What are our Goals?
We are building the System Level Intelligence for Mixture-of-Models (MoM), bringing Collective Intelligence into LLM systems.
Core Questions​
Our project addresses five fundamental challenges in LLM systems:
1. How to capture the missing signals?​
In traditional LLM routing, we only look at the user's query text. But there's so much more information we're missing:
- Context signals: What domain is this query about? (math, code, creative writing?)
- Quality signals: Does this query need fact-checking? Is the user giving feedback?
- User signals: What are the user's preferences? What's their satisfaction level?
Our solution: A comprehensive signal extraction system that captures 6 types of signals from requests, responses, and context.
2. How to combine the signals?​
Having multiple signals is great, but how do we use them together to make better decisions?
- Should we route to the math model if we detect both math keywords and math domain?
- Should we enable fact-checking if we detect either a factual question or a sensitive domain?
Our solution: A flexible decision engine with AND/OR operators that lets you combine signals in powerful ways.
3. How to collaborate more efficiently?​
Different models are good at different things. How do we make them work together as a team?
- Route math questions to specialized math models
- Route creative writing to models with better creativity
- Route code questions to models trained on code
- Use smaller models for simple tasks, larger models for complex ones
Our solution: Intelligent routing that matches queries to the best model based on multiple signals, not just simple rules.
4. How to secure the system?​
LLM systems face unique security challenges:
- Jailbreak attacks: Adversarial prompts trying to bypass safety guardrails
- PII leaks: Accidentally exposing sensitive personal information
- Hallucinations: Models generating false or misleading information
Our solution: A plugin chain architecture with multiple security layers (jailbreak detection, PII filtering, hallucination detection).
5. How to collect valuable signals?​
The system should learn and improve over time:
- Track which signals lead to better routing decisions
- Collect user feedback to improve signal detection
- Build a self-learning system that gets smarter with use
Our solution: Comprehensive observability and feedback collection that feeds back into the signal extraction and decision engine.
The Vision​
We envision a future where:
- LLM systems are intelligent at the system level, not just at the model level
- Multiple models collaborate seamlessly, each contributing their strengths
- Security is built-in, not bolted on
- Systems learn and improve from every interaction
- Collective intelligence emerges from the combination of signals, decisions, and feedback
Why This Matters​
For Developers​
- Build more capable LLM applications with less effort
- Leverage multiple models without complex orchestration
- Get built-in security and compliance
For Organizations​
- Reduce costs by routing to appropriate models
- Improve quality through specialized model selection
- Meet compliance requirements with built-in PII and security controls
For Users​
- Get better, more accurate responses
- Experience faster response times through caching
- Benefit from improved safety and privacy
Next Steps​
Learn more about the core concepts:
- What is Semantic Router? - Understanding semantic routing
- What is Collective Intelligence? - How signals create intelligence
- What is Signal-Driven Decision? - Deep dive into the decision engine