Measured on the live roles against the "Copenhagen engineering" agent's
intent: the English-only all-MiniLM-L6-v2 separates engineering from
non-engineering roles with an AUC of 0.68 overall but only 0.74 restricted
to Danish-country roles (a Danish engineering role beats the rest's
non-engineering median only 68% of the time). multilingual-e5-small reaches
0.70 overall and 1.00 on Danish-country roles, for about 3x the CPU time.
paraphrase-multilingual-MiniLM-L12-v2 measured worse than the English-only
baseline (AUC 0.60) and was dropped.
e5 needs "query: "/"passage: " prefixes; app/jobs/embed.py adds them by
matching the model name so a model change can't leave a stale prefix
setting behind. Added tests for the prefixing and for the two invariants
the model change relies on: embed drops the old model's vectors before
re-embedding, and an agent's cached intent vector is keyed by model so a
model change is a cache miss, never a stale-dimension read.
Co-Authored-By: Claude Sonnet <noreply@anthropic.com>
A require gate on an empty column gated the role ("country unknown"), and extract, the stage that fills the column, is bounded to the roles that rank, so 2,281 of 15,556 open roles could never come back: every board that writes "Remote", "Hybrid" or "Distributed" as its location. An empty column now neither passes nor fails a gate; the role ranks, extract reads it, the rescore decides.
`any` is a third gate mode: the any gates are one group a role must match one of, so a ranking can say "in Denmark, or remote", which require alone could not. The group is gated only when some member had evidence against it.
`remote_of(place)` fills `remote` from the location line the way `country_of` fills `country`, and the startup backfill does both, since Greenhouse writes "Hybrid" as the location and gives no flag.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Run one stage goes: the cascade is the unit of work and app job keeps the
per-stage entry points. The pull button is disabled while a pull runs and
the lede no longer hardcodes the cadence. Messages folds into Runs as
Problems. pull_batch checks for a cancel before every board, one plugin per
target, and both plugins now close their HTTP client.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
agents.limits holds extract_top, extract_min, summarise_top and
summarise_min; the scope query reads them per ranked row, and a row saved
before this behaves like a fresh agent. The four global config keys go.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
`app companies prune`, `/api/prune` and the Sources disclosure switched off every board with no open role in a region. It existed because the first seed was 186 US boards; the European seed makes it moot, and it was wrong on its own terms: a board with no European role today (a company between openings, or a fetch that timed out) was switched off for good. A company that hires nowhere near should not be in the list, and a ranking's country gate handles the rest. `EUROPE` in models had no other reader and goes with it.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The seed was jobhunt-hq's 186 US boards, chosen because they could be read. It is now a hand-picked list of European tech employers, with `board` filled by the new `app.jobs.discover` from each company's own careers site and every board fetched once before it landed: 175 have one, and the rest keep the ATS host their site runs on (Teamtailor 32, Workday 12, SuccessFactors 9, ...) and a careers URL, as rows the app cannot read yet. Companies shows them as "not readable yet" with a link, behind a filter, so the next parser to write is the host that filter shows most.
`app companies discover NAME DOMAIN [--add]` reads the home page, its careers links and the usual paths for a link into a known host, one level down on a careers page, and probes Greenhouse, Lever, Ashby, SmartRecruiters, Recruitee and Personio by slug when the pages link nothing. It is what the LinkedIn flow will call.
Migration 7 adds companies.ats, country and city. The Workable parser now reads the v3 shape its list endpoint actually sends (location object, department list, shortcode and no URL or text); the first populated board showed it.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The free stages run first and their score bounds the two model stages.
