reference data models

Data models

The data model has three layers: Python dataclasses in src/lode/model.py, SQLite tables in src/lode/storage.py, and JSON responses from src/lode/cli.py and src/lode/daemon.py.

Python records

Record	File	Fields
Node	src/lode/model.py	id, kind, name, qname, path, start_line, end_line, signature, doc, confidence, content_hash, extra
Edge	src/lode/model.py	src, dst, kind, confidence, detail
FileIndex	src/lode/model.py	file metadata plus nodes and edges

SQLite tables

Table	Created in	Purpose
repos	src/lode/storage.py	Repository roots.
files	src/lode/storage.py	Indexed file metadata.
nodes	src/lode/storage.py	Graph nodes.
edges	src/lode/storage.py	Graph edges.
node_fts	src/lode/storage.py	FTS5 search index.
embedding_queue	src/lode/storage.py	Nodes waiting for embedding.
embeddings	src/lode/storage.py	Stored vectors.

Response records

Response	Source	Notes
Search result	src/lode/storage.py	Includes node identity, path, line range, and rank.
Context pack	src/lode/context.py	Includes summary, must-read items, top hits, related neighbors, and confidence.
Impact report	src/lode/graph.py	Includes target, callers, callees, blast radius, affected files, entrypoints, and truncation state.

Related pages

• Node and edge records
• Storage and search
• Daemon endpoints