mark8 · realtime call analyst
Every support call, read and scored as it happens.
Sukoon Unlimited’s experts talk to elderly members on the phone all day. mark8 reads each call, tags what it was about, keeps a persona for both sides, and scores the conversation out of 100, so the team sees quality without listening to every call.
100–500calls handled a day
all dayrealtime, not batch
12topic families tracked
7scoring parameters
01 — the pipeline
From hang-up to score
02 — what members talk about
Twelve topic families, every call tagged
Each scored call is tagged against the topic families in mark/texts/topics.txt.
Mental Health
01Physical Health
02Personal Wellbeing
03Family
04Work/Retirement
05Community
06Legal
07Life as usual
08General News
09Forward Looking
10Financial Planning
11Interests
1203 — the server around it
One process, Flask and Quart side by side
Starlette mounts a sync Flask app and an async Quart app together. Besides scoring calls, the same server ran Sukoon’s operations: bulk users and schedules, WhatsApp queues, push notifications and PDF invoices.
POST /flask/processscore one call
/bulk_usersimport users
/bulk_schedulebook many calls
/queue_wa_msgswhatsapp queue
/send_fcm_msgspush notifications
/excel_uploadsheet imports
/gather_datareporting
/chatchat service
async /invoicepdf invoices · quart
# mark8 + ARK-I share code as git submodules cd mark8 && git pull git submodule update --init --recursive cd ../ARK-I && git pull git submodule update --init --recursive pm2 restart mark markb ARK