LMSerg/iola-1b-router-2026-05-28-merged
LMSerg/iola-1b-router-2026-05-28-merged is a 1 billion parameter router model based on Google's Gemma 3 architecture, fine-tuned for specific public city-data tool routing. This model specializes in generating strict JSON router actions for tools like `resolve_entity_field`, `search_entities`, `rag_search`, and official/person lookups. It is designed to direct queries to external APIs and RAG tools for mutable public facts rather than answering from its internal memory. The model is intended for direct local execution within the `iola-cli` using Python `transformers`.
Loading preview...
IOLA 1B Router Model Overview
This model, LMSerg/iola-1b-router-2026-05-28-merged, is a 1 billion parameter router model built upon the google/gemma-3-1b-it base architecture. It integrates a LoRA adapter from LMSerg/iola-1b-2026-05-27 to specialize its functionality.
Key Capabilities
- Strict JSON Router Actions: The model is specifically trained to output strict JSON formats for routing actions.
- Public City-Data Tool Integration: It is designed to interface with public city-data tools, including:
resolve_entity_fieldsearch_entitiesrag_search- Official and person lookup tools
- External Fact Resolution: The model prioritizes routing queries for mutable public facts to external API/RAG tools, preventing it from answering such queries directly from its internal memory.
- Local Execution: Optimized for direct local execution within the
iola-clienvironment using Pythontransformers, eliminating the need for external runtimes like Ollama or GGUF.
Good For
- Applications requiring precise, JSON-formatted routing decisions for public city-data queries.
- Integrating with existing
iola-cliworkflows that leveragetransformersfor local model inference. - Scenarios where external tools and RAG systems are preferred for dynamic and mutable information retrieval over internal model knowledge.