Runtime/feed URLs:

Public agenda page:

https://www.marseille-tourisme.com/vivez-marseille-blog/agenda/

Playlist JSON feed:

https://www.marseille-tourisme.com/api/render/website_v2/marseille-tourisme/playlist/78894/fr_FR/json

Public sitemap index:

https://www.marseille-tourisme.com/sitemap.xml/

Event detail pages:

Individual public event URLs returned by the playlist feed or discovered from sitemap FMA pages.

How it works:

• Starts from the configured source URL, usually the public agenda page.
• Finds the hawwwai-playlist element and uses its public playlist configuration.
• Requests paginated FMA agenda items from the playlist JSON endpoint with page query parameters.
• Scans the public sitemap for French agenda FMA detail pages that are not present in the playlist feed.
• Fetches event detail pages with bounded concurrency.
• Extracts the detail-page HwSheet description, price text, accepted public/clientele, hours, official images, ticket URL, venue/address, GPS coordinates and date instances.
• Stores multiple date ranges as event instances when the official feed exposes them.

Supported configuration:

max_items:

Maximum playlist items to inspect per sync. Default: 1500.

max_pages:

Maximum playlist pages to request. Default: 200.

include_sitemap_details:

Whether to import valid sitemap-only Marseille Tourisme FMA detail pages. Default: true.

max_sitemap_items:

Maximum sitemap detail pages to inspect per sync. Default: 1500, above the current complete catalog.

sitemap_url:

Optional sitemap index override. Default: https://www.marseille-tourisme.com/sitemap.xml/.

detail_concurrency:

Concurrent detail-page fetches, clamped from 1 to 20. Default: 8.

encoding:

Optional response decoding hint. Default: UTF-8.

Runtime/feed URLs:

Public agenda pages:

https://www.jds.fr/marseille/agenda/?page={page}

Event detail pages:

Individual JDS URLs ending in a stable numeric _A identifier.

How it works:

• Traverses 25-row agenda pages through their rel=next links.
• Fetches detail pages with bounded concurrency and parses Schema.org Event JSON-LD.
• Accepts only addressLocality Marseille, Marseille postcodes 13000/13001-13016, or an explicit Schema.org City named Marseille.
• Rejects nearby cities even though JDS includes them in its Marseille regional agenda.
• Imports full descriptions, dates and repeated date/hour rows, venue/address, GPS, ticket link, free status and largest available images.
• Exposes the deepest JDS agenda breadcrumb as jds_category for Admin Categories mapping.

Supported configuration:

max_items:

Maximum event detail pages to inspect per sync. Default: 250.

max_pages:

Maximum 25-row agenda pages to request. Default: 20.

detail_concurrency:

Concurrent event detail requests, clamped from 1 to 20. Default: 8.

locality:

Documented admission scope; this connector is intentionally fixed to Marseille.

Runtime/feed URLs:

Configured event feed:

The OpenAgenda v2 agenda events URL stored on the source row.

Event details:

OpenAgenda v2 /agendas/{agenda_uid}/events/{event_uid} endpoints.

Public event pages:

https://openagenda.com/fr/{agenda_slug}/events/{event_slug}

How it works:

• Authenticates with OPENAGENDA_ACCESS_KEY from the ignored .env file.
• Preserves repeated relative filters from the configured source URL.
• Traverses OpenAgenda cursor pages through repeated after[] parameters.
• Fetches event detail records with bounded concurrency.
• Uses French localized titles and full descriptions, with English/first-value fallback.
• Imports every official timing as an event instance and keeps the final timing as the event end.
• Selects the largest OpenAgenda CDN image variant first.
• Extracts registration links, conditions/prices, attendance mode and coordinates.
• Exposes source keywords when present, otherwise an explicit category inferred by Planet Mars for admin mapping.

Supported configuration:

agenda_uid:

OpenAgenda numeric agenda identifier. Inferred from the URL when omitted.

public_agenda_slug:

Slug used to build canonical public event URLs.

locale:

Preferred localized content language. Default: fr.

max_items:

Maximum events enriched per sync. Default: 250.

page_size:

Events requested per cursor page, clamped to 1-100. Default: 100.

max_pages:

Maximum cursor pages requested. Default: 10.

detail_concurrency:

Concurrent detail requests, clamped to 1-20. Default: 10.

Runtime/feed URLs:

Instagram profile:

The configured instagram.com account URL.

Instagram private API:

instagrapi-managed endpoints; exact URLs are not stored in Planet Mars config.

