duplix 0.1.0__tar.gz

duplix-0.1.0/.gitignore

@@ -0,0 +1,8 @@
+__pycache__/
+*.py[cod]
+*.egg-info/
+dist/
+build/
+.venv/
+.pytest_cache/
+*.egg

duplix-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Bhishan
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

duplix-0.1.0/PKG-INFO

@@ -0,0 +1,61 @@
+Metadata-Version: 2.5
+Name: duplix
+Version: 0.1.0
+Summary: Scaffold a new project by copying a bundled template project.
+Project-URL: Homepage, https://github.com/BhishanSharma/duplix
+Project-URL: Issues, https://github.com/BhishanSharma/duplix/issues
+Author-email: Bhishan <you@example.com>
+License: MIT
+License-File: LICENSE
+Keywords: cli,scaffold,template
+Classifier: Environment :: Console
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Requires-Python: >=3.8
+Provides-Extra: dev
+Requires-Dist: build; extra == 'dev'
+Requires-Dist: pytest>=7.0; extra == 'dev'
+Requires-Dist: twine; extra == 'dev'
+Description-Content-Type: text/markdown
+
+# duplix
+
+Scaffold a new project by copying a bundled template project, exactly as it is.
+
+## Installation
+
+```bash
+pip install duplix
+```
+
+## Usage
+
+```bash
+python -m duplix my_new_project
+# or, after install:
+duplix my_new_project
+```
+
+This copies duplix's bundled template folder verbatim into `./my_new_project/`
+(files, code, and structure unchanged - nothing is renamed or rewritten).
+
+### Options
+
+```bash
+python -m duplix my_new_project --dir ./somewhere
+```
+
+## Development
+
+```bash
+pip install -e ".[dev]"
+pytest
+```
+
+## Publishing
+
+```bash
+python -m build
+python -m twine upload dist/*
+```

duplix-0.1.0/README.md

@@ -0,0 +1,40 @@
+# duplix
+
+Scaffold a new project by copying a bundled template project, exactly as it is.
+
+## Installation
+
+```bash
+pip install duplix
+```
+
+## Usage
+
+```bash
+python -m duplix my_new_project
+# or, after install:
+duplix my_new_project
+```
+
+This copies duplix's bundled template folder verbatim into `./my_new_project/`
+(files, code, and structure unchanged - nothing is renamed or rewritten).
+
+### Options
+
+```bash
+python -m duplix my_new_project --dir ./somewhere
+```
+
+## Development
+
+```bash
+pip install -e ".[dev]"
+pytest
+```
+
+## Publishing
+
+```bash
+python -m build
+python -m twine upload dist/*
+```

duplix-0.1.0/pyproject.toml

@@ -0,0 +1,42 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "duplix"
+version = "0.1.0"
+description = "Scaffold a new project by copying a bundled template project."
+readme = "README.md"
+requires-python = ">=3.8"
+license = { text = "MIT" }
+authors = [
+    { name = "Bhishan", email = "you@example.com" }
+]
+keywords = ["scaffold", "template", "cli"]
+classifiers = [
+    "Programming Language :: Python :: 3",
+    "License :: OSI Approved :: MIT License",
+    "Operating System :: OS Independent",
+    "Environment :: Console",
+]
+dependencies = []
+
+[project.optional-dependencies]
+dev = [
+    "pytest>=7.0",
+    "build",
+    "twine",
+]
+
+[project.scripts]
+duplix = "duplix.__main__:main"
+
+[project.urls]
+Homepage = "https://github.com/BhishanSharma/duplix"
+Issues = "https://github.com/BhishanSharma/duplix/issues"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/duplix"]
+
+[tool.hatch.build.targets.sdist]
+include = ["src/duplix/**", "README.md", "LICENSE"]

duplix-0.1.0/src/duplix/__init__.py

@@ -0,0 +1,7 @@
+"""duplix - scaffold a new project by copying a bundled template project.
+
+Usage:
+    python -m duplix <project_name>
+"""
+
+__version__ = "0.1.0"

