Merge branch 'worktree-agent-a3bd292551395dda0'

# Conflicts:
#	app/storage/schema.py
This commit is contained in:
Wolf committed 2026-09-29 14:16:35 +02:00
commit 43773efc53
8 files changed
+338 -34

No files matched your search

+29 -10
View File
@@ -23,7 +23,7 @@ from desk.models import Item
from app.models import (
AFFINITY, REMOTE, STATUSES, Agent, Application, ApplicationEvent, Candidate,
Company, Gate, Limits, Role, ScoredRole, Subscription, Weights, company_key,
country_of, place_key, remote_of, title_key,
country_of, place_key, posted_at, posted_host, remote_of, title_key,
)
from app.storage import schema
@@ -263,6 +263,12 @@ class Store:
from `item.raw` under their own names, with `company` falling back to
the subscription's and `country` to what `place` names; an empty value
never overwrites a filled column.
`posted_at` (see `app.models.posted_at`) is set from this pull's own
`published_at` and `stamp`; on an update the `ON CONFLICT` clause below
keeps the earlier of that and what is already stored, so an edit on
the board — which can only ever move `published_at` later, never
earlier — cannot move `posted_at` later either.
"""
if not item.source:
raise ValueError(f"item {item.source_id!r} has no source; the runner sets it")
@@ -270,6 +276,7 @@ class Store:
company = str(raw.get("company") or (subscription.company if subscription else ""))
place = str(raw.get("place") or "")
stamp = self.now()
published = _iso(item.published_at)
row: dict[str, Any] = {
"uid": item.uid,
"source": item.source,
@@ -284,7 +291,8 @@ class Store:
"description": item.body,
"feed_text": item.excerpt,
"original_url": item.location,
"published_at": _iso(item.published_at),
"published_at": published,
"posted_at": posted_at(published, stamp, posted_host(raw)),
"first_seen": stamp,
"last_seen": stamp,
"raw": json.dumps(raw, default=str),
@@ -302,6 +310,7 @@ class Store:
for c in (*_RAW_INT, "subscription_id")]
updates += [f"{c}=excluded.{c}"
for c in ("title", "title_key", "published_at", "raw", "last_seen")]
updates.append("posted_at=MIN(excluded.posted_at, roles.posted_at)")
updates.append("closed_at=NULL")
columns = ", ".join(row)
marks = ", ".join(f":{c}" for c in row)
@@ -429,12 +438,12 @@ class Store:
tags: Sequence[str] = (), tag_mode: str = "any",
agent: int | None = None, by_score: bool = True, gated: bool = False,
judged: str | None = None, offset: int = 0) -> list[ScoredRole]:
"""Newest first by `first_seen`. `company` and `place` are
"""Newest first by `posted_at`. `company` and `place` are
case-insensitive substrings of the columns a board wrote — not the
key, so a search does not have to know the exact slug a company folds
to. `q` is matched word by word against `title`; a role must match
every word. `function` and `remote` match a gate column exactly.
`days` keeps only roles first seen within that many days of now.
`days` keeps only roles posted within that many days of now.
`agent` says whose votes, hidden companies and scores are meant. Each
row carries that agent's `vote` and `voted_at`; `judged` (a key of
@@ -471,10 +480,20 @@ class Store:
tags=tags, tag_mode=tag_mode, agent=agent, by_score=by_score, gated=gated,
judged=judged)
select = "SELECT roles.*, v.vote, v.voted_at"
order = "roles.first_seen DESC, roles.uid"
if agent is not None and by_score:
order = "roles.posted_at DESC, roles.uid"
scored = agent is not None and by_score
if scored:
select += f", {_VOTED_SCORE} AS score, s.gated_reason"
order = "score DESC, " + order
elif open_only:
# Newest first, open only: idx_roles_posted covers this exactly
# (posted_at DESC, uid, WHERE closed_at IS NULL), so it earns the
# LIMIT rather than sorting every open role first. Left to the
# planner, sqlite3's bundled build (no STAT4) reads the near-100%
# selectivity of `closed_at IS NULL` from idx_roles_closed's plain
# average and picks that instead, sorting the lot — checked with
# EXPLAIN QUERY PLAN on a copy of the live database.
sql = sql.replace("FROM roles ", "FROM roles INDEXED BY idx_roles_posted ", 1)
found = [ScoredRole.from_row(r) for r in self._rows(
f"{select} {sql} ORDER BY {order} LIMIT ? OFFSET ?", [*params, limit, offset])]
by_uid = self._tags_by_uid([r.uid for r in found])
@@ -521,7 +540,7 @@ class Store:
where.append("remote = ?")
params.append(remote)
if days is not None:
where.append("first_seen >= ?")
where.append("posted_at >= ?")
params.append(_iso(self.clock() - timedelta(days=days)))
for word in (q or "").split():
where.append("title LIKE ? ESCAPE '\\'")
@@ -641,7 +660,7 @@ class Store:
return (f"""r.uid IN (SELECT uid FROM (
SELECT s.uid, s.score, a.limits, ROW_NUMBER() OVER (
PARTITION BY s.agent_id
ORDER BY s.score DESC, top.first_seen DESC, s.uid) AS n
ORDER BY s.score DESC, top.posted_at DESC, s.uid) AS n
FROM scores s
JOIN agents a ON a.id = s.agent_id AND a.enabled = 1
JOIN roles top ON top.uid = s.uid AND top.closed_at IS NULL
@@ -936,7 +955,7 @@ class Store:
columns.append(
f"CASE WHEN {empty} THEN -1 WHEN {match} THEN 1 ELSE 0 END AS g{n}")
params += values
select = ", ".join(["r.uid", "r.first_seen",
select = ", ".join(["r.uid", "r.posted_at",
"COALESCE(ac.affinity, 0) AS affinity", *columns])
rows = self._rows(
f"""SELECT {select} FROM roles r
@@ -944,7 +963,7 @@ class Store:
ON ac.company_key = r.company_key AND ac.agent_id = ?
WHERE r.closed_at IS NULL AND COALESCE(ac.affinity, 0) <> -1""",
[*params, agent_id])
return [Candidate(uid=r["uid"], first_seen=r["first_seen"],
return [Candidate(uid=r["uid"], posted_at=r["posted_at"],
affinity=int(r["affinity"]),
gates=[int(r[f"g{n}"]) for n in range(len(gates))])
for r in rows]