Open Source · MIT License

The understanding layer
for adaptive AI products

Add persistent cognitive state to your AI applications. Maintain evolving understanding that learns patterns, detects identity, and adapts over time.

View on GitHub

pip install sxth-mind

The Problem

Conversations don't accumulate understanding

Traditional AI chatbots treat every conversation as new. They retrieve history but don't truly understand it.

Your users interact for days, weeks, months—but your AI never learns who they are.

The Solution

Derived cognitive state that evolves

Detects behavioral patterns over time
Identifies user identity and style
Tracks journey stages and progression
Adapts responses based on accumulated understanding

Architecture

Four integrated layers

Everything you need to add adaptive intelligence

UserMind

User-level cognitive state persisting across all interactions. Tracks identity type, behavioral patterns, preferences, and trust score.

ProjectMind

Context-specific state for deals, habits, learning paths, or any domain. Tracks journey stage, momentum score, and custom progress data.

ConversationMemory

Sliding window of recent messages with summaries of older conversations. Extracts topics and themes for context-aware responses.

Pluggable Architecture

Customize everything: LLM providers (OpenAI, Anthropic, etc.), storage backends (SQLite, PostgreSQL), and domain-specific adapters.

Usage

Simple API, powerful understanding

example.py

from sxth_mind import Mind
from sxth_mind.providers import OpenAIProvider
from examples.sales import SalesAdapter
 
# Initialize with your domain adapter
mind = Mind(
    adapter=SalesAdapter(),
    provider=OpenAIProvider(api_key="...")
)
 
# Chat - state accumulates automatically
response = await mind.chat(
    user_id="user_123",
    message="I just closed a big deal!",
    project_id="deal_456"
)
 
# Get accumulated understanding
state = await mind.get_state(user_id="user_123")
print(state.identity_type)   # "hunter"
print(state.patterns)        # {"closes_quickly": 0.85, ...}
print(state.trust_score)     # 0.72

Choose an Adapter

Use built-in adapters (Sales, Habits, Learning) or create your own for any domain.

Interact

Each conversation updates the cognitive state. Patterns emerge over time.

Adapt

Responses become personalized based on accumulated understanding.

Applications

Built for long-term understanding

Any AI experience where understanding should evolve over time

💼

Sales Assistants

Track deal stages, detect outreach patterns, adapt coaching based on seller archetype.

🎓

Learning Platforms

Personalize curriculum based on learning style, track skill progression, detect when learners are stuck.

🎯

Habit Coaches

Understand user motivation patterns, track streaks, send re-engagement nudges at the right time.

📈

Customer Success

Predict churn by tracking engagement momentum, personalize outreach based on user journey stage.

🤖

Personal Assistants

Remember preferences, communication style, and context across sessions without explicit memory retrieval.

🚀

Your Use Case

Create custom adapters for any domain that benefits from accumulated user understanding.

Clarity

What sxth-mind is not

Not an agent framework

Doesn't plan or execute actions

Not a vector database

No similarity search or RAG

Not a prompt library

Manages state, not templates

Not a chat history store

Stores understanding, not transcripts

Start building adaptive AI

Open source. MIT licensed. Ready to integrate.

View on GitHub View on PyPI

The understanding layerfor adaptive AI products