duplix-0.1.0/src/duplix/__main__.py

@@ -0,0 +1,62 @@
+"""Command-line entry point for duplix.
+
+Running `python -m duplix <project_name>` copies duplix's bundled template
+project, as-is, into a new folder named <project_name> in the current
+directory (or --dir). No files are renamed or rewritten - it's a plain
+copy of the template's files and code.
+"""
+
+from __future__ import annotations
+
+import argparse
+import importlib.resources
+import shutil
+import sys
+from pathlib import Path
+
+
+def get_template_dir() -> Path:
+    """Locate the bundled template directory inside the installed package."""
+    template = importlib.resources.files("duplix").joinpath("_template")
+    return Path(str(template))
+
+
+def scaffold(project_name: str, target_dir: Path) -> Path:
+    dest = target_dir / project_name
+    if dest.exists():
+        raise FileExistsError(f"'{dest}' already exists.")
+
+    template_dir = get_template_dir()
+    if not template_dir.exists():
+        raise FileNotFoundError(
+            f"Bundled template not found at '{template_dir}'. "
+            "The duplix package may be installed incorrectly."
+        )
+
+    shutil.copytree(template_dir, dest)
+    return dest
+
+
+def main(argv: list[str] | None = None) -> int:
+    parser = argparse.ArgumentParser(
+        prog="duplix",
+        description="Copy duplix's bundled template project into a new folder.",
+    )
+    parser.add_argument("project_name", help="Name of the new project folder to create")
+    parser.add_argument(
+        "-d", "--dir", default=".", help="Directory to create the project in (default: current dir)"
+    )
+    args = parser.parse_args(argv)
+
+    try:
+        dest = scaffold(args.project_name, Path(args.dir).resolve())
+    except (FileExistsError, FileNotFoundError) as exc:
+        print(f"Error: {exc}", file=sys.stderr)
+        return 1
+
+    print(f"Copied template to: {dest}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

