Recherche

Pourquoi `_search`

Jarvi fait tourner un moteur de recherche plein texte sur vos données. Les noms, emails, numéros de téléphone, intitulés de poste, compétences et valeurs de champs personnalisés sont indexés, et _search interroge ces index. Filtrer avec _ilike "%terme%" fait tout autre chose : PostgreSQL parcourt chaque enregistrement de votre espace et le compare au motif. Sur une base de production, un seul _ilike sur un nom de famille a été mesuré à 2,4 secondes ; la même recherche via l’index répond en quelques dizaines de millisecondes.

_ilike, _like, _nilike et _nlike seront retirés de l’API publique le 31 juillet 2027.Sur les endpoints REST vous n’avez rien à faire — les filtres _ilike sont déjà traduits en recherche indexée pour vous, et la réponse indique quel filtre a remplacé le vôtre. Via l’endpoint /graphql, la requête vous appartient : elle peut viser n’importe quelle table, c’est donc à vous de la changer.

Utiliser l’opérateur

Un filtre prend le nom d’une cible de recherche et un terme :

{ "quickSearch": { "_search": "Dupont" } }

Combinez-en plusieurs avec _and, _or et _not, comme n’importe quel autre filtre :

{
  "_and": [
    { "globalSearch": { "_search": "javascript" } },
    { "isTalent": true },
    { "addedAt": { "_gte": "2026-01-01" } }
  ]
}

Syntaxe de requête

Le terme est lui-même une requête, pas une simple chaîne :

Syntaxe	Correspond à
`developer designer`	l’un ou l’autre mot
`developer AND designer`	les deux mots
`developer AND (java OR kotlin)`	les deux, avec l’une ou l’autre technologie
`NOT intern`	les enregistrements sans le mot
`"chef de projet"`	l’expression exacte, dans cet ordre
`java^10 python`	les deux mots, java pesant dix fois plus dans le classement

Des mots entiers, pas des fragments

Les noms sont indexés mot à mot, depuis le début : Dupo trouve Dupont, upont non. Il en va de même pour les noms d’entreprises et de projets. Les emails et les numéros de téléphone sont indexés autrement et acceptent un fragment n’importe où : un numéro se retrouve par ses derniers chiffres, un email par son domaine.

Les filtres, ressource par ressource

Tous les filtres ci-dessous s’appuient sur un index. Les autres colonnes et relations Hasura restent acceptées, mais elles ne sont ni optimisées ni garanties.

