Ragex. Store. Backend. Dllb
(Ragex v0.19.0)
View Source
dllb-backed storage implementation for Ragex's knowledge graph.
Delegates all operations to Dllb.MetaAST.Query and the dllb connection
pool. Provides native HNSW vector search, graph traversal, and full-text
search capabilities that are orders of magnitude faster than the ETS
brute-force equivalents on large datasets.
Requirements
The dllb server must be running and configured:
config :dllb,
enabled: true,
host: "127.0.0.1",
port: 3009,
pool_size: 5
Summary
Functions
Returns the ordered list of schema statements executed by bootstrap/0.
Functions
@spec schema_statements() :: [String.t()]
Returns the ordered list of schema statements executed by bootstrap/0.
The table, secondary-index, and edge-index definitions come from
Dllb.Schema. The full-text and vector search indexes are built here so the
vector index dimension can be resolved dynamically from the configured
embedding model (see the :embedding_model config and
Ragex.Embeddings.Registry). A static dimension that does not match the
model's output makes the engine silently skip indexing every row.