duplix-0.1.0/src/duplix/_template/.gitignore

@@ -0,0 +1,3 @@
+.venv/
+__pycache__/
+*.pyc

duplix-0.1.0/src/duplix/_template/README.md

@@ -0,0 +1,154 @@
+# Flight Allocation
+
+Engine + web console for IndiGo CLC flight allocation. Python 3.11–3.14,
+no database, no Docker, no Node — the UI is a zero-dependency stdlib
+`http.server` app on `127.0.0.1:8765` with plain HTML/CSS/JS under
+`src/web/static/`.
+
+**Everything goes in and comes out through the browser.** The three
+input files are uploaded in the browser and held in the server's memory
+for the session; results are read back as JSON and, if you need a file
+to send on, downloaded as a one-off `.xlsx`. There is no console
+workbook, no `excel/` folder to manage, and no per-date file tree.
+
+The files are split by how often they change. The flight schedule is a
+fresh export every morning, so it sits on the dashboard. The two rosters
+cover a whole period and are set once, in the **Setup** sidebar.
+
+## Quick start (Windows)
+
+1. Double-click `start.bat`.
+2. First run takes ~1 minute: it creates a `.venv` and installs
+   `requirements.txt`. Every run after that opens the browser straight
+   to http://127.0.0.1:8765/.
+
+## Manual setup (Mac/Linux, or to run it yourself)
+
+```
+python3.11 -m venv .venv
+source .venv/bin/activate        # Windows: .venv\Scripts\activate
+pip install -r requirements.txt
+python -m src.cli                # --port 9000 --no-browser also work
+```
+
+## Daily flow
+
+0. **Once per period** — open **Setup** (top right) and upload the two
+   roster files. They stay loaded for the session, so this is not part
+   of the daily flow:
+
+   | Slot | What it is |
+   |---|---|
+   | Staff roster | regular crew roster |
+   | AM + ZC roster | AM and ZC roster |
+
+1. **Upload** today's flight schedule (the SV portal export) on the
+   dashboard's *Today's flight schedule* card.
+
+   Any sheet name works — each reader takes its canonical sheet if
+   present and otherwise falls back to the first sheet in the file.
+
+2. Check the allocation date (top right). It is prefilled from the
+   **server's** date, not the browser's, so a machine on a stale clock
+   can't plan the wrong day. Change it to back-fill another date.
+3. Click **Plan** — cleans the flights and extracts the roster, ~3s.
+4. Review the dashboard: per-shift stats, charts, roster by shift,
+   staffing recommendation.
+5. Click **Allocate** — CP-SAT solves, typically 30–90s on a first run.
+6. Adjust anything that needs it from the **Override** drawer:
+   - `p2f` / `norse` — handler nominations
+   - `sick` — pull someone out of the day
+   - `change_role` — promote STAFF→ZC, or push someone to AM
+   - `max_flights` / `cutoff_time` — per-staff caps
+   - add / remove a flight or a staff member
+7. Click **Allocate** again. Prior assignments are pinned and used as a
+   CP-SAT warm start, so only the affected staff's flights move and the
+   re-solve takes seconds rather than a minute. Rows whose staff changed
+   are flagged in the Allocations tab.
+8. **Download .xlsx** when you want a file to email or print.
+
+## Session lifetime
+
+State lives in the server process. Restarting it clears the uploads,
+the overrides and the results — start again by uploading the three
+files. **Reset** is narrower: it clears the results but keeps your
+uploads and override rows, so you can go straight back to Plan.
+
+## Layout
+
+```
+src/
+  state.py                  the in-memory store everything reads and writes
+  plan.py                   Plan stage: step 1 + step 2 + the summary
+  step1_clean_flights.py    clean the schedule into ops classes
+  step2_extract_roster.py   rosters -> availability
+  step3_allocate_flights.py solve, post-passes, outputs
+  staged_overrides.py       apply drawer mutations to the working data
+  recommender_staffing.py   per-shift required headcount
+  schemas.py                the typed row models the whole app shares
+  allocator/                eligibility, caps, pairings, post-passes
+  solver/                   CP-SAT model
+  io/
+    readers.py              parse the uploaded workbooks + override rows
+    export.py               render the state as a downloadable .xlsx
+  web/
+    server.py               the request loop + launcher (no endpoint logic)
+    core/                   Response, the Router, static file serving
+    api/                    one module per concern; each registers routes
+    readback/               state -> the JSON the UI renders
+    overrides/              override + config.yml mutations
+    runner.py               runs a stage on a background thread
+    static/
+      index.html            a shell; views bring their own markup
+      js/core/              dom, api, store, shared constants
+      js/ui/                tab registry, layout mounting, modals
+      js/views/             one module per tab (views/index.js registers)
+      js/charts/            one module per dashboard chart
+      js/panels/            dashboard blocks with their own load cycle
+      js/drawer/            the Override drawer and its sections
+      js/sidebar/           the Setup sidebar (the set-once roster files)
+      css/                  partials, listed in styles.css
+configs/
+  config.yml                column maps, status aliases, solver knobs
+  shift_limits.json         per-(shift, role) preferred / acceptable / cap
+```
+
+```stdlib http.server web UI.
+
+Routes (JSON unless noted):
+
+    GET  /                          text/html — single-page app
+    GET  /static/<path>             the JS module / CSS partial tree
+
+    GET  /api/inputs                which of the 3 files are uploaded,
+                                    grouped daily (schedule) / setup (rosters)
+    POST /api/inputs/<kind>         upload one (raw .xlsx body)
+    DEL  /api/inputs/<kind>         drop one
+    GET  /api/server_date           the operating date, per the server's clock
+    GET  /api/export.xlsx           download the current results
+
+    GET  /api/dashboard             headline counts + run status
+    GET  /api/allocations           every allocation row
+    GET  /api/workload              per-staff workload summary
+    GET  /api/pairs                 pair map
+    GET  /api/warnings              warnings
+    GET  /api/unallocated           unallocated flights
+    GET  /api/recommendations       Phase-R suggestions per unallocated
+    GET  /api/plan                  plan summary payload
+    GET  /api/handlers              P2F / NORSE handler picture
+    GET  /api/staffing              staffing recommendation
+    GET  /api/staff_names           roster names for the drawer dropdowns
+    GET  /api/shift_limits          current (shift, role) bands
+
+    GET  /api/overrides             override rows
+    POST /api/overrides             append one {values: [...]}
+    PUT  /api/overrides/<i>         overwrite row i
+    DEL  /api/overrides/<i>         delete row i
+    POST /api/overrides/apply_staged   apply staged rows to the data
+
+    POST /api/run                   {date, step} — plan | all | reset | …
+    GET  /api/run/status            current run state
+
+Bind: 127.0.0.1:8765. Everything lives in memory for the life of the
+process; there is no workbook and no per-date file.
+```

