Updated3 min read

OpenAI Engine & Models

Memron has transitioned to OpenAI high-performance, cost-effective models across all analysis pipelines and Playground RAG orchestrations.

Model Configuration

Task	Selected Model	Temperature	Purpose
`Playground RAG & Context Recall`	gpt-4o-mini	0.0 (Deterministic)	Grounds assistant responses strictly on retrieved memory context.
`Analysis & Fact Extraction`	gpt-4o-mini	0.1 (Structured JSON)	Extracts atomic facts, entities, and relationships into JSON schemas.
`Vector Embeddings`	text-embedding-3-small	N/A (1536 dims)	Generates semantic vectors for HNSW indexing in PostgreSQL pgvector.
`Memory Title Generation`	gpt-4o-mini	0.3	Summarizes extracted memories into concise 3-word titles.

Updated3 min read

Memron has transitioned to OpenAI high-performance, cost-effective models across all analysis pipelines and Playground RAG orchestrations.

Task	Selected Model	Temperature	Purpose
`Playground RAG & Context Recall`	gpt-4o-mini	0.0 (Deterministic)	Grounds assistant responses strictly on retrieved memory context.
`Analysis & Fact Extraction`	gpt-4o-mini	0.1 (Structured JSON)	Extracts atomic facts, entities, and relationships into JSON schemas.
`Vector Embeddings`	text-embedding-3-small	N/A (1536 dims)	Generates semantic vectors for HNSW indexing in PostgreSQL pgvector.
`Memory Title Generation`	gpt-4o-mini	0.3	Summarizes extracted memories into concise 3-word titles.