embabel/embabel-llm-database
Large Language Model Database
What's novel
Large Language Model Database
Code Analysis
34 files read · 7 roundsA Spring Boot REST API + React UI that stores LLM model metadata (fetched from LLM Stats API and AWS Bedrock), and uses LLM-powered agents to recommend models based on natural language user requests by mapping requests to tags, filtering by provider, and narrowing by description.
Strengths
Clean multi-module Maven architecture with clear separation (core domain, agent logic, batch processing, REST server, UI). The agent-based approach of using LLMs as a 'shim' to parse heterogeneous JSON into a unified domain model is a creative pattern, and the Spring Batch pipeline for incremental data loading is well-structured.
Weaknesses
Heavy reliance on LLM calls for what could be deterministic JSON parsing (ModelParserAgent uses 4 separate LLM calls per model to extract sub-objects), making it slow, expensive, and non-deterministic. Error handling is inconsistent — null returns instead of exceptions, a Thread.sleep-based polling loop in ModelSuggestionService, in-memory session storage with no TTL, and the 'auth' is a randomly generated UUID that changes on every restart.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
40
Contributors
2
Files
154
Active weeks
11
Repository
Language
Java
Stars
3
Forks
3
License
Apache-2.0