Apify (second-priority LinkedIn channel)

The linkedin-apify channel takes over the main LinkedIn search after the Bright Data free quota is used up (or it fails), fetching jobs through an actor on Apify.

Account and token

1. Sign up for an Apify account (the Free plan includes monthly credit), and copy the Personal API token from Console -> Settings -> API & Integrations.
2. Put it in .env: APIFY_TOKEN=... (.env is gitignored; do not commit it or paste it into chat or logs).
3. Without a token (or with an empty actor_id) the channel shows “Not configured”; the cascade simply skips it, produces no failure record, and does not affect the circuit breaker.
4. On the channels page of /settings, confirm linkedin-apify is enabled (enabled by default); the monthly budget monthly_budget_usd (default 10 USD) is bounded by the global paid_total_monthly_budget_usd.

Configuration (data/config/sources.yaml -> linkedin.apify)

Key	Meaning
enabled / monthly_budget_usd / throttle / max_results	Same as other channels; max_results is the maximum results taken per query
actor_id	(protected: edit the file by hand) The actor, username/actor-name or a 17-character ID; default valig/linkedin-jobs-scraper; empty = not configured
api_base_url	Apify API address
price_per_1k_usd / run_start_usd	For accounting: price per 1000 results, and the start fee charged once per run. Use the price for your plan under Pricing on the actor page
memory_mb / timeout_secs	Memory and timeout of one run; the synchronous endpoint caps at 300 seconds, and a timeout is recorded as a retryable failure
country_name	Location of city queries = <city>, <country name>
remote_locations	The location used by remote queries per search.yaml: remote_scope
date_posted	How the actor’s “date posted” filter is written: window <=24 hours / <=7 days / <=30 days / longer (empty = omit this key)
input_template	The actor input template; strings may use {keywords} {location} {date_posted} {limit} (when the whole string is "{limit}" it becomes a number); keys with empty values are omitted
remote_input	Keys layered onto the template by remote queries (default urlParam: f_WT=2)

input_template, remote_input and date_posted are not shown on the settings page (they are actor-specific data); edit the YAML directly.

How it works

• Each query = one synchronous run: POST /v2/acts/<actor>/run-sync-get-dataset-items (201 on success, 408 after 300 seconds, 402 when there is no credit). Keyword groups are merged as "A" OR "B" (over 200 characters it is split into several runs sharing the limit).
• Before triggering, the maximum number of results this run may request is computed from the remaining monthly budget (start fee deducted first), and sent to Apify as the hard caps maxItems/maxTotalChargeUsd; if the budget cannot cover one run plus one result it is budget_exceeded directly, and the same query goes to the next channel in the cascade.
• Accounting = start fee + results actually delivered x unit price; every result that was billed (duplicates and ones dropped by the local time window included) counts toward the channel’s monthly usage and the global paid cap.
• A run is sent only once, with no retry (a retry would bill twice); three consecutive failures open the breaker, and it tries again after 24 hours.
• external_id is LinkedIn’s numeric job id, the same as Bright Data, so the same job obtained through the two channels is merged into one.

Live smoke test

pnpm smoke:apify              # reads .env; one query, at most 10 results, about $0.005
pnpm exec tsx --env-file=/path/to/.env scripts/smoke-apify.ts --limit 5

It spends money on your Apify account; it writes nothing to the database, and usage is shown only in the output.