duplix-0.1.0/src/duplix/_template/configs/config.yml

@@ -0,0 +1,256 @@
+# Flight Allocation system — runtime config
+# All maps live here so onboarding a slightly different export shape only
+# requires editing this file, not Python source. See plan §1.5.1.
+
+allocation:
+  # Cycle year used to resolve year-less roster headers ("Mon,30Mar" -> 30-Mar-<cycle_year>).
+  cycle_year: 2026
+
+  # Default integration-run date. The launcher and CLI take --date to override.
+  # Sample SV portal contains only this date; AM roster covers it; staff roster
+  # ends 26-Apr so STAFF availability for this D will be empty (W010).
+  default_date: 2026-04-30
+
+# Step 3 solver parameters.
+solver:
+  # 2026-05-28 (user direction — restoring distribution quality): 350s
+  # was too aggressive. The solver was reaching FEASIBLE early but the
+  # S1b bucket-spread polish phase (which equalizes per-shift staff
+  # counts) kept getting cut short, producing -7 / -5 / -19 outliers in
+  # otherwise healthy shifts. 700s gives the solver the late-phase time
+  # it needs to drive max-min toward 0 across each bucket.
+  #
+  # Auto-retry (3 attempts in runner.py) is the safety net when a
+  # single attempt leaves UNALLOCATED >= 15 — worst-case total
+  # wall-clock = 3 * (700s + post-pass) ~ 40 min, ~12 min nominal.
+  # Never set to 0 — caused CP-SAT to search indefinitely on a prior
+  # attempt (99 min CPU, 4.6 GB RAM).
+  #
+  # Soft objective drives equality:
+  #   - S1 below-preferred: 100 per missing flight (pushes count up)
+  #   - S1 above-preferred: 50 / 500 / 1500 marginal (cap at +1/+2/+3)
+  #   - S1b bucket-spread:  600 per unit of (max - min) inside bucket
+  #     (bumped from 400 alongside the time bump — equalization is
+  #     more aggressive per-second so the budget pays off harder)
+  # Result: {23, 23} costs 100 vs {22, 24} costing 1750 — strongly
+  # prefers equal pairs over unequal pairs at the same sum.
+  max_seconds: 350
+  deficit_penalty_lambda: 100   # legacy knob; ignored by lex objective, kept for API compat
+
+# State / cycle-history.
+state:
+  staleness_max_days: 2          # plan §7.6 W003 threshold
+
+# ---------- I/O column maps ----------
+io:
+  sv_portal:
+    sheet: IN_SVportal           # was "Sheet1"; renamed under IN_*-prefix convention
+    # 2026-05-14: SV portal export format simplified — fewer columns,
+    # different header names. All aliases below (FLT/Flight No., DEP/Origin,
+    # ARR/Dest., STD/Dep., TYPE/A/C Type, LOAD/Pax Load) map to the same
+    # raw-row field so the reader picks up whichever is present. STA,
+    # AC-subtype, and Aircraft Owner are not in the new export — flights
+    # default to no-NORSE-owner / blank subtype.
+    columns:                     # source-col -> raw-row field (per §4.3)
+      FLT:           flight_id
+      "Flight No.":  flight_id
+      DEP:           departure
+      Origin:        departure
+      ARR:           arrival
+      "Dest.":       arrival
+      STD:           dep_time
+      "Dep.":        dep_time
+      STA:           arr_time
+      TYPE:          aircraft_type
+      AC:            aircraft_subtype
+      Aircraft:      aircraft_subtype
+      # 2026-05-14 new SV portal export: "A/C Type" holds the aircraft
+      # MODEL number (321, 789, 320 — what the legacy export called AC).
+      # Map to aircraft_subtype, not aircraft_type. aircraft_type
+      # (the J/B/K/X ops-class letter) is absent from the new export,
+      # so ops_class falls through to date-based DAY/NIGHT classification.
