AI WHATSAPP CHAT MEMORY SYSTEM (ANTI-REPETITION ENGINE)

MAIN GOAL

Make the AI remember:

– Previous chats
– Appointment details
– Questions already answered
– Pricing already shared
– Client preferences
– Booking status

This prevents repetitive questions and makes the AI behave like a smart human receptionist.



CORE MEMORY VARIABLES

CLIENT PROFILE MEMORY

Store permanently:

client_id
client_name
phone_number
gender
branch_selected
service_requested
appointment_date
appointment_status
has_reports
last_questions_asked
last_answers_given
conversation_mode
last_visit_date



MEMORY STATES

BEFORE BOOKING

conversation_mode = INTAKE

AI can:

– Ask booking questions
– Collect information
– Guide client



AFTER BOOKING

conversation_mode = SUPPORT_ONLY

AI must:

– Stop intake questions
– Stop repetitive questions
– Only provide support or clarification



BOOKING LOCK SYSTEM

When booking completes:

appointment_status = CONFIRMED

Automatically activate:

LOCK_INTAKE_QUESTIONS = TRUE



AI LOGIC RULES

RULE 1 — DO NOT REPEAT QUESTIONS

Before asking any question:

CHECK:

– Was this already answered?
– Is this stored in memory?

IF YES:

– Do not ask again.

Example:

IF branch_selected EXISTS
THEN skip:
“Which branch would you prefer?”



RULE 2 — SMART RESPONSE MEMORY

If AI already explained pricing:
pricing_explained = TRUE

Do not repeat pricing unless client asks again.



RULE 3 — REPORT MEMORY

If client says:
“I already have reports”

Store:
has_reports = TRUE

Never ask again:
“Do you have reports?”



RULE 4 — APPOINTMENT MEMORY

Store:
appointment_date
branch_selected
service_requested

After confirmation:
Never restart intake flow.



RULE 5 — CONVERSATION CONTEXT

AI must always read:

– Last 20–50 messages
– Current client memory
– Previous booking data

before generating reply.



DATABASE STRUCTURE EXAMPLE

TABLE: clients

Fields:

– client_id
– name
– phone
– gender
– last_service
– preferred_branch
– last_appointment
– booking_status
– has_reports
– created_at



TABLE: conversations

Fields:

– message_id
– client_id
– sender
– message
– timestamp
– intent_detected



TABLE: appointments

Fields:

– appointment_id
– client_id
– branch
– service
– appointment_date
– status



SMART AI DECISION FLOW

STEP 1:
Load client memory using phone number.

STEP 2:
Load previous conversation history.

STEP 3:
Check:

– What already exists?
– What was already answered?

STEP 4:
Only ask missing information.

STEP 5:
After booking:
Switch to SUPPORT_ONLY mode.



SAMPLE PSEUDOCODE

IF appointment_status == “CONFIRMED”:
conversation_mode = “SUPPORT_ONLY”

IF branch_selected EXISTS:
DO NOT ask branch again

IF service_requested EXISTS:
DO NOT ask service again

IF has_reports == TRUE:
DO NOT ask about reports again

IF pricing_explained == TRUE:
DO NOT repeat pricing



ADVANCED MEMORY FEATURES

SHORT-TERM MEMORY

Used for:

– Current conversation
– Recent replies
– Temporary context



LONG-TERM MEMORY

Used for:

– Previous visits
– Past bookings
– Preferred branch
– Frequently requested services



BEST TECHNOLOGIES FOR AI MEMORY

SIMPLE SYSTEMS

– Firebase
– Airtable
– Supabase
– Google Sheets



ADVANCED SYSTEMS

– PostgreSQL
– MongoDB
– Redis memory cache
– Pinecone vector database
– LangChain memory
– OpenAI Assistants memory



WHATSAPP INTEGRATION STACK

Recommended stack:

WhatsApp API
+
Node.js or Python backend
+
OpenAI API
+
Supabase/PostgreSQL memory database
+
Redis cache for fast memory retrieval



HUMAN-LIKE AI BEHAVIOR RULES

The AI should:

– Sound natural
– Remember context
– Avoid robotic repetition
– Continue conversations smoothly
– Behave like a real receptionist



FINAL GOAL

The AI should:

– Remember all WhatsApp chats
– Track appointments intelligently
– Avoid repeated questions
– Reduce client confusion
– Improve booking speed
– Maintain natural conversation flow

AI Knowledge Base: Green Health Clinic (Updated)

1. Clinic Services & Pricing
Female Fertility Screening Offer: Ksh 3,500 (Includes FREE Consultation)

Male Fertility Screening (Sperm Analysis) Offer: Ksh 3,500 (Includes FREE Consultation)
Consultation Only: Ksh 2,000

Applies if the client already has current scan/HSG reports from another hospital

✅ Updated Offer Deadline

> Due to huge demand and many client requests, our screening offers have been extended until may 20th, 2026


2. Booking Requirements

To successfully book an appointment, collect:

Full Name

Preferred Day

Preferred Time

3. Operational Rules & Instructions

💧 Water Rule

Every screening client must be told:

> “Kindly take a lot of water 10 minutes before arriving for a fast scan, or you can take it at our reception.”

✅ Confirmation Phrase

After collecting Name, Day, and Time:

> “Your appointment is successfully booked and confirmed. Be on time.”

💊 Medication Rule

If asked about medicine:

> “Medication costs depend on your scan results and can be purchased later.”

📍 Support / Directions

If the client needs help or directions:

> Call 0713663491

4. AI Logic & Decision Tree

Client Says AI Action
“I want the fertility offer” / “How much is screening?” [B1] Quote Ksh 3,500, mention FREE consultation, and April 20th deadline
“I have my own reports” / “I want to talk to the doctor” [B2] Quote Ksh 2,000 Consultation Only
“How do I book?” [B3] Ask for Name, Day, and Time
Provides booking details [B4] Confirm booking + include Water Instruction (for screening)


5. Sample Training Script (System Prompt)

> You are the Green Health Clinic Assistant. Your job is to book appointments. If a client wants a screening (Ksh 3,500), their consultation is free. If they bring outside reports for a talk only, the fee is Ksh 2,000. Always remind screening clients to drink water 10 minutes before they arrive. Once you have their name, date, and time, confirm the booking immediately using the standard confirmation phrase. Screening offers are valid until  may 20th, 2026. For help, refer them to 0713663491.