Find the website
Find the website
Finds a company's official website from its name and city, with no AI and without opening any website. A website is only written when it clearly comes up for this company; the company's listing on a directory is stored in a separate column.
Purpose
emails, socials, legal_mentions, techstack and pricing all require a site_web. A list
that has names but no websites cannot reach any of them. This module is the bridge.
It runs in three stages:
- Search. For each company, a single web search on its name and city. No website is opened.
The establishment
qualifier registries append to the legal name is stripped before searching:
NALDEO - SITE DE TARNOSbecomesNALDEO, otherwise the query targets a label nobody ever writes. - Count. Once the whole list has been searched, the block counts for how many different
companies each website came up, and through how many different pages. A website seen for at
least two companies through at least two different pages is a directory or a platform (a
directory has one page per company): its page goes into
page_annuaire, never intosite_web. No hand-maintained list of directories: the count takes care of it, in every language, and it accumulates from one list to the next. - Choose. Among the websites that are not directories,
site_webreceives the first one that meets either condition: it is the only non-directory website of the row and its home page is among the first five results; or it holds at least two results, including the first or second of the page. Otherwise the cell stays empty.
All traffic goes through the VPN proxy pool, one exit per worker.
Inputs
| Field | Required | Notes |
|---|---|---|
nom |
yes | Company name. Aliases accepted (name, raison sociale, entreprise…). |
ville / commune |
no | Added to the name in the search. |
code_postal |
no | Not used in the search. |
adresse |
no | Fallback: the town is read from it when the ville column is missing. |
telephone |
no | Used to recognise the same company from one list to the next and to pick its directory page. |
siren / siret |
no | Same role when the list has no phone. |
Like opendata, this module accepts rows with no contact detail at all: creating one is its
job. Rows with no usable name are kept in the output file untouched.
Outputs
| Column | Type | Description |
|---|---|---|
site_web |
url | The official website, written only when the rule above keeps it. An existing value is never overwritten. |
page_annuaire |
url | The company's page on a directory or platform, when the search shows one. |
site_finder_match |
text | What the decision was based on: site propre, page d'annuaire seule, aucun site retenu, site déjà présent, nom inexploitable or recherche indisponible. |
Every input column is carried through unchanged (column passthrough contract).
Lifecycle
POST /api/jobs/site-finder → pending → running → completed.
Rows that already have a website come back untouched and are never searched, so chaining this module after a cheaper one costs nothing on the rows the cheaper one already filled.
Results are written at the end of the job, once the count is done. A row whose search still fails
after the retries is written with site_finder_match = recherche indisponible rather than dropped.
On cancellation, the block decides with what it has already collected and writes every row.
Pipeline
| input | pois_any |
| output | pois |
| needs | nom |
| produces | site_web, page_annuaire, site_finder_match |
Recommended order: sirene → opendata → site_finder → emails. The first two are free and
instant; this one costs network time, so it should only see the rows they could not fill.
Errors
| Case | Behaviour |
|---|---|
| No usable name in the whole list | 400 with the list of columns that were read |
| Search engine unavailable | The row is retried on another VPN exit, then written empty with an explicit reason |
What's next
The output feeds any module that needs a website. emails is the usual next step.