mirror of https://github.com/jackwener/OpenCLI.git synced 2026-09-14 18:25:42 +08:00

Files

T

jakevin 29b4869efd feat(indeed): add search and job adapters (US site) (#1298 )

* feat(indeed): add `search` and `job` adapters (US site)

Adds an Indeed adapter that fills the US job-search gap (alongside
existing 51job / boss-zhipin / linkedin coverage). Both commands run
through a real browser session because Indeed sits behind Cloudflare
and answers bare HTTP fetches with `403` + `cf-mitigated: challenge`.

## Commands

- `indeed search <query>` — keyword job search
  - args: `query`, `--location`, `--fromage`, `--sort`, `--start`, `--limit`
  - columns: `rank, id, title, company, location, salary, tags, url`
- `indeed job <jk>` (alias `detail`, `view`) — full job posting
  - args: `id` (positional, the 16-char hex `jk` from `search`)
  - columns: `id, title, company, location, salary, job_type, description, url`

## Listing↔detail id pairing

`search.id` is the Indeed `jk` (job key, 16-char lowercase hex). It feeds
directly into `indeed job <jk>`. Conforms to the listing↔detail id
pairing convention proposed in #1297.

## CF challenge handling

The adapter polls the result selectors for up to 15s after navigation,
giving the browser time to clear the Cloudflare interstitial. If the
challenge is still up after the wait, the adapter throws a
`CommandExecutionError` with a hint pointing the user at the connected
browser to clear it once. Subsequent calls reuse the warmed cookies via
`Strategy.COOKIE`, mirroring the v2ex / boss / linkedin patterns.

## Validation

`utils.js` keeps argument validation pure and unit-testable:

- `requireJobKey` rejects anything that isn't a 16-char lowercase hex
- `requireFromage` only accepts `1` / `3` / `7` / `14` (Indeed's enum)
- `requireSort` only accepts `relevance` / `date`
- `requireBoundedInt(limit, default=15, max=25)` — Indeed serves at most
  one page (10 jobs/page); ArgumentError on out-of-range, no silent
  clamping, per the typed-error feedback in #1289.

## Tests

18 unit tests in `clis/indeed/indeed.test.js` cover registration,
validators, URL builders, and DOM-card normalizers. Browser-driven
verification stays out of CI by design (CF challenge is interactive).

## Docs

- `docs/adapters/browser/indeed.md` — full adapter doc with prerequisite
  CF-challenge notes and listing↔detail id pairing callout.
- Sidebar entry + adapter index row.

* fix(indeed): tighten timeout fail-fast and runtime tests

* fix(indeed): align readiness with search parser

2026-05-04 20:52:16 +08:00

3.1 KiB

Raw Permalink Blame History

Indeed

Mode: 🍪 Browser (cookie) · Domain: www.indeed.com

Indeed sits behind Cloudflare and answers bare HTTP fetches with 403 and a cf-mitigated: challenge header, so the adapter drives a real browser session. DOM extraction happens against the rendered page.

Commands

Command	Description
`opencli indeed search <query>`	Keyword job search on the US site
`opencli indeed job <jk>` (alias `detail`, `view`)	Read the full job posting by `jk` (job key)

Usage Examples

# Quick start — first run will route through the browser session
opencli indeed search "rust developer" --limit 10

# Narrow to recent jobs in a location, sorted by date
opencli indeed search "site reliability engineer" \
  --location "Remote" --fromage 7 --sort date

# Read a job posting using the `id` surfaced by `search`
opencli indeed job dccc07ac5a