+      "A/C Type":    aircraft_subtype
+      DATE:          date
+      Date:          date
+      LOAD:          booked_pax_raw
+      "Booked Pax":  booked_pax_raw
+      "Pax Load":    booked_pax_raw
+      "Aircraft Owner": owner
+    date_format: "%m/%d/%Y"      # US-style m/d/Y per §1.5.1
+    # Step 1 drops rows whose Booked Pax does not match this regex.
+    pax_regex: '^\d+/\d+/\d+$'
+    # Phase 3 / Change 1 (2026-05-14, INTL overhaul): flights with
+    # DEP in this list are DROPPED at Step 1 extraction. The ARR-side
+    # pink-highlight machinery was removed in the same amendment;
+    # Gulf-3 ARR flights are plain domestic now.
+    routing_drop_dep_codes: [AUH, DOH, DXB]
+    # International airport list. Step 1 tags a flight with
+    # is_international=True if its **DEP** (3-letter code) is in this list
+    # (Change 9, 2026-05-14: DEP-side only; was DEP or ARR). Drives H12
+    # in Step 4 (newbies excluded) + the Change 4 D-75→STD coverage rule.
+    international_airport_codes:
+      - ATH
+      - ALA
+      - BAH
+      - BKK
+      - CAN
+      - CGK
+      - CMB
+      - DAC
+      - DMM
+      - DPS
+      - FJR
+      - GYD
+      - HAN
+      - HKG
+      - HKT
+      - JAF
+      - JED
+      - KBV
+      - KTM
+      - KUL
+      - KWI
+      - LGK
+      - MCT
+      - MED
+      - MLE
+      - MRU
+      - NBO
+      - PEN
+      - PVG
+      - RKT
+      - RUH
+      - SAI
+      - SEZ
+      - SGN
+      - SHJ
+      - SIN
+      - TAS
+      - TBS
+      - ZRH  # smoke test
+      - GVA
+
+  staff_roster:
+    sheet: IN_Staff             # was "Sheet1" — IN_*-prefixed naming since 2026-05-13
+    name_column: STAFF NAME     # was "Name" — input files use STAFF NAME
+    # Header pattern matches "Mon,30Mar" / "Tue,1Apr".
+    date_header_format: "%a,%d%b"
+    skip_blank_columns: true     # blank cols between Name and first date are dropped
+
+  am_roster:
+    sheet_glob: "IN_AM_Roster*"  # was "AM Roster*" — IN_*-prefixed naming
+    name_column: STAFF NAME      # was "NAME" — input files use STAFF NAME
+    # Header pattern matches "30-Apr".
+    date_header_format: "%d-%b"
+    # Number of leading rows after the header to skip (weekday sub-header in row 2).
+    subheader_rows_to_skip: 1
+
+# ---------- Status normalization ----------
+status:
+  # Aliases applied at read time before enum lookup.
+  # Note: keys are quoted so YAML 1.1 doesn't parse "OFF" / "ON" as booleans.
+  aliases:
+    "OFF": F                      # day off literal -> §4.1 F
+    # User direction 2026-05-10: future samples will write ZC entries
+    # as ZC/M, ZC/A, ... (current sample uses M/ZC, A/ZC). Both are
+    # accepted — the alias maps the new form to the canonical M/ZC.
+    "ZC/M":  "M/ZC"
+    "ZC/A":  "A/ZC"
+    "ZC/N":  "N/ZC"
+    "ZC/M1": "M1/ZC"
+    "ZC/A1": "A1/ZC"
+
+  # Recognized shift codes (current_shift will be populated when status is one of these).
+  shifts: [M, A, N, M1, A1]
+
+  # Recognized non-shift, non-assignable statuses (already in §4.1 enum).
+  non_shift_recognized: [U, F, "P/L", "C/L", "C/OFF", TR, ""]
+
+  # Recognized AM/ZC shift variants.
+  am_shifts: ["M/IGT", "A/IGT"]
+  zc_shifts: ["M/ZC", "A/ZC", "N/ZC", "M1/ZC", "A1/ZC"]
+
+  # Anything not matched above maps to CrewStatus.OTHER, assignable=False.
+  # Common occurrences in real AM data: N/CCC, L&T TRG, *Audit suffixes.
+  unrecognized_treatment: OTHER
+
+# ---------- Aircraft type -> ops_class lookup ----------