Extract reads only the best extract.per_agent roles of each agent at
extract.min_score or more, then score runs again so the gates read what it
filled; summarise gets a min_score of its own. The cascade steps by
position, since score is in it twice.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
A search agent is a persona, and every like or pass belongs to the one
being viewed. The rail now opens on Viewing as: a full-width chooser,
the name alone with one agent, a note with none, remembered in the URL
and localStorage. The vote buttons, the reader's affinity control, the
feed, the tag counts and the reader fetch all carry that agent; before,
votes sent none and every one landed on the default agent. Best first
and Newest first are back as the segmented control under Order,
independent of the agent. The rail selects no longer clip descenders:
Chrome ignores line-height on a select, so the fixed box is gone in
favour of padding and a min-height.
Settings render inside a sticky rail (Search agents; Sources with Job
boards, Company boards, Runs, Messages; Account), a strip under the
header on a phone, so Runs is one click away however long the company
list gets. Sources shows one section at a time, chosen by the fragment,
so the old redirects still land. Search agents are cards, newest first:
name, an On/Off switch with its word, the intent in full, the weights,
Edit in place, New agent on top, and a link to Roles as that agent.
Switching one on rescores it.
replace_scores writes nothing for an agent deleted while score was
running, instead of failing the run on the foreign key.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The per-agent migration becomes v6 behind main's v5. A vote still stamps
the role's read_at, tag counts hide the agent's hidden companies, and
/api/filters takes the agent like /api/roles.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Votes on roles and company affinity move to role_votes and agent_companies,
one row per search agent, so two agents rank the same roles differently. The
score gains taste, the cosine to mean(liked) - mean(passed) of the agent's
votes, as newsdesk ranks articles. The API takes agent on roles, votes and
companies and falls back to the default agent, the enabled one with the
lowest id.
Settings become Search agents (a list again, with New and Delete), Sources
(job boards, company boards and runs with the message log, one section per
plugin) and Account. Old paths answer 301 to their section.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The Roles tab takes newsdesk's shell, ported from its index.html and
app.css rather than rewritten: a filter rail (Show read, Best first or
Newest first, first seen, tag chips with counts, the old filters behind a
disclosure), the feed in newsdesk's row layout (score gutter, title,
company, place, badges, age, tags, the summary as excerpt, judge buttons)
and the reader as a third column that replaces the list on a phone. The
router pushes /roles/{uid}, which the server also renders with the reader
open, so the separate read page and role.js go. j/k, u/n/d, S, Enter,
Escape, o, /, f and c work as in newsdesk. The phone rules come with their
comments.
Two things the row lacked: roles.read_at, stamped when the reader opens a
role or a vote is cast, behind include_read on /api/roles; and role_tags,
written by summarise alongside the summary and served with counts by
/api/filters, filtered in SQL before LIMIT. Migration 5.
PyCharm run configurations in .run/ for serve, pull, seed and pytest.
core.autocrlf=input is set locally so a Windows-side checkout commits LF.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
README's tabs table now matches two tabs and the settings cog, and
mentions the read page, votes, and the Fetch/Summarise buttons.
docs/decisions.md gets a new "The phone" section recording each
decision the mobile plan made, confirmed against the merged code.
docs/mobile-plan.md is deleted per its own package C.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
What a pull reads can now be changed on a page instead of only the command
line: try a search and see its first ten results before saving it, edit or
remove one, follow a company's ATS board, switch any board or search on or
off, and preview and apply a prune by region. Each row shows its open roles
and the countries they are in.
The page and `app companies`/`app searches` call the same functions in
app/jobs/sources.py. Switching a target off closes its open roles, and the
pull's settle closes them again if a batch in flight reopened them. Roles
pulled before `country` existed are backfilled at startup.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The ats boards only read companies someone listed, and the seed list is
US-centred: 31 of 29,000 open roles were in Denmark. A `search` plugin runs
saved queries on boards that span companies, pulled in the same task as the
boards: Jobindex (Denmark, RSS, no key) and Adzuna (UK and most of the EU,
free key, built to its documented API). Jobnet and Jooble were checked and
left out.