How it works:

• Loads non-secret strategy selection from config/instagram_connectors.json.
• Uses credentials or a persisted session from ignored .env/storage values.
• Fetches recent public account posts and complete carousel image sets.
• Reads the current profile grid once per sync, then sends only new or changed posts to image download, OCR and OpenAI extraction.
• Records successfully analyzed posts even when they contain no event, preventing repeated token spend on non-event posts.
• Runs enabled local PaddleOCR per slide, then sends caption, OCR and images to the source-selected OpenAI model.
• Imports only validation-ready future Marseille events through the normal deduplication and visibility pipeline.
• Caches extraction by account, post content, prompt version and model so unchanged posts do not spend OpenAI tokens again.

Supported configuration:

default_strategy:

Global Instagram strategy in config/instagram_connectors.json.

sources.<account>.strategy:

Per-account strategy override.

openai_model:

Multimodal extraction model selected per Instagram source.

instagram_options.post_limit:

Latest profile posts checked for new or changed content per synchronization, from 1 to 50. Default for new sources: 20.

instagram_account_type:

Venue, organizer, or mixed account behavior.

instagram_default_venue:

Verified fixed venue used only by venue accounts.

instagram_default_address:

Verified Marseille address used only by venue accounts.

Runtime/feed URLs:

Configured public page:

The source URL configured on the source row.

How it works:

• Downloads the configured public URL.
• Selects article/event-like HTML nodes.
• Builds raw event candidates from visible text and the first link.
• Relies on the shared extraction pipeline for dates and event fields.

Supported configuration:

selectors:

Optional CSS selectors to find event candidate nodes.

max_items:

Maximum candidate nodes to inspect. Default: 25.

encoding:

Optional response decoding hint. Default: UTF-8.

Runtime/feed URLs:

Configured public page:

The source URL configured on the source row.

How it works:

• Requires a configured public URL.
• Uses the same candidate selection and extraction flow as the generic website scraper.

Supported configuration:

selectors:

Optional CSS selectors to find event candidate nodes.

max_items:

Maximum candidate nodes to inspect. Default: 25.

Runtime/feed URLs:

Configured public page:

The source URL configured on the source row.

How it works:

• Requires a configured public URL.
• Uses the same candidate selection and extraction flow as the generic website scraper.

Supported configuration:

selectors:

Optional CSS selectors to find event candidate nodes.

max_items:

Maximum candidate nodes to inspect. Default: 25.

Runtime/feed URLs:

Configured public page:

The source URL configured on the source row.

How it works:

• Requires a configured public URL.
• Uses the same candidate selection and extraction flow as the generic website scraper.

Supported configuration:

selectors:

Optional CSS selectors to find event candidate nodes.

max_items:

Maximum candidate nodes to inspect. Default: 25.

Runtime/feed URLs:

Public Marseille listings:

https://ra.co/events/fr/marseille

GraphQL endpoint:

https://ra.co/graphql

Event details:

Resident Advisor event(id) GraphQL queries and https://ra.co/events/{event_id} pages.

How it works:

• Queries Resident Advisor area 338, which RA identifies as Marseille.
• Paginates upcoming listings over a configurable future horizon and deduplicates repeated listing-day rows by stable RA event ID.
• Fetches event details with bounded concurrency and falls back to listing data when one detail request fails.
• Keeps only events whose venue area is explicitly Marseille.
• Treats RA start/end values as Europe/Paris local times and stores the interval as an event instance.
• Imports paragraph-preserved descriptions, original RA flyer images, venue/address, lineup, price, minimum age and ticket link.
• Exposes only RA genres as source taxonomies and uses attending count as the popularity score.

Supported configuration:

area_id:

Resident Advisor numeric area identifier. Marseille: 338.

max_items:

Maximum unique RA events enriched per sync. Default: 250.

page_size:

Listing rows requested per page, clamped to 1-100. Default: 50.

max_pages:

Maximum listing pages requested. Default: 10.

horizon_days:

Future listing horizon from the current Marseille day. Default: 365.

detail_concurrency:

Concurrent event-detail queries, clamped to 1-20. Default: 8.

graphql_url:

Optional endpoint override. Default: https://ra.co/graphql.

Runtime/feed URLs:

Public city listings:

https://shotgun.live/fr/cities/aix-marseille

Parse API:

https://api.parse.bot/scraper/da664d42-1552-4c38-a1d9-1ca0909c41f6/{endpoint}

Event details:

Parse get_event_detail calls followed by canonical https://shotgun.live/fr/events/{slug} links.

