vlozi/docs

Getting started

Introduction

What Contact Intelligence does, how it works with the chatbot, and key concepts.

Contact Intelligence (CI) is the memory and relationship layer behind your Vlozi chatbot. While the chatbot processes individual messages, Contact Intelligence tracks what customers say over time, how they feel, and how strong your relationship with them is growing.

When the bot responds to a customer, it automatically draws on CI to:

  • Know the customer's name, preferences, and past topics — without asking again
  • Adapt its tone and empathy based on the customer's current mood
  • Recognise that today's message is part of a ongoing relationship, not a new conversation

What CI tracks

Concept What it stores Example
Contact Who the customer is (channel + identifier) WhatsApp +91-9876543210
Mood Current emotional state (7 moods + energy level) frustrated, high energy
Memories Things said across all conversations "Has a golden retriever named Bruno"
Entities Named things the customer mentioned pet: Bruno, workplace: Infosys
Relationship stage How deep the relationship is (6 stages) established (20+ messages, active daily)
Churn risk Probability the customer will stop engaging 0.72 (high risk — 14 days inactive)
Outreach triggers Scheduled messages to re-engage or celebrate Inactivity check-in, 7-day streak celebration

How it works with the chatbot

Every message your customer sends flows through this pipeline:

Customer sends message


Chat Engine calls CI: GET /context/:id
        │  Returns mood, memories, relationship state, entities

Bot generates a response that knows who this person is


Chat Engine calls CI: POST /ingest
        │  CI classifies mood, extracts new memories, updates state

Customer receives a personalised, context-aware reply

This happens automatically when your chatbot's memory mode is set to deep (see Bot configuration).

Channels

CI creates one contact per customer per channel. A customer on WhatsApp and a customer on Telegram are tracked separately unless you have built a cross-channel merge on your end.

Channel Contact identifier
WhatsApp Phone number
Telegram Telegram user ID
Website widget Session-based (anonymous until name is shared)
Instagram Instagram user ID

Relationship stages

CI tracks six relationship stages automatically — no configuration required:

Stage When What it means for the bot
new < 5 messages Introduce yourself, ask questions, learn basics
building 5+ messages, active recently Growing rapport, reference shared topics
established 20+ messages, 5+ day streak Familiar tone, recall preferences naturally
deep 50+ messages, 14+ day streak Full personalisation, treat as a trusted contact
fading 7–29 days inactive Reduced context; re-engagement tone
dormant 30+ days inactive Minimal context; no proactive outreach

Key concepts

Memory — A single piece of information CI learned from a conversation. Memories have a type (fact, preference, episode, pattern), an importance score (0.0–1.0), and decay over time if not re-accessed. The bot can retrieve the most relevant memories before generating a reply.

Entity — A named thing attached to a contact: a pet, a person they mentioned, their workplace, a hobby, a place they live. Entities help group related memories ("all Bruno-related memories") and give the bot a structured understanding of the customer's world.

Mood — Classified on every customer message using an LLM (with keyword-based fallback). The current mood — happy, neutral, frustrated, sad, anxious, excited, or confused — is injected as a hint into the system prompt so the bot adapts its tone automatically.

Outreach trigger — A scheduled message to a contact: a re-engagement after inactivity, a milestone celebration (7-day streak, 100th message), or a manually scheduled check-in. Triggers are created automatically by CI and require no manual action.

Contact Intelligence · Getting startedEdit on GitHub