StarIntel local search

Goal

Search canonical local DB records and Auto-Dig packets before research, document creation, linking, or remote ingest.

Requires Python 3.11+, a current StarIntel Auto-Dig checkout, and starintel-spec-version.

Input

A query, identifier, dtype, dataset, predicate, source, or confidence requirement.

Output

Matching canonical documents with stable IDs and optional repository locations.

Command

Run from the Auto-Dig repository root:

python3 scripts/starintel.py search '<terms>' \
  --dtype <dtype> \
  --dataset <dataset-fragment> \
  --predicate <predicate-fragment> \
  --source <source-fragment> \
  --min-confidence 0.8 \
  --with-location

Use only the filters needed for the question. Results are JSONL. --with-location adds path, line, and surface around each document.

Workflow

1. Read the checkout's AGENTS.md, use starintel-spec-version to resolve the active release, and run python3 scripts/schema-release.py check before interpreting dtype/field semantics. A v0.9.0 schema filename identifies the immutable base family, not necessarily the active release.
2. Start with exact entity names, StarIntel IDs, external identifiers, source URLs, or predicates.
3. Search without a confidence floor first so contrary or uncertain records remain visible.
4. Narrow with --dtype, --dataset, --id, --predicate, or --source.
5. Preserve result IDs and locations in downstream notes or document proposals.
6. For link work, use scripts/search-db-links.py search, resolve, and neighbors; those commands prefer normalized DB records and fail on ambiguity.

Rules

• Resolve version authority before using schema-dependent filters or interpreting stored fields.
• Search both db/ and digs/ unless the task explicitly needs --db-only or --packets-only.
• Resolve ambiguous names with identifiers and source evidence; never choose a record merely because it ranked first.
• A relation match is evidence that a record asserts an edge, not proof that the edge is true.