How it works:

• Authenticates with PARSE_API_KEY from the ignored .env file; the secret is never stored on the source row.
• Uses get_events with city=aix-marseille/-/ and follows page/has_more pagination.
• Calls get_event_detail for address, coordinates, schedule, images, performers and ticket offers.
• In incremental mode, skips detail calls for recently known listings and periodically refreshes them after the configured window.
• Requires explicit Marseille locality, address or postcode evidence after detail enrichment.
• Imports canonical Shotgun links, event images, prices, performers and normalized event instances.
• Creates daily instances only when separate ticket offers explicitly name at least two event dates.
• Exposes Shotgun tags under shotgun_genre for Admin Categories mapping.
• Paces requests according to the configured Parse plan to avoid HTTP 429 responses and excess credit use.
• Parse.bot is an independent paid wrapper, not an official Shotgun API; source health therefore depends on both services.

Supported configuration:

city:

Parse/Shotgun city slug. Default: aix-marseille/-/.

locality:

Required physical locality after detail enrichment. Default: Marseille.

physical_locality_only:

Require the configured locality/postcodes instead of the full city feed. Default: true.

max_items:

Maximum unique listings enriched per synchronization. Default: 50.

max_pages:

Maximum get_events pages requested. Default: 10.

requests_per_minute:

Parse request pace matching the account plan. Free: 5; Hobby: 20; Developer: 100.

incremental_detail_skip:

Skip detail calls for already imported listings while still checking listing pages. Default: true.

incremental_refresh_days:

Refresh already imported listings after this many days so date/ticket/image changes are eventually captured. Default: 30.

api_base:

Optional Parse scraper endpoint override for a pinned or private API copy.

Runtime/feed URLs:

Public Marseille search:

https://www.helloasso.com/e/recherche/activites?place_city=Marseille&place_department=Bouches-du-Rh%C3%B4ne

Structured search proxy:

https://www.helloasso.com/algolia/1/indexes/*/queries

Event pages and ticketing:

Canonical https://www.helloasso.com/associations/{organization}/evenements/{event} URLs returned by the index.

Official directory alternative:

https://api.helloasso.com/v5/directory/forms (requires partner FormOpenDirectory privilege and is not used by this connector).

How it works:

• Reads the public page first and discovers its current public Algolia application ID, search key, proxy and production index dynamically.
• Does not require or persist a private HelloAsso credential.
• Requests only form_type=Event records whose end date is today or later and paginates the complete configured result set.
• Loads each canonical event page to read the event venue instead of trusting the association address exposed by the search index.
• Fails safely instead of reconciling a partial catalog when upstream results exceed max_items or max_pages.
• Requires explicit Marseille city/postcode evidence from the event page and rejects other or ambiguous locations.
• Imports exact UTC schedules, event venue/address, paragraph-preserved descriptions, full CDN banner/logo URLs, min/max price and participant popularity.
• Never treats the association headquarters coordinates in Algolia as event coordinates.
• Uses each canonical HelloAsso event URL as both source URL and Ticket/Billetterie destination.
• Exposes activity type, categories and public tags under helloasso_* taxonomies for Admin Categories mapping.
• The Explore Algolia contract is public but undocumented and may require connector maintenance if HelloAsso changes its frontend.

Supported configuration:

locality:

Required physical city. Default: Marseille.

department:

HelloAsso department filter. Default: Bouches-du-Rhone.

physical_locality_only:

Documents strict Marseille admission. Default: true.

max_items:

Maximum complete current/upcoming catalog size. Default: 1000.

page_size:

Records requested per search page, clamped to 1-100. Default: 100.

max_pages:

Maximum complete search pages. Default: 10.

detail_concurrency:

Concurrent canonical event-page validations, clamped to 1-20. Default: 10.

detail_requests_per_second:

Maximum paced canonical detail requests per second. Default: 5.

browser_impersonation:

curl-cffi browser profile used for the Cloudflare-protected public page. Default: safari17_0.

timeout_seconds:

Per-request timeout, clamped to 5-120 seconds. Default: 30.

algolia_proxy:

Optional structured-search proxy override.

index_name:

Optional index override. Default: prod_activities.

Runtime/feed URLs:

Configured public page:

The source URL configured on the source row.

How it works:

• Requires a configured public URL.
• Uses the same candidate selection and extraction flow as the generic website scraper.

Supported configuration:

selectors:

Optional CSS selectors to find event candidate nodes.

max_items:

Maximum candidate nodes to inspect. Default: 25.

Runtime/feed URLs:

Public event search:

https://tarpin-bien.com/recherche/?evenementCheck=1&pageEvenement={page}&show=evenement

WordPress event API:

https://tarpin-bien.com/wp-json/wp/v2/evenement

Event details:

Individual https://tarpin-bien.com/evenement/{slug}/ pages discovered from the upcoming index.