+# Per §4.13: day/night split is by date; other ops_classes (p2f, ferry, norse, test)
+# are derived from aircraft_type. Initial mapping below is a placeholder seeded from
+# the sample's TYPE values (G, H, J, K, P, T, X) — refine as user confirms semantics.
+# Any TYPE not listed falls through to date-based day/night routing.
+ops_class_by_aircraft_type:
+  # 2026-05-11 (revised per user): TYPE-letter routing rules.
+  #   H → P2F                                (passenger-to-freighter)
+  #   J → date-based DAY/NIGHT               (passenger normal)
+  #   X → handled in code: dep==arr → TEST, dep!=arr → FERRY
+  #   B → CHARTER                            (charter flight, own class)
+  #   K / P / T / G → FERRY                  (legacy ferry markers)
+  # TEST / FERRY / CHARTER are now allocated as normal flights and
+  # counted under workload (no per-staff special cap). The cleaned
+  # output groups them on one SpecialOps group for visibility.
+  H: p2f
+  B: charter
+  K: ferry
+  P: ferry
+  T: ferry
+  G: ferry
+  # 2026-05-27 (user direction): TYPE=W flights route to special ops
+  # (the FERRY group). Previously they fell
+  # through to date-based DAY/NIGHT routing and ended up in DayOps.
+  W: ferry
+
+# Post-pass §3 (user direction 2026-05-12) — when a handler has a P2F
+# flight, the surrounding domestic flights get removed from their list
+# (briefing / paperwork time) and redistributed via §4. Tolerance is
+# the half-width of the matching window around each anchor (D-3h,
+# D-1h, D+20min).
+#
+# 2026-05-28: ``logic_v2`` switches to the union-window-per-chain rule.
+# For each P2F at STD=T, the {owner P2F handler, pre-planner P2F
+# handler} chain loses ALL flights with STD in [T-2h, T+1h]. The P2F
+# post-pass also runs BEFORE the INTL post-pass when this flag is on,
+# so INTL D-75 doesn't remove a flight the P2F window would already
+# free up. Default false in main builds; the "p2f-logic" variant zip
+# ships with this set true.
+p2f_adjustment:
+  tolerance_minutes: 30
+  logic_v2: false
+
+ops_class_norse_owner_codes:
+  # NORSE flights are identified by the Aircraft Owner column (per user
+  # direction 2026-05-11). Case-insensitive equality match against any
+  # code below routes the flight to NORSE. Replaces the legacy
+  # (TYPE, Aircraft) compound rule — owners are explicit in the new
+  # export, so we no longer triangulate via type+subtype.
+  - N0
+  - AC-789
+
+# ---------- GULF routing (2026-05-19) ----------
+# A flight is GULF when its DEP airport is in the list below OR its
+# Aircraft Owner / TYPE letter equals one of the listed carrier codes
+# (case-insensitive). GULF flights land on the dedicated
+# GULF class so the assigner can see them for reference,
+# but Step 3's allocator does NOT read that sheet — they are never
+# assigned to any staff.
+ops_class_gulf_dep_codes:
+  - AUH
+  - DOH
+  - DXB
+# 2026-05-21 (user direction): owner-based Gulf detection (QR) dropped.
+ops_class_gulf_owner_codes: []
+
+# ---------- User-defined extraction filters (2026-05-25) ----------
+# A flight matching ANY filter below is routed to the GULF class
+# (extract-only — no staff allocation). Each filter is a dict with any
+# subset of fields. Empty / missing field = "don't match on that
+# field". Match is AND across specified fields, case-insensitive
+# exact. Example entries:
+#   - ac_owner: N0
+#   - ac_type: X
+#   - dep: BLR
+#     arr: COK
+#   - custom_header: REG
+#     custom_value: VT-IKW
+# Edit via the Override drawer's "Extraction Filters" form or by hand.
+extraction_filters: []