DMIND Project Overview

Evidence snapshot: 2026-08-19. Repository facts use the recorded commit in the open repository map →. Production resources were verified in AWS account 564141170168, region ap-southeast-1, and Atlas project DMIND Production.

Purpose

DMIND is a mental-health prescreening and interview platform. It guides participants through AI-assisted interviews, presents severity guidance and follow-up services, and gives clinical staff a dashboard for reviewing cases and recording outcomes.

The maintainership boundary is organized into three modules:

Module	Primary users	Responsibilities
Open prescreening →	Participants, clinical staff, corporate partners	Participant web journey, staff dashboard, corp-specific behavior, sessions, case records, follow-up routing, content, and reporting.
Open IVR →	Telephone interview participants and calling systems	IVR interview session, answer submission, severe-result lookup, feedback, and AI result retrieval.
Open the AI platform →	Prescreening and IVR backends	Prediction orchestration, media storage/conversion, ASR, depression/suicidal classification, GPT analysis, and result aggregation.

Current product surfaces

Surface	Repository	Runtime relationship
Participant prescreening web	dmind-prescreening-web-frontend	Calls the /interview API of the selected corp deployment.
Staff dashboard	agnos-mental-dashboard-front	Calls https://api.prescreening.dmind.app/dashboard directly in production.
Prescreening API	dmind-prescreening-backend	Owns participant, dashboard, system, corp, and partner flows and persists current state in MongoDB and Redis.
IVR API	dmind-ivr-backend	Owns IVR sessions and delegates prediction work to the AI controller.
AI controller and workers	Multiple dmind-aiservice-* repositories	Fan-out/fan-in processing through SQS, S3, MongoDB, ECS, Lambda, and model-serving components.

Environments and variants

The participant product has a main deployment plus tenant-specific variants. Open prescreening infrastructure → for the frontend, backend, database, and AI bundle used by each current or retained variant.

• Main DMIND (main, dev, and UAT configuration)
• TU and TU UAT
• MDCU
• CPIRD
• THPF
• KKU
• Thangrath
• TSU
• Navy is retained but inactive. Chula and Somdet remain legacy namespaces without a current service identified.

How to use this handover

1. Start with open system architecture → for module and ownership boundaries.
2. Use the open repository map → before changing code; several active components do not contain dmind in their repository name.
3. Read the business, data, and infrastructure pages for the module being changed.
4. Treat resource lists as a dated snapshot and re-run read-only discovery before deployment or operational work.
5. Never copy values from Secrets Manager, Parameter Store secure values, database records, or historical compose files into documentation or tickets.

Document directory

Topic	Documents
Architecture and ownership	Open system architecture →, open the repository map →
Prescreening and dashboard	Open business logic →, open data and integrations →, open infrastructure →
Prescreening flow animation	Source HTML: dmind/other/dmind-data-flow-animation/index.html. It covers the prescreening flow only and does not represent IVR.
IVR	Open business logic →, open data and integrations →, open infrastructure →
AI platform	Open the pipeline →, open data contracts →, open infrastructure →
Product design	Open the UX/UI and Figma directory →

Deprecated and inactive paths retained in code

Path	Current status
Mental Med/Agnos HAMD-7/8Q participant step	Deprecated and removed from the current participant frontend. /interview/submit-self-rating remains a registered external entry point before the interview; no caller was found in the accessible application repositories, so external runtime use is unverified.
AIMET 9Q self-rating	Deprecated and absent from the current participant route tree. Its retained page submits the rating and then navigates to method selection, placing it before the interview in that implementation. The API client, backend route /interview/submit-self-rating-9q, stored fields, and result-source enum remain for compatibility.
CHOICE participant interview	Deprecated and excluded from current configured method selection. Frontend components and backend/AI contracts remain for compatibility.
GET /interview/question	Deprecated and disabled; the route and controller are commented out while response types remain.
agnos-mental-dashboard-back	Historical and not in the current staff-dashboard path. The active frontend uses dmind-prescreening-backend and MongoDB.
Python GPT worker	Retained but not running in production-v2; the Go GPT worker is the current production consumer.
lambda-media-convert	Historical and not used by the current DMIND production path; the active converter is the ECS service.