How it works:

• Paginates only the public current/upcoming event results instead of traversing the full historical WordPress collection.
• Fetches detail pages with bounded concurrency and reads their Schema.org Event subtypes.
• Treats the published event clock values as Europe/Paris wall time because the site labels them +0000 while rendering the same local clock time.
• Requires explicit Marseille locality or a Marseille 13000/13001-13016 postcode and rejects Aix, Cassis, Martigues and other regional listings.
• Imports paragraph-preserved descriptions, date instances, venue/address, coordinates, direct ticket links, prices and original images.
• Uses bounded WordPress REST batches to enrich categorie-evenement/tag-evenement taxonomies and original featured media.

Supported configuration:

max_items:

Maximum event detail pages to inspect per sync. Default: 250.

max_pages:

Maximum upcoming-result pages to request. Default: 20.

detail_concurrency:

Concurrent event-detail requests, clamped from 1 to 20. Default: 8.

locality:

Documented admission scope; this connector is intentionally fixed to Marseille.

physical_locality_only:

Documents strict physical-Marseille admission. Default: true.

Runtime/feed URLs:

Configured public page:

The source URL configured on the source row.

How it works:

• Downloads the configured public URL.
• Selects article/event-like HTML nodes.
• Builds raw event candidates from visible text and the first link.
• Relies on the shared extraction pipeline for dates and event fields.

Supported configuration:

selectors:

Optional CSS selectors to find event candidate nodes.

max_items:

Maximum candidate nodes to inspect. Default: 25.

encoding:

Optional response decoding hint. Default: UTF-8.

Runtime/feed URLs:

Configured public page:

The source URL configured on the source row.

How it works:

• Downloads the configured public URL.
• Selects article/event-like HTML nodes.
• Builds raw event candidates from visible text and the first link.
• Relies on the shared extraction pipeline for dates and event fields.

Supported configuration:

selectors:

Optional CSS selectors to find event candidate nodes.

max_items:

Maximum candidate nodes to inspect. Default: 25.

encoding:

Optional response decoding hint. Default: UTF-8.

Runtime/feed URLs:

Instagram profile:

The configured instagram.com account URL.

Instagram private API:

instagrapi-managed endpoints; exact URLs are not stored in Planet Mars config.

How it works:

• Loads non-secret strategy selection from config/instagram_connectors.json.
• Uses credentials or a persisted session from ignored .env/storage values.
• Fetches recent public account posts and complete carousel image sets.
• Reads the current profile grid once per sync, then sends only new or changed posts to image download, OCR and OpenAI extraction.
• Records successfully analyzed posts even when they contain no event, preventing repeated token spend on non-event posts.
• Runs enabled local PaddleOCR per slide, then sends caption, OCR and images to the source-selected OpenAI model.
• Imports only validation-ready future Marseille events through the normal deduplication and visibility pipeline.
• Caches extraction by account, post content, prompt version and model so unchanged posts do not spend OpenAI tokens again.

Supported configuration:

default_strategy:

Global Instagram strategy in config/instagram_connectors.json.

sources.<account>.strategy:

Per-account strategy override.

openai_model:

Multimodal extraction model selected per Instagram source.

instagram_options.post_limit:

Latest profile posts checked for new or changed content per synchronization, from 1 to 50. Default for new sources: 20.

instagram_account_type:

Venue, organizer, or mixed account behavior.

instagram_default_venue:

Verified fixed venue used only by venue accounts.

instagram_default_address:

Verified Marseille address used only by venue accounts.

Runtime/feed URLs:

Instagram profile:

The configured instagram.com account URL.

Instagram private API:

instagrapi-managed endpoints; exact URLs are not stored in Planet Mars config.

How it works:

• Loads non-secret strategy selection from config/instagram_connectors.json.
• Uses credentials or a persisted session from ignored .env/storage values.
• Fetches recent public account posts and complete carousel image sets.
• Reads the current profile grid once per sync, then sends only new or changed posts to image download, OCR and OpenAI extraction.
• Records successfully analyzed posts even when they contain no event, preventing repeated token spend on non-event posts.
• Runs enabled local PaddleOCR per slide, then sends caption, OCR and images to the source-selected OpenAI model.
• Imports only validation-ready future Marseille events through the normal deduplication and visibility pipeline.
• Caches extraction by account, post content, prompt version and model so unchanged posts do not spend OpenAI tokens again.

Supported configuration:

default_strategy:

Global Instagram strategy in config/instagram_connectors.json.

sources.<account>.strategy:

Per-account strategy override.

openai_model:

Multimodal extraction model selected per Instagram source.

instagram_options.post_limit:

Latest profile posts checked for new or changed content per synchronization, from 1 to 50. Default for new sources: 20.

instagram_account_type:

Venue, organizer, or mixed account behavior.

instagram_default_venue:

Verified fixed venue used only by venue accounts.

instagram_default_address:

Verified Marseille address used only by venue accounts.