> For the complete documentation index, see [llms.txt](https://docs.replicats.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.replicats.ai/old-technical-foundations/beyond-llms/replicats-hybrid-approach.md).

# Replicats' Hybrid Approach

How we combine specialized models with targeted LLM usage for optimal trading decisions.

### The Power of Specialization

At Replicats, we recognize that no single technology can effectively capture all aspects of cryptocurrency markets. Instead, we've developed a hybrid approach that combines specialized models, each designed for specific aspects of market analysis, with selective use of LLMs where they add value.

### Foundation Models

Our approach centers on two specialized foundation models that form the core of our market understanding:

#### Graph Model

Our graph transformer processes market relationships through a sophisticated knowledge graph. This model captures:

* Token interactions and dependencies
* Protocol relationships and integrations
* Market participant behaviors

The graph model enables us to understand market structure beyond simple price relationships, providing insights into the complex web of interactions that drive cryptocurrency markets.

### Intelligent Integration

What sets our approach apart is how we combine these specialized models with other technologies:

#### Model Fusion

We combine insights from different models through a sophisticated fusion architecture. This allows us to:

1. Merge temporal and structural insights
2. Maintain the strengths of each model
3. Create unified market representations
4. Enable consistent decision-making

#### Strategic Use of LLMs

Unlike platforms that try to solve everything with LLMs, we use them selectively where they add real value:

**Where We Use LLMs:**

* Natural language interface for strategy definition
* Market narrative analysis and summarization
* Strategy explanation and reasoning
* Processing qualitative market information

**Where We Use Specialized Models:**

* Price prediction and pattern recognition
* Risk assessment and portfolio optimization
* Market structure analysis
* Real-time decision making

### Looking Forward

The future of cryptocurrency trading lies not in relying on a single technology but in intelligently combining specialized approaches. Our hybrid architecture provides a foundation for continuous innovation while maintaining the reliability and performance needed for effective trading.

Through this approach, we deliver a trading platform that combines the best of multiple technologies - the precision of specialized models, the flexibility of LLMs, and the reliability of traditional trading systems - all working together to create a more effective trading solution.