Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language: en
|
| 4 |
+
tags:
|
| 5 |
+
- agentic-ai
|
| 6 |
+
- llm-routing
|
| 7 |
+
- model-selection
|
| 8 |
+
- cost-optimization
|
| 9 |
+
- multi-provider
|
| 10 |
+
pipeline_tag: any-to-any
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# AIF-Router
|
| 14 |
+
|
| 15 |
+
**Intelligent Model Selection Engine** — Routes requests across 10+ LLM providers with quality-aware optimization and automatic fallback chains.
|
| 16 |
+
|
| 17 |
+
## 10+ Providers
|
| 18 |
+
|
| 19 |
+
The Router maintains up-to-date profiles for all supported providers including cost per 1K tokens, average latency, health status, and capability matrix.
|
| 20 |
+
|
| 21 |
+
## Routing Strategy
|
| 22 |
+
|
| 23 |
+
1. **Capability Match**: Filter providers by required features (vision, function calling, streaming)
|
| 24 |
+
2. **Quality Scoring**: Weight by benchmark performance for the task type
|
| 25 |
+
3. **Cost Optimization**: Select cheapest provider meeting quality threshold
|
| 26 |
+
4. **Latency Budget**: Enforce max latency constraints
|
| 27 |
+
5. **Fallback Chain**: Pre-configured cascade on failure (3+ deep)
|
| 28 |
+
|
| 29 |
+
## Provider Profiles
|
| 30 |
+
|
| 31 |
+
Each provider carries:
|
| 32 |
+
- **Health**: Current availability and error rate
|
| 33 |
+
- **Cost**: Input/output pricing per 1K tokens
|
| 34 |
+
- **Latency**: P50/P95/P99 response times
|
| 35 |
+
- **Capabilities**: Vision, tools, streaming, JSON mode
|
| 36 |
+
- **Quality**: Task-specific benchmark scores
|
| 37 |
+
|
| 38 |
+
## Repository
|
| 39 |
+
|
| 40 |
+
[https://github.com/frostyjay7813/AIF](https://github.com/frostyjay7813/AIF)
|