| name | 0.4.6-understand-vector_database |
| description | [0.4.6] Storage system indexing embeddings for similarity search; backbone of RAG. Examples: Pinecone, Weaviate, Chrom |
understand-vector_database
CALL NUMBER: llm_memory_and_context_engineering.vector_database : deep_long_term_memory(2)
DEFINITION: Storage system indexing embeddings for similarity search; backbone of RAG. Examples: Pinecone, Weaviate, Chroma, FAISS. Enables retrieval from vast document collections.
Invoke this skill to understand vector_database down to its primitives. The RELATIVE ROOT below is the least-fixed-point closure of everything it bundles from — the full import cone, grouped by the lib each prim comes from. Projected from a prover-typed KB (MAP/SWI-Prolog consistency gate): every reference below resolves.
THE RELATIVE ROOT (the import cone, by lib)
from deep_long_term_memory
- em_significance_rank (d3): Ordering structure by event importance or impact; enables retrieval by significance rather than recency.
- em_temporal_index (d3): Organizational structure mapping events to time ranges; enables retrieval by temporal proximity.
from llm_memory_and_context_engineering
- approximate_nearest_neighbor (d1): Fast retrieval of roughly similar vectors accepting some error; enables scaling vector_database beyond exact search limits
- chunking_strategy (d1): How documents are split into retrieval units; choices affect recall/precision: fixed-size, semantic (by sentences/paragraphs), recursive, or agentic splitting.
- cosine_similarity (d1): A similarity measure between embedding vectors based on the cosine of the angle between them; invariant to vector magnitude
- embedding_dimensionality (d1): Number of dimensions in the embedding_space; higher captures more nuance but increases storage and retrieval_cost
- embedding_model (d1): A model trained to produce text_embedding vectors; may be a general encoder like CLIP, a domain-specific encoder, or a late-intercept from an LLM
- hierarchical_chunking (d1): Recursive splitting of text into nested chunks enabling multi-granularity retrieval from summary to detail
- locality_sensitive_hashing (d1): Hashing scheme preserving similarity relationships; used in approximate_nearest_neighbor for fast vector_database queries
- overlapping_chunks (d1): Chunking strategy where adjacent chunks share boundary tokens to prevent information loss at chunk_boundaries
- semantic_similarity (d1): Measure of meaning overlap between two text fragments in embedding_space; the basis for associative_memory retrieval
- context_fidelity (d2): The accuracy with which context content reflects ground truth; low fidelity leads to hallucinated reasoning from corrupted context
- memory_retrieval (d2): The act of fetching stored memory entries — by semantic similarity, keyword, metadata, or temporal proximity — to include in context
- associative_memory (d2): Recall triggered by related concepts; attention mechanism is a form of content-addressable associative memory enabling retrieval by similarity.
CONSUMERS (what needs this)
long_term_memory, retrieval_augmented_generation
Projected from the llm memory and context engineering KB (438 concepts / 213 relations) — consistency-typed by MAP; the facet list after the colon IS the cross-lib dependency web.
(leaf — this is an actual skill.)