Profiles

Filter	Operators	What it matches
`quickSearch`	`_search`	Searches name, email, phone and short id — the target for simple person or name lookups.
`globalSearch`	`_search`	Searches name, headline, summary, position titles, company names, skills, educations and custom field values — the target for skill, role or technology searches.
`emailPhoneAndNameSearch`	`_search`	Searches name, email, phone and short id — the target for simple person or name lookups.
`fullPositionTitleSkillsAndTagsSearch`	`_search`	Searches name, headline, summary, position titles, company names, skills, educations and custom field values — the target for skill, role or technology searches.
`headline`	`_search`	Headline of the profile: the tagline shown under the name on LinkedIn, or the summary line extracted from the CV for profiles that have no LinkedIn page. Editable by hand only when the profile has no LinkedIn identifier.
`currentPositions.global`	`_search`	Full-text over title, description
`currentPositions.global.title`	`_search`	Job title held at this position, coming from LinkedIn, from a parsed CV or entered by hand. This is the main field behind job title search.
`currentPositions.global.description`	`_search`	What the person wrote about this role on their profile or CV. Often empty on imported records.
`emails`	`_search`	Email address of the profile or the company it belongs to. Normalized on save: lowercased, and when the input is not a clean address the first valid one found inside it is kept, the row being dropped if there is none.
`phones`	`_search`	Phone number search. Only the digits are compared, so a number matches whatever spacing, separators or country prefix it was typed with, and a suffix matches too.
`profiles.shortId`	`_search`	Short readable reference for the profile — what users type in quick search and what appears on generated documents. Assigned once at creation and never changes.
`historyEntries`	`_search`	Full-text over subject, message
`historyEntries.subject`	`_search`	Subject of the interaction: the email subject, the meeting title, or the job offer title. Empty for channels that carry no subject, such as LinkedIn messages, WhatsApp and SMS.
`historyEntries.message`	`_search`	Content of the interaction: the email body, the text of the LinkedIn, WhatsApp, SMS or Telegram message, or what the recruiter typed for a manual note.
`historyEntries.type`	`_eq`, `_in`, `_is_null`	What kind of interaction this entry records: emails, LinkedIn messages and InMails, connection invitations, SMS, WhatsApp, Telegram, phone calls, scheduled meetings, job applications, or a free note.
`historyEntries.subTypeId`	`_eq`, `_in`, `_is_null`	Optional sub-category of the interaction, defined by the group itself — naming its own kinds of notes or calls, for instance. Each sub-category belongs to a single parent type.
`historyEntries.occurredAt`	`_eq`, `_gte`, `_lte`, `_is_null`	Single date used to place the interaction on the timeline: the meeting start for a scheduled event, the creation date otherwise. Maintained by a trigger.
`historyEntries.userId`	`_eq`, `_in`, `_is_null`	Jarvi user this entry belongs to. Determines the group the entry is attached to, and is used as its sender by default.
`historyEntries.hasFiles`	`true`, `false`	The profile has at least one interaction carrying an attached file.
`historyEntries.hasRecording`	`true`, `false`	The profile has at least one interaction with a meeting recording.
`customField`	`_search`	Value of a custom field. Target one field by suffixing its id, as in “customField.<fieldId>”.
`location`	`_near`	Geographic search. Use range 0 for an exact city or region match, or a radius in meters.
`hasEmail`	`true`, `false`	The profile has at least one email address.
`hasEmailType`	`_eq`, `_in`, `_is_null`	The profile has an email address of this type. Takes the same values as the type of an email address.
`hasPhone`	`true`, `false`	The profile has at least one phone number.
`hasValidPhone`	`true`, `false`	The profile has at least one phone number not marked as invalid.
`hasLocation`	`true`, `false`	The profile has at least one location.
`hasPhoneType`	`_eq`, `_in`, `_is_null`	The profile has a phone number of this type. Takes the same values as the type of a phone number.
`hasWhatsApp`	`true`, `false`	The profile has a phone number reachable on WhatsApp.
`isTalent`	`true`, `false`	The profile is a talent — someone you recruit, followed in the ATS. A profile can be a talent and a contact at once.
`isContact`	`true`, `false`	The profile is a contact — a client or a hiring manager, followed in the CRM. A profile can be a contact and a talent at once.
`hasLinkedin`	`true`, `false`	The profile is matched to a LinkedIn account.
`hasResume`	`true`, `false`	The profile has at least one CV attached.
`isConnectedOnLinkedinWithMe`	`true`, `false`	The calling user is connected with this profile on LinkedIn, as recorded in the profile’s history.
`hasReplied`	`true`, `false`	The profile has answered at least once, on any channel.
`hasScheduledMessage`	`true`, `false`	A message or a connection invitation is still waiting to be sent to this profile.
`hasUnreadMessagesNotifications`	`true`, `false`	The profile has an unread message notification.
`firstPositionStartDate`	`_eq`, `_gte`, `_lte`, `_is_null`	Start date of the earliest position on the profile — the usual way to reason about years of experience.
`lastPositionStartDate`	`_eq`, `_gte`, `_lte`, `_is_null`	Start date of the current position, that is the one with no end date.
`currentPositions.companyId`	`_eq`, `_in`, `_is_null`	The profile currently works at this company. Takes the identifier of the LinkedIn company page, not a Jarvi company id.
`projectId`	`_eq`, `_in`, `_is_null`	The profile is associated with this project.
`statusId`	`_eq`, `_in`, `_is_null`	Status of the profile. A parent status also matches profiles carrying any of its children, and the search is narrowed by the project or entity given in the context. Pass “other” for profiles whose status has no parent.
`hasProject`	`true`, `false`	The profile is associated with at least one project.
`projectCount`	`_eq`, `_gte`, `_lte`, `_is_null`	How many distinct projects the profile is associated with.
`emptyCustomFields`	`_all`, `_in`	Custom field UUIDs. _all requires every field, _in requires any of them.
`filledCustomFields`	`_all`, `_in`	Custom field UUIDs. _all requires every field, _in requires any of them.
`isInCrm`	`true`, `false`	The profile is followed in the CRM, meaning it is a contact. Same as isContact.
`isInAts`	`true`, `false`	The profile is followed in the ATS, meaning it is a talent. Same as isTalent.
`hasAttachedFiles`	`true`, `false`	The profile has at least one attached file.
`isOpenToNewOpportunities`	`true`, `false`	The profile is flagged as open to new opportunities.
`isHiring`	`true`, `false`	The profile is flagged as hiring.
`id`	`_eq`, `_in`, `_is_null`	Jarvi identifier of the profile, used in the public API and in app URLs.
`createdAt`	`_eq`, `_gte`, `_lte`, `_is_null`	Technical timestamp of the row creation. For the date the profile entered the base, use addedAt (added_at) instead.
`addedAt`	`_eq`, `_gte`, `_lte`, `_is_null`	Date the profile entered the group’s base — what recruiters read as “added on”. Set automatically when the profile is created or restored, unless the caller provides its own value, which is what imports do to keep the original date.
`externalId`	`_eq`, `_in`	Identifier of the profile in the client’s own system — their ATS, their CRM, or whatever calls the public API. Never a Jarvi identifier. Unique per group, and used to recognise a profile that is sent again.
`publicIdentifier`	`_eq`, `_in`	LinkedIn public slug, the part after linkedin.com/in/. Unique per group. The person can change it on LinkedIn, which makes it the least reliable of the LinkedIn identifiers.
`linkedinId`	`_eq`, `_in`	Numeric LinkedIn identifier of the person. Unique per group. Not to be confused with memberId (member_id), which holds the obfuscated “ACoAA…” identifier, nor with publicIdentifier (public_identifier), which holds the URL slug.
`primaryLocale`	`_eq`, `_in`, `_is_null`	Main language of the profile itself — the language its LinkedIn page or CV is written in. Different from the profiles_locales table, which lists the languages the person speaks and how well.
`auditLogs.createdAt`	`_eq`, `_gte`, `_lte`, `_is_null`	When the tracked change happened. This is the reference date for incremental synchronisation: fetching everything modified since a given date goes through this column.