`country` is now filled from the board's location line by a lookup, with EU
for Europe-wide roles, so country gates work without extract. `app companies
prune --in` unsubscribes boards with no open role in a region. The starter
agents and searches are Copenhagen and Europe.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
One entry replaces "own repo, depending on news" and "desk and pipeline
are never edited here": built on the boilerplate with desk copied from
news, all four downstream copies, fixes upstream first, package app. The
superseded 8 MB cap entry is gone and the 96 MB one is unwrapped. Long
entries are cut to the decision and one reason, and the stale names
(roles.yaml, the one alpha page, app.js) follow the restructure. New: pull
as a job on a 6h schedule, and the coupling rule's place in the layers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With fit as (cos + 1) / 2 the real run spanned about 0.55 to 0.79, so a
prefer gate at weight 0.2 moved a score more than the whole spread of fit.
fit is now the cosine min-max scaled across the roles an agent scores in
that run: the closest is 1, the farthest 0, a single role 1, and a role a
require gate rules out stays out of the range. The starter agents' prefer
weights drop to 0.05, enough to break a tie. A regression test pins the
case: two close cosines that the old mapping let a 0.05 gate reorder.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
embed stores one vector per open role; score rates every open role for
every enabled agent from require and prefer gates, fit against the
agent's intent, recency and company affinity, keeping gated rows with a
reason. Schema v3 adds vectors and the cached intent vector. The page
gets an agent order, the score on each row, and an Agents tab. Two
starter agents in roles.yaml are imported once.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
roledesk job extract reads open, unanswered roles 20 to a call and fills
place, country, remote, seniority, salary, currency, function, permit and
languages, only into columns the board left empty. Schema v2 adds
roles.permit, roles.languages and roles.extract_tries. Off by default.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
/, GET /api/roles|companies|applications, POST /api/companies/{key}/affinity
and /api/applications[/{id}/status]. The Jinja2 template renders an empty
shell; static/app.js fills it and drives Save, affinity and status changes.
<desk-jobs> from pipeline.web mounts at the bottom so Pull can be started
from the page.
Store.roles gains q, function, remote, days and hide_muted, and company/
place move from a key match to a substring on the text a board wrote.
hide_muted filters in SQL before LIMIT: filtering after it shipped first and
broke immediately on real data, since Databricks' freshly pulled 884 roles
were the newest 200-plus rows in the table and muting it left nothing to
show.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
openings.go's 8 MB cap was written against arbitrary URLs; every URL here is
on a fixed allow-list of seven ATS hosts, and the real pull of 2026-09-23
refused Anduril, Anthropic, OpenAI, Databricks, Carvana, Axon and Airwallex
for it. A board that still exceeds 96 MB now retries once without
content=true (Greenhouse only), trading the description for the listing.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The real run against the seeded 186 companies refused seven of the
largest boards outright, Anduril worst at 41 MB, because the cap and
requesting full descriptions pull against each other on exactly the
employers that post the most roles. Left as is for now; the fix is a
retry without content=true, tracked as an open decision.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
seed/companies.json: 179 companies from jobhunt-hq's seed.json (MIT),
reduced to key/name/board/careers_url, plus a hand-added DACH set.
`roledesk companies seed|add|check` import it, add one company from its
board with a live check first, and report dead boards.
roledesk/jobs/pull.py is now a pipeline task: ten boards per pull_batch job
on the http resource, upserting every role the ats plugin yields. Its final
job's settle closes every board this run answered in full and no longer
lists a role from, reading the plugin's own self.state stamps rather than
each batch job's return value, so one bad board in a batch does not cost the
other nine their closure. `roledesk job pull` runs it, and it registers
itself in pipeline.TASKS so <desk-jobs> can start it too.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
roledesk/storage/schema.py creates the tables at v1 with WAL and the
versioned-migration machinery from newsdesk; store.py adds upsert_role
(unpacks raw into gate columns, first_seen on insert, last_seen every
time), roles(), mark_closed, companies and subscriptions CRUD, and
save_application writing an append-only application_events row on every
status entered. Key helpers company_key, title_key, place_key live in
models.py. docs/decisions.md records the key decisions.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>