Company Job Openings

{  "type": "object",  "additionalProperties": false,  "properties": {    "boardSource": {      "type": "string",      "enum": [        "targets",        "ai-data",        "registry",        "discover"      ],      "title": "Boards to collect",      "description": "The list below is one run against named employers. AI data & labeling companies is a maintained list of the data-annotation and model-training vendors (Surge AI, Appen, Cohere, xAI, Encord, …). The registry is every board discovery has ever found — tens of thousands, and where the volume is. Discovery sweeps the crawl index for new ones first, which adds an hour or so to the front of the run.",      "default": "targets"    },    "targets": {      "type": "array",      "items": {        "type": "string"      },      "maxItems": 1000,      "title": "Boards",      "description": "One entry per employer, as `board:token` or a pasted careers URL. Boards: greenhouse, lever, ashby, smartrecruiters, personio, workday, workable, breezy, recruitee, bamboohr, teamtailor (a Teamtailor careers site on its own domain is named by host, e.g. `teamtailor:careers.example.com`). Ignored unless the source above is set to this list.",      "default": [        "greenhouse:stripe",        "lever:spotify",        "ashby:ramp",        "smartrecruiters:smartrecruiters",        "personio:personio"      ],      "examples": [        "greenhouse:stripe, https://jobs.lever.co/spotify"      ]    },    "keywords": {      "type": "array",      "items": {        "type": "string"      },      "maxItems": 1000,      "title": "Title keywords",      "description": "Keeps only postings whose title contains one of these. Empty keeps everything the board publishes.",      "default": [],      "examples": [        "engineer, data, scraping"      ]    },    "remoteOnly": {      "type": "boolean",      "title": "Remote roles only",      "description": "Only keeps postings the board explicitly marks remote. Boards that publish no remote flag contribute nothing under this.",      "default": false    },    "postedWithinDays": {      "type": "number",      "minimum": 0,      "maximum": 3650,      "title": "Posted within (days)",      "description": "Drop postings older than this. 0 keeps every open role. A low value turns the run into a \"who started hiring this week\" feed.",      "default": 0    },    "maxJobsPerBoard": {      "type": "number",      "minimum": 1,      "maximum": 100000,      "title": "Max jobs per board",      "description": "A safety stop, not a target. Set above the largest board you expect: a board that hits it is truncated, and the run says so by name.",      "default": 25000    },    "maxBoards": {      "type": "number",      "minimum": 0,      "maximum": 1000000,      "title": "Max boards per run",      "description": "0 collects every board the source offers. Set it to cap a first registry run, which otherwise reads tens of thousands of boards.",      "default": 0    },    "discoverSources": {      "type": "array",      "items": {        "type": "string",        "enum": [          "commoncrawl",          "wayback"        ]      },      "uniqueItems": true,      "title": "Indexes to sweep",      "description": "Common Crawl is a periodic snapshot and answers in minutes. Wayback is twenty years of the same hosts and holds an order of magnitude more — 367 pages of Lever boards against Common Crawl’s one — but takes about an hour. Lever is only reachable through Wayback and older crawl editions, since it disallows crawling now.",      "default": [        "commoncrawl"      ]    },    "discoverCrawls": {      "type": "number",      "minimum": 1,      "maximum": 30,      "title": "Crawl editions to sweep",      "description": "Each edition is a partial sample of the web, so boards keep accumulating as more are unioned. Reaching back several years is also the only way Lever boards are found at all — it disallows crawling now, so only older editions hold them.",      "default": 4    },    "retryFailed": {      "type": "boolean",      "title": "Only retry boards that errored",      "description": "Reads only the registry boards whose last read ended in an error rather than in an answer — the cheap pass after a big one, since a rate-limited board is unread rather than empty.",      "default": false    },    "skipDeadAfter": {      "type": "number",      "minimum": 0,      "maximum": 100,      "title": "Skip boards empty this many runs",      "description": "Registry boards that keep coming back empty are employers who closed the board or changed ATS. Skipping them is what keeps a full sweep from spending most of its requests on the dead. 0 collects everything on file.",      "default": 3    },    "includeDescription": {      "type": "boolean",      "title": "Store job descriptions",      "description": "Off by default: one large employer's descriptions run to megabytes of repeated boilerplate. On, each is capped.",      "default": false    },    "maxDescriptionChars": {      "type": "number",      "minimum": 200,      "maximum": 100000,      "title": "Description cap (characters)",      "default": 5000    },    "concurrency": {      "type": "number",      "minimum": 1,      "maximum": 50,      "title": "Concurrent boards",      "description": "Boards are independent, and these are the employers’ own endpoints rather than an aggregator defending itself, so this can be raised further than a search scraper’s. A registry sweep of tens of thousands of boards wants the top of this range.",      "default": 8    },    "requestDelayMs": {      "type": "number",      "minimum": 0,      "maximum": 30000,      "title": "Delay between requests (ms)",      "description": "Only applies between pages of a paged board; a one-shot board makes a single request.",      "default": 500    }  }}