Companies

Filter	Operators	What it matches
`companySearch`	`_search`	Searches company name and description, LinkedIn tagline and description, legal activity and NAF code, and custom fields — the target for keyword or industry searches.
`companyQuickSearch`	`_search`	Searches company name, email and phone — for name lookups only.
`companies.global`	`_search`	Company name and description.
`location`	`_near`	Geographic search. Use range 0 for an exact city or region match, or a radius in meters.
`customField`	`_search`	Value of a custom field. Target one field by suffixing its id, as in “customField.<fieldId>”.
`hasLinkedinCompanyId`	`true`, `false`	The company is matched to a LinkedIn company page.
`hasContacts`	`true`, `false`	The company has at least one profile associated as a contact.
`hasProject`	`true`, `false`	The company has at least one project.
`isParentCompany`	`true`, `false`	The company has at least one subsidiary.
`isSubsidiary`	`true`, `false`	The company has a parent company.
`isProtected`	`true`, `false`	The company is under protection, meaning its protection date is still in the future.
`statusId`	`_eq`, `_in`, `_is_null`	Status of the company. A parent status also matches companies carrying any of its children.
`filledCustomFields`	`_all`, `_in`	Custom field UUIDs. _all requires every field, _in requires any of them.
`emptyCustomFields`	`_all`, `_in`	Custom field UUIDs. _all requires every field, _in requires any of them.
`id`	`_eq`, `_in`, `_is_null`	Jarvi identifier of the company, used in the public API and in app URLs.
`createdAt`	`_eq`, `_gte`, `_lte`, `_is_null`	Date the company record was created. Data imports replace it with the creation date coming from the source ATS or CRM.
`externalId`	`_eq`, `_in`	Identifier of the company in the client’s own system. Unique per group, and used as the matching key by imports and by the public API.
`linkedinCompanyId`	`_eq`, `_in`	Link to the LinkedIn page data of this company. That data is shared across all groups, and a given LinkedIn page can be attached to at most one company per group.
`parentId`	`_eq`, `_in`, `_is_null`	Parent company when this record is a subsidiary. Used as a single level: searching on a parent can include its direct subsidiaries, but not their own.
`auditLogs.createdAt`	`_eq`, `_gte`, `_lte`, `_is_null`	When the tracked change happened. This is the reference date for incremental synchronisation: fetching everything modified since a given date goes through this column.

