{
  "id": "page:library-travel",
  "_kind": "Page",
  "_file": "wiki/library/travel.md",
  "_cluster": "wiki",
  "attributes": {
    "nodeKind": "Page",
    "title": "Travel (Curated-Dataset + SQL-Tool Pattern) (Library)",
    "displayName": "Travel (Curated-Dataset + SQL-Tool Pattern) (Library)",
    "slug": "library/travel",
    "articlePath": "wiki/library/travel.md",
    "article": "\n---\ndomain: business/travel\n---\n\n# Travel (Curated-Dataset + SQL-Tool Pattern)\n\nTwo processes that together turn a local SQLite file into a sophisticated\ntravel planner. The pattern is: do the curation once, then let an LLM\ncompose SQL against the curated database through Python scripts at plan\ntime. No MCP is involved at any stage.\n\n## Processes\n\n- **`flight-dataset-build.js`** -- discover open flight/airport/climate\n  data sources, design a SQLite schema, author and run Python 3 + stdlib\n  `sqlite3` ETL scripts, build indexes and denormalized helper views, and\n  produce `SCHEMA.md` so downstream agents can author SQL without guessing.\n- **`travel-plan-compose.js`** -- read `SCHEMA.md`, resolve a traveler's\n  loose intent into SQL-filterable constraints, compose and execute\n  Python query scripts against the DB, and return ranked itineraries that\n  include stopover-as-vacation options (two-destination trips that can\n  beat a direct flight on price). Every itinerary carries the verbatim\n  SQL that produced it as audit evidence.\n\n## Hard constraints (honoured by both processes)\n\n- Only `kind: 'agent'` tasks are used. No shell tasks, no sqlite3 CLI,\n  no MCP adapters.\n- All database creation, loading, indexing, and querying happens through\n  Python 3 scripts that use ONLY the standard library `sqlite3` module.\n- Database is opened read-only during planning (`mode=ro` URI) so the\n  planner cannot mutate the dataset.\n- SQL used in the plan output is carried verbatim as audit evidence.\n\n## Agents\n\nUnder `agents/`:\n\n- `trip-scope-planner` -- turns origin + window + interests into concrete dataset scope.\n- `open-data-scout` -- discovers authoritative open data sources.\n- `sqlite-schema-architect` -- designs the schema, indexes, views, and writes `SCHEMA.md`.\n- `python-etl-engineer` -- writes and executes Python + stdlib `sqlite3` ETL scripts.\n- `data-quality-inspector` -- validates the DB against expected queries + integrity checks.\n- `traveler-profiler` -- resolves loose traveler inputs into SQL-filterable constraints.\n- `sql-query-composer` -- authors and runs Python query scripts against the DB.\n- `stopover-strategist` -- finds stopover-as-vacation itineraries.\n- `itinerary-narrator` -- explains tradeoffs in plain language, ranks, and exports markdown.\n\n## Inspired by\n\n- https://github.com/mluggy\n- https://www.linkedin.com/posts/mluggy_%D7%9B%D7%9E%D7%95-%D7%9E%D7%99%D7%9C%D7%99%D7%95%D7%A0%D7%99-%D7%99%D7%A9%D7%A8%D7%90%D7%9C%D7%99%D7%9D-%D7%92%D7%9D-%D7%90%D7%A0%D7%97%D7%A0%D7%95-%D7%9E%D7%AA%D7%9B%D7%A0%D7%A0%D7%99%D7%9D-%D7%97%D7%95%D7%A4%D7%A9%D7%95%D7%AA-ugcPost-7448843353275858944-b7d4\n",
    "documents": [
      "specialization:travel"
    ]
  },
  "outgoingEdges": [
    {
      "from": "page:library-travel",
      "to": "specialization:travel",
      "kind": "documents"
    }
  ],
  "incomingEdges": [
    {
      "from": "page:index",
      "to": "page:library-travel",
      "kind": "contains_page"
    }
  ]
}