Deep Dives
Detailed technical deep dives into EdgeQuake’s implementation.
- LightRAG Algorithm — The algorithm behind entity extraction and graph construction.
- Entity Extraction — How LLMs extract entities from text.
- Entity Normalization — Deduplication and canonicalization of entities.
- Chunking Strategies — Document splitting approaches and trade-offs.
- Query Modes — The 6 retrieval modes explained.
- Graph Storage — PostgreSQL AGE integration for graph operations.
- Vector Storage — pgvector for embedding-based retrieval.
- Embedding Models — Supported embedding providers and configuration.
- Community Detection — Graph clustering for global queries.
- PDF Processing — Vision LLM-powered PDF to markdown conversion.
- Gleaning — Multi-pass entity extraction for higher recall.
- Cost Tracking — Monitor and control LLM API costs.
- Pipeline Progress — Real-time processing status and progress tracking.