Router Settings
Configure embedding parameters and model routing thresholds.
Retrieval Engine (ChromaDB)
Tune semantic search behaviour
512 tokens
2562048
20 tokens
0500
Model Routing
Pick which model handles each chunk
Fast Model
ActiveQwen3-4B
Used for low-complexity queries and rapid extraction.
Deep Model
Llama-3-70B
Used for complex reasoning and multi-hop synthesis.