1. Prepare sources
Documents are collected, cleaned and classified before ingestion. Public, internal and restricted material must stay visibly separated.
Method
How Dr Albert works
Dr Albert uses retrieval-augmented generation: documents are prepared, searched locally, then used to generate a cited answer for human review.
2. Retrieve evidence
The private RAG stack searches approved document chunks. The public website never sends visitor questions to that stack.
3. Generate answer
A local model can draft an answer from retrieved evidence. It should show citations and declare uncertainty.
4. Review and publish
Any public example must be reviewed, redacted if needed, and published as static content only.
Limits
Cannot answer everything
Dr Albert answers from its approved corpus. If the source is not present, the honest answer is that the system does not have enough information.
Human review remains required
Outputs from large language models can be incomplete or wrong. Public-sector use needs human judgement and source checking.
Not real-time
V1 public examples are based on reviewed source dates. They do not show current traffic, weather, service status or private operational data.
Not personal advice
Dr Albert does not provide legal, medical, financial or personal council-service advice on this public site.