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+ and a current StarIntel Auto-Dig checkout.
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
- Read the checkout's
AGENTS.md and confirm it is current.
- Start with exact entity names, StarIntel IDs, external identifiers, source URLs, or predicates.
- Search without a confidence floor first so contrary or uncertain records remain visible.
- Narrow with
--dtype, --dataset, --id, --predicate, or --source.
- Preserve result IDs and locations in downstream notes or document proposals.
- For link work, use
scripts/search-db-links.py search, resolve, and neighbors; those commands prefer normalized DB records and fail on ambiguity.
Rules
- 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.
1---2name: starintel-local-search3description: starintel, local-search, jsonl, ndjson, relations, corpus4---56# StarIntel local search78## Goal910Search canonical local DB records and Auto-Dig packets before research, document creation, linking, or remote ingest.1112Requires Python 3.11+ and a current StarIntel Auto-Dig checkout.1314## Input1516A query, identifier, dtype, dataset, predicate, source, or confidence requirement.1718## Output1920Matching canonical documents with stable IDs and optional repository locations.2122## Command2324Run from the Auto-Dig repository root:2526```bash27python3 scripts/starintel.py search '<terms>' \28 --dtype <dtype> \29 --dataset <dataset-fragment> \30 --predicate <predicate-fragment> \31 --source <source-fragment> \32 --min-confidence 0.8 \33 --with-location34```3536Use only the filters needed for the question. Results are JSONL. `--with-location` adds path, line, and surface around each document.3738## Workflow39401. Read the checkout's `AGENTS.md` and confirm it is current.412. Start with exact entity names, StarIntel IDs, external identifiers, source URLs, or predicates.423. Search without a confidence floor first so contrary or uncertain records remain visible.434. Narrow with `--dtype`, `--dataset`, `--id`, `--predicate`, or `--source`.445. Preserve result IDs and locations in downstream notes or document proposals.456. For link work, use `scripts/search-db-links.py search`, `resolve`, and `neighbors`; those commands prefer normalized DB records and fail on ambiguity.4647## Rules4849- Search both `db/` and `digs/` unless the task explicitly needs `--db-only` or `--packets-only`.50- Resolve ambiguous names with identifiers and source evidence; never choose a record merely because it ranked first.51- A relation match is evidence that a record asserts an edge, not proof that the edge is true.
lost-rob0t