chore: prepare 0.2.0

CHANGELOG.md

@@ -2,7 +2,7 @@
 
 All notable changes are listed here. Format based on [Keep a Changelog](https://keepachangelog.com/).
 
-## Unreleased
+## 0.2.0 (2026-11-05)
 
 ### Added
 - `BidsSource` and one small class per city for 12 municipalities that share the `/bids/` page layout (Akko, Arad,
@@ -24,6 +24,18 @@ All notable changes are listed here. Format based on [Keep a Changelog](https://
 - `Michraz.search` no longer fails when only some sources are down. Their errors go to `last_errors`, and it raises
   only if every searched source failed.
 
+### Fixed
+- Kiryat Motzkin: rows about the tender committee (`ועדת מכרזים`, meetings and protocols) are no longer returned as
+  tenders.
+
+### Upgrading from 0.1.0
+- A default `Michraz()` search now reads 15 sources instead of 2. The first search takes about 20 seconds, mostly
+  because mr.gov.il is read page by page. Responses are cached for an hour, so later searches are fast. Use
+  `source="rmi"` (or any other name) to search one source.
+- Every `Tender` has a `kind`, and `to_dicts` and the CSV include it. `Michraz.search` hides staff vacancies by
+  default. Pass `kind=None` to get everything.
+- `RmiSource.fetch()` returns a lazy iterator like the other sources. Use `search()` for a list.
+
 ## 0.1.0 (2026-11-05)
 
 ### Added

docs/sources.md

@@ -106,16 +106,16 @@ correct and not confirmed.
   `סגירת המכרז`). `source_id` is the `id` in the link, or the tender number when a row has no such link (Dimona).
 - Dates have no hour, so a deadline is read as 23:59:59 of that day. This is an assumption: the real cut-off hour
   is in the tender documents, not on the page.
-- Rows without a closing date are kept with `deadline=None`. Kiryat Motzkin lists 26 of 29 rows that way.
+- Rows without a closing date are kept with `deadline=None`. Kiryat Motzkin listed 26 of 29 that way before its committee rows were filtered out.
 - Not covered: the 25 other sites that answered the same path with HTML but no `bids` table (for example Herzliya,
   Raanana, Holon, Rehovot, Ashkelon, Lod, Modiin, Karmiel, Tiberias, Afula) use a different layout. Jerusalem,
   Tel Aviv, Haifa, Petah Tikva, Rishon LeZion and Ashdod were not reachable the same way.
 - Several sites mix staff vacancies (job ads) into the same tables: Beit Shemesh, Dimona, Kfar Saba and Nof HaGalil.
   A row is a vacancy (`kind=vacancy`) when its table caption or category says `דרושים`, `כוח אדם`, `כח אדם`,
   `משאבי אנוש` or `מכרזי כ"א`, or its title says `דרושים`, `דרוש/ה` or `מכרז חיצוני/פנימי`. On 2026-11-05 this
   marked 28 rows as vacancies. It is a heuristic: a vacancy under an unexpected caption is still listed as a tender.
-- Kiryat Motzkin lists 29 rows titled `ועדת מכרזים` (tender committee pages). They are not tenders and are not
-  filtered out.
+- Kiryat Motzkin lists rows about its tender committee (category or title `ועדת מכרזים`: meetings and protocols, 24 of
+  29 rows on 2026-11-05). They are not tenders and are left out.
 - Azor uses the same platform but its `robots.txt` disallows every path, so it is not included.
 - The other 12 sites returned an HTML page (not a rules file) for `/robots.txt`, so they state no rules.
 - `tests/fixtures/bids_<city>.html` holds the real table of each city, cut to its header and the first three rows

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "michraz"
-version = "0.1.0"
+version = "0.2.0"
 description = "One clean interface to Israeli public tenders."
 readme = "README.md"
 requires-python = ">=3.10"

src/michraz/sources/bids.py

@@ -30,6 +30,7 @@
 _PUBLISHED_HEADER = re.compile(r"(מועד|תאריך) (פרסום|פתיחת)|התחלת")
 _DEADLINE_HEADER = re.compile(r"אחרון להגשה|סגירת")
 _VACANCY_SECTION = re.compile(r"דרושים|כוח אדם|כח אדם|משאבי אנוש|מכרזי כ\"א")
+_COMMITTEE = re.compile(r"^ועדת מכרזים")
 _VACANCY_TITLE = re.compile(r"דרושים|דרוש/ה|מכרז (חיצוני|פנימי|פנים)")
 
 
@@ -213,6 +214,9 @@ def __init__(self, **settings: Any) -> None:
     def records(self) -> Iterator[tuple[Table, list[Cell]]]:
         """Yields ``(table, row)`` for every row of every table that has a tender name column.
 
+        Rows about the tender committee itself (``ועדת מכרזים``, its meetings and protocols) are not tenders and
+        are left out.
+
         Raises:
             SourceUnavailableError: If the site cannot be reached.
             SourceFormatError: If the page has no bids table.
@@ -222,8 +226,12 @@ def records(self) -> Iterator[tuple[Table, list[Cell]]]:
         if not usable:
             raise SourceFormatError(self.name, "no bids table with a tender name column on the page")
         for table in usable:
+            name, category = _column(table.headers, _Header.NAME), _column(table.headers, _Header.CATEGORY)
             for row in table.rows:
-                yield table, row
+                if not (
+                    _COMMITTEE.search(_cell_text(row, name) or "") or _COMMITTEE.search(_cell_text(row, category) or "")
+                ):
+                    yield table, row
 
     def record_id(self, record: tuple[Table, list[Cell]]) -> object:
         _, row = record

tests/test_bids.py

@@ -164,3 +164,15 @@ def test_a_vacancy_is_recognized_by_its_title_when_the_table_has_no_caption():
 def test_the_caption_is_read_even_when_the_site_hides_it():
     tables = parse_tables('<table class="bids"><caption class="sr-only">דרושים</caption><tr><th>x</th></tr></table>')
     assert tables[0].caption == "דרושים"
+
+
+def test_tender_committee_pages_are_not_tenders(respx_mock, http):
+    body = (
+        '<table class="bids"><tr><th>שם המכרז</th><th>קטגוריה</th><th>תאריך סגירת המכרז</th></tr>'
+        '<tr><td><a href="./bids/?id=1">ועדת מכרזים 26\\1</a></td><td>ועדת מכרזים</td><td></td></tr>'
+        '<tr><td><a href="./bids/?id=2">פרוטוקול</a></td><td>ועדת מכרזים</td><td></td></tr>'
+        '<tr><td><a href="./bids/?id=3">מכרז 26\\6 מתקני פרסום</a></td><td>מכרזים</td><td>09/12/2026</td></tr></table>'
+    )
+    source, tenders = fetch(respx_mock, http, "kiryat-motzkin", body)
+    assert [t.source_id for t in tenders] == ["3"]
+    assert source.errors == []