CosmicMind Python SDK¶

Persistent Memory for AI Systems

CosmicMind is an AI context management platform that gives neural networks persistent, structured memory. Unlike traditional stateless AI interactions, CosmicMind maintains relationships, learns from conversations, and builds contextual understanding over time—making any neural network system smarter, more personalized, and more capable.

What is CosmicMind?¶

CosmicMind transforms ephemeral AI conversations into persistent knowledge. By capturing entities, relationships, and context across interactions, it enables AI systems to:

Remember previous conversations and learned information
Understand complex relationships between people, places, and concepts
Personalize responses based on accumulated user context
Scale contextual understanding beyond token limits
Connect information across multiple conversations and users

How It Improves Neural Network Systems¶

Traditional LLMs are stateless—they forget everything after each conversation. CosmicMind adds persistent memory infrastructure that:

Extracts entities and relationships from conversations automatically
Stores knowledge in CosmicMind's optimized data infrastructure for fast retrieval
Injects relevant context into prompts based on conversation history
Maintains user-specific memory across sessions and interactions
Enables multi-user knowledge sharing while preserving privacy

This means your AI applications can provide context-aware responses that reference past interactions, understand user preferences, and maintain coherent long-term conversations—all without manual prompt engineering or hitting token limits.

Use Cases¶

AI Assistants: Build assistants that remember user preferences and history
Customer Support: Provide personalized support based on customer history
Game Masters: Create NPCs with persistent memory and evolving storylines
Education: Track student progress and adapt teaching approaches
Research: Maintain knowledge connections of research findings and relationships
Enterprise: Enable AI that understands your organization's context

Installation¶

pip install cosmicmind

Or install from source:

git clone https://github.com/pansynapse/CosmicMind-SDK
cd CosmicMind-SDK/python
pip install -e .

Features¶

Structured Memory: Persistent context stored securely in CosmicMind's proprietary infrastructure
Multi-LLM Support: Works with Cerebras, OpenAI, Anthropic, Google, and more
Avatar System: Create AI personalities with custom traits and knowledge
Private & Secure: User-specific memory isolation with API key authentication
Fast: Optimized context retrieval for low-latency responses
Token Tracking: Built-in usage monitoring and cost estimation
Simple API: Minimal code to add persistent memory to any AI app
Type Safety: Pydantic models for validation and IDE autocomplete

Type Safety with Pydantic Models¶

The SDK supports Pydantic models for type-safe API interactions with client-side validation:

from cosmicmind import CosmicMindClient
from cosmicmind.models import ChatRequest, ChatResponse

client = CosmicMindClient(
    api_key="sk-your-key",
    base_url="https://cosmicmind.pansynapse.com/api"
)

# Type-safe request with validation
request = ChatRequest(
    messages=["Hello, who am I?"],
    user_id="alice_123",
    llm="cerebras",
    llm_model="llama-3.3-70b"
)

# Returns typed ChatResponse
response: ChatResponse = client.chat.send(request)

# IDE autocomplete works!
print(response.message)
print(f"Tokens used: {response.token_usage['total_tokens']}")

Benefits: - Client-side validation - Catch errors before API calls - IDE autocomplete - Know exactly what fields are available - Type hints - Better code quality and fewer bugs - Self-documenting - Clear field names and descriptions

Backward Compatible: Dict-based and legacy parameter styles still work!

Supported LLM Providers¶

CosmicMind integrates with multiple LLM providers, allowing you to choose the best model for your use case:

Provider	Models	Best For
Cerebras	llama-3.3-70b, llama-3.1-8b	Fast, affordable inference
OpenAI	gpt-4, gpt-4-turbo, gpt-3.5-turbo	General-purpose AI tasks
Anthropic	claude-3-opus, claude-3-sonnet	Long-form content, analysis
Google	gemini-2.0-flash, gemini-pro	Multimodal, quick responses
Perplexity	sonar-medium, sonar-small	Web-grounded responses

All models benefit from CosmicMind's persistent context management.

Support¶

Documentation: https://docs.pansynapse.com/cosmicmind
Email: support@pansynapse.com
Issues: https://github.com/pansynapse/CosmicMind-SDK/issues

License¶

Proprietary - See LICENSE file for details.