Create a workspace
Register an expert workspace, set up the organization, and create the first Assistant.
How it works
From curated knowledge files to a client-facing AI Assistant
SonaMinds is designed around a simple loop: create a workspace, import selected knowledge, configure an Assistant, test grounded answers, then share access with clients, students, members, or website visitors.
Workflow
Six steps from expert knowledge to usable access
Import curated knowledge
Upload focused text, Markdown, and DOCX sources such as articles, course notes, FAQs, templates, or onboarding guides.
Configure the Assistant
Set the Assistant name, introduction, answer style, knowledge scope, disclaimers, and unsupported topics.
Test grounded answers
Ask representative questions in the expert workbench and check whether responses use the right material and cite useful sources.
Share access
Use invite links, client access passes, public pages, or website widgets to let the right audience ask questions.
Monitor and refine
Review usage, quota, frequent questions, and weak answers. Improve source material instead of letting the assistant drift.
Under the surface
The product value is the structure around the model
A model alone is not enough. SonaMinds combines knowledge selection, semantic processing, answer generation, citations, access rules, quota checks, client passes, usage events, and deletion semantics into one controlled product flow.
Source-grounded semantic processing
SonaMinds is not a traditional information retrieval system; it is an intelligent knowledge processing layer for modern AI. When someone asks a question, the Assistant interprets the selected sources through the expert's knowledge framework, core principles, concept relationships, and service boundaries.
Beyond repeating existing sentences
Even when the source material does not contain a direct sentence-level answer, the Assistant can synthesize a grounded response when the answer follows from the expert's materials. Experts can also ask their own Assistant to test reasoning, review boundaries, and find gaps in the knowledge set.
Access and quota checks
The system checks whether a user, organization, Assistant, and client pass are allowed to ask and whether answer limits remain available.
Knowledge lifecycle
When a document is deleted, it should stop being retrieved. Source storage and indexed chunks are part of the product lifecycle.
Source quality
Curated knowledge beats bulk uploading
SonaMinds works best when the expert uploads high-signal material with clear titles, stable concepts, and enough context.
- Start with the core method, main FAQ, onboarding document, and best public explanations.
- Separate audiences into different Assistants when their questions or boundaries differ.
- Review citations during testing and rewrite source material when answers are vague.
- Keep sensitive customer data out unless a separate written agreement exists.
Ready to build your first Assistant?
Start with one focused use case, curated source material, and a real audience.