A3M Router: Biology-Inspired LLM Routing for Cost Savings and Performance
This repository profile is provided by osrepos.com, an open source repository discovery platform.

Summary
A3M Router is an open-source, biology-inspired LLM router designed to optimize costs and performance across over 80 providers. It leverages principles like swarm intelligence and neural plasticity to offer adaptive, resilient, and cost-effective AI routing. This tool helps users save significantly on AI costs by intelligently selecting the best model for each query.
Repository Information
Topics
Click on any tag to explore related repositories
Use at your own risk
OSRepos shares public repositories for knowledge and discovery only. Any installation, execution, configuration, or use of code from these repositories is the user's own responsibility. Always review the repository, source code, dependencies, licenses, and security implications before running or installing anything. OSRepos is not responsible for issues, damages, or losses resulting from third-party repositories.
Introduction
A3M Router is an open-source, biology-inspired LLM router that intelligently directs queries across more than 80 AI providers. It addresses the challenges of centralized AI routing and high costs, offering a decentralized and highly efficient alternative. Inspired by 3 billion years of biological intelligence, A3M Router aims to build the "nervous system" for AI, ensuring optimal performance and significant cost savings.
Why Use A3M Router and Key Benefits
A3M Router stands out by offering substantial advantages for developers and organizations working with LLMs. Its core philosophy, inspired by biological intelligence, translates into tangible benefits:
- Significant Cost Reduction: By intelligently routing queries, A3M Router can reduce AI costs by 70-95%. Simple questions, for example, can be routed to cheaper models like Groq, costing 300x less than premium models.
- Superior Performance: Benchmarks show A3M Router achieving a P99 latency of 162ms, making it 14% faster than some alternatives.
- Exceptional Uptime: Leveraging "Swarm Intelligence," similar to how ant colonies find optimal paths, the router ensures 99.99% uptime. If a model fails, requests automatically avoid it, and new paths emerge.
- Adaptive Learning: Inspired by "Neural Plasticity," A3M Router constantly adapts. Time-decayed weights ensure that recent provider performance is prioritized, preventing overfitting to outdated data.
- Provider Diversity: The "Competitive Exclusion" principle, using the EXP3 algorithm, promotes a natural equilibrium among providers, preventing any single one from dominating and ensuring a diverse ecosystem.
- Intelligent Resource Allocation: The "Handicap Principle" treats cost as a credibility signal. High-cost models are reserved for high-stakes, complex queries, ensuring optimal resource use.
- Extensive Provider Coverage: With support for over 80 providers, including OpenAI, Anthropic, Google, Groq, and Mistral, A3M Router offers unparalleled flexibility.
- Open-Source and Self-Hostable: Unlike some centralized solutions, A3M Router is 100% open-source and can be self-hosted, giving users full control and transparency.
Installation
Getting started with A3M Router is straightforward. Choose your preferred package manager:
# npm
npm install adaptive-memory-multi-model-router
npx a3m-router serve
# Python
pip install a3m-router
python -m a3m_router.serve
# Docker
docker run -p 8787:8787 ghcr.io/das-rebel/a3m-router
Examples
Once installed, you can integrate A3M Router into your applications. It's designed to be compatible with OpenAI's API format.
Basic Routing
Simply change your base_url and set model="auto":
from openai import OpenAI
client = OpenAI(base_url="http://localhost:8787/v1", api_key="not-needed")
response = client.chat.completions.create(
model="auto", # A3M picks the best provider
messages=[{"role": "user", "content": "What is Python?"}]
)
print(response.choices[0].message.content)
Parallel Ensemble, Best Answer Mode
For critical queries where the best possible answer is paramount, A3M Router can query multiple providers simultaneously:
from a3m.router import A3MRouter
router = A3MRouter(
model="auto",
parallel_ensemble=3, # Call 3 providers simultaneously
)
result = router.route(
messages=[{"role": "user", "content": "Explain quantum entanglement"}],
ensemble_timeout_ms=10000,
)
print(f"Best from: {result.provider}")
print(f"Response: {result.content}")
Semantic Memory and Context
A3M Router also includes semantic memory capabilities to maintain context across interactions:
from a3m.router import A3MRouter
router = A3MRouter(
model="auto",
memory={
"type": "semantic",
"window": 10, # Last 10 exchanges
"similarity_threshold": 0.85,
}
)
# Second call uses cached context automatically
result = router.route(
messages=[{"role": "user", "content": "What framework should I use?"}]
)
# A3M knows "Python web app" from previous context
Links
Explore A3M Router further through its official channels:
Related repositories
Similar repositories that may be relevant next.
Source repository
Open the original repository on GitHub.