Projects

Filter	Operators	What it matches
`projectNameSearch`	`_search`	Searches project name, its company name and its reference id — the target for project lookups.
`customField`	`_search`	Value of a custom field. Target one field by suffixing its id, as in “customField.<fieldId>”.
`hasJobBoardEnabled`	`true`, `false`	The project is published on at least one job board.
`hasLinkedinId`	`true`, `false`	The project is matched to a LinkedIn job posting.
`filledCustomFields`	`_all`, `_in`	Custom field UUIDs. _all requires every field, _in requires any of them.
`emptyCustomFields`	`_all`, `_in`	Custom field UUIDs. _all requires every field, _in requires any of them.
`statusId`	`_eq`, `_in`, `_is_null`	Status of the project. A parent status also matches projects carrying any of its children.
`id`	`_eq`, `_in`, `_is_null`	Jarvi identifier of the project, used in the public API and in app URLs.
`createdAt`	`_eq`, `_gte`, `_lte`, `_is_null`	Date the project was created. Data imports replace it with the creation date coming from the source ATS or CRM.
`externalId`	`_eq`, `_in`	Identifier of the project in the client’s own system. Unique per group, and used as the matching key by imports and by the public API.
`referenceId`	`_eq`, `_in`	The project reference id is used to map external job applications on website or jobboard to the project.
`companyId`	`_eq`, `_in`, `_is_null`	Client company the project is for. Optional — a project can exist without a company, and deleting the company clears the link rather than the project.
`auditLogs.createdAt`	`_eq`, `_gte`, `_lte`, `_is_null`	When the tracked change happened. This is the reference date for incremental synchronisation: fetching everything modified since a given date goes through this column.

Ces listes sont générées depuis le moteur de recherche lui-même : elles correspondent toujours à ce que l’API accepte. Les mêmes définitions sont servies en JSON Schema par GET /openapi, sous ProfilesFilters, CompaniesFilters et ProjectsFilters.

Trier

Le paramètre orderBy de chaque endpoint REST associe un champ et un sens :

?orderBy={"addedAt":"desc"}

Les champs acceptés par une ressource sont listés dans la documentation du paramètre, sur la page de l’endpoint. Trier sur autre chose retombe sur l’ordre par défaut plutôt que d’échouer.

Chercher via GraphQL

La query search renvoie des identifiants et un total, que vous chargez ensuite en un second appel. Ce découpage en deux temps est ce qui garde une recherche rapide, quelle que soit la taille des fiches :

query FindDevelopers {
  search(
    resource: "profiles"
    filters: { globalSearch: { _search: "javascript AND react" }, isTalent: true }
    sort: { field: "addedAt", order: "DESC" }
    limit: 50
  ) {
    ids
    total
  }
}

query LoadThem($ids: [uuid!]!) {
  profiles(where: { id: { _in: $ids } }) {
    id
    firstName
    lastName
  }
}

resource vaut profiles, companies ou projects. filters prend le même objet que le paramètre where du REST.

Quitter `_ilike`

Au lieu de	Écrivez
`{"lastName": {"_ilike": "%Dupont%"}}`	`{"quickSearch": {"_search": "Dupont"}}`
`{"emailAddresses": {"email": {"_ilike": "%@acme.com"}}}`	`{"emails": {"_search": "acme.com"}}`
`{"headline": {"_ilike": "%developer%"}}`	`{"globalSearch": {"_search": "developer"}}`
`{"name": {"_ilike": "%Acme%"}}` (entreprises)	`{"companyQuickSearch": {"_search": "Acme"}}`
`{"externalId": {"_ilike": "ABC-123"}}`	`{"externalId": {"_eq": "ABC-123"}}`

Un identifiant ou une URL n’est jamais une recherche : utilisez _eq, qui passe par son propre index.

Il vous manque quelque chose, ou un filtre que vous attendiez n’y est pas ? Écrivez-nous — la liste ci-dessus est pilotée par le moteur, y ajouter un filtre est un petit changement.

Général

Cas d'usages

Référence de l'API REST

Référence de l'API GraphQL

Pourquoi `_search`

Utiliser l’opérateur

Syntaxe de requête

Des mots entiers, pas des fragments

Les filtres, ressource par ressource

Profiles

Companies

Projects

Trier

Chercher via GraphQL

Quitter `_ilike`

​Pourquoi _search

​Utiliser l’opérateur

​Syntaxe de requête

​Des mots entiers, pas des fragments

​Les filtres, ressource par ressource

​Profiles

​Companies

​Projects

​Trier

​Chercher via GraphQL

​Quitter _ilike

Pourquoi `_search`

Utiliser l’opérateur

Syntaxe de requête

Des mots entiers, pas des fragments

Les filtres, ressource par ressource

Profiles

Companies

Projects

Trier

Chercher via GraphQL

Quitter `_ilike`