Bring the comments the review left stale to the code: judge without a prompt, extract reading again, searches closing by age, the cap at 4, the affinity bonus and the cursor a sweep hands back

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
WolfandClaude Fable 5.1 committed 2026-10-05 10:19:01 +02:00
1 parent b4774f1da9
commit e8157d9509
7 files changed
+16 -11

No files matched your search

+1 -1
View File
@@ -70,7 +70,7 @@ class SummariseConfig(ModelTaskConfig):
class JudgeConfig(ModelTaskConfig):
"""Each search agent's cut read against its CV, as three numbers and a
total. See app/jobs/judge.py. Off by default like extract; on, a pull
judges the cut of each enabled agent with a CV and a prompt, and only
judges the cut of each enabled agent with a CV, and only
the roles whose judgement is missing or was read from another CV or
text."""
+2 -1
View File
@@ -9,7 +9,8 @@ refused batch, and failing with a notice when no model is reachable are
`extract` is also the afterburner for the level and the remote area. A role
is waiting while it is open, has not been given up on, has never been
answered (`extracted_at IS NULL`), has no model level (`seniority_model`
answered or was answered before a better text came (`extracted_at IS
NULL`, which `Store.upsert_role` sets again), has no model level (`seniority_model`
empty), was never asked where a remote hire may live (`remote_where_model`
NULL) or for its languages and deadline (`languages_required` NULL), and is
in the cut of an enabled agent (`scores.cut`, see `app.jobs.score`). That
+1 -1
View File
@@ -32,7 +32,7 @@
* person likes and what they do not like (the work, the workplace,
* situations), which the judge reads for the career score. The must-have text
* is no longer on the form: a save leaves `must_text` out, so the stored one
* is kept (it is set through the API, and still caps career at 39 in code
* is kept (it is set through the API, and still caps career at 4 in code
* when the judge reports one broken).
* Every save sends the CV, languages and the two texts, and the card's switch
* re-posts the row as the server gave it, so neither drops them.
+1 -1
View File
@@ -104,7 +104,7 @@
<section class="card" id="company-boards" data-section>
<h2>Company boards</h2>
<p class="lede">Every company roledesk has seen, and how many of its roles are open. Switching a board off stops reading it and closes its roles; hiding a company keeps reading it but leaves its roles out of the list and unscored. Like or favourite raises a company's roles a little in the cut and in the feed order, in each search agent's own ranking (a liked company about 0.05, a favourite about 0.1), and passes nothing the fetch gate rejects. All locations and All roles are for every agent: the first skips the location gate and keeps the world gate, the second skips the role exclusion and the fetch terms.</p>
<p class="lede">Every company roledesk has seen, and how many of its roles are open. Switching a board off stops reading it and closes its roles; hiding a company keeps reading it but leaves its roles out of the list and unscored. Like or favourite raises a company's roles a little in the cut and in the feed order, in each search agent's own ranking (a liked company about 0.056, a favourite about 0.111 on the start weights), and passes nothing the fetch gate rejects. All locations and All roles are for every agent: the first skips the location gate and keeps the world gate, the second skips the role exclusion and the fetch terms.</p>
<form id="board-form" class="agent-form source-form" autocomplete="off">
<h3>Follow a company</h3>
+1 -1
View File
@@ -260,7 +260,7 @@ def list_roles(
when not judged), and `lang_flag` and `lang_note`, the posting's languages
against the agent's (`app.models.language_flag`).
`company` and `place` are substrings, kept for links from other pages.
`company` and `place` are substrings, kept for older links and saved filters.
The rail's fields are `company_keys` (repeated, exact keys) and
`places` (repeated place tokens, `app.places`: `country:DK`,
`region:skane`, `place:malmo`, matched on whole folded words), several
+6 -2
View File
@@ -167,8 +167,12 @@ shared code logs, records and decides.
handed to `rows`, its raw results as dicts with an `id` (counted, by distinct
id, to decide completeness), the total the host states or None, and the
cursor of the next page or None after the last. The first call gets cursor
None. A board with a total is complete when as many distinct ids were read;
one without, when the read ended on an empty page. A 429 after the first page
None, and a later one may get its cursor back from JSON, a tuple as a list,
when a sweep continues. A board with a total is complete when as many
distinct ids were read; one without, when the read ended on an empty page.
A reader that sees a page come again hands it over with no results and a
cursor that is not None, which ends the read incomplete; None there would
read as the last page. A 429 after the first page
ends the read, incomplete; `_runner.MAX_PAGES` pages a pull at most,
`_runner.PAGE_DELAY` apart, and a board cut there is swept across pulls
(`_runner.read_listing`). A paged module may also have `parse(body, slug)` for one page's body.
+4 -4
View File
@@ -13,10 +13,10 @@ and the saved `pages` param bounds the pages read (`DEFAULT_PAGES`, at most
`MAX_PAGES`).
A search result list is never a board's complete list, so a role missing from
one run is not closed here. A search that read everything that matches was
complete (`complete="stated"`, as many distinct results as the stated total),
and the pull closes what it stopped listing; the rest close once no search
has seen them for `STALE_DAYS` (see `app/jobs/pull.py`).
one run is not closed, also when the search read everything that matches
(`complete="stated"`, as many distinct results as the stated total): its
roles close once no search has seen them for `STALE_DAYS` (see
`app/jobs/pull.py`).
In a pull a search is read by `_runner.read` under the prefix `search`, as a
board is under `ats`: the 10% guard holds it, `_runner.stamp` records it,