transcribe voice input.
- Combine the bounded conversation with the published profile context.
- Display the response and optionally synthesize speech.
Why it is built this way
- Distinguish profile-based replies from live model responses.
- Bound conversation history and voice-answer length.
- Use the published project catalogue rather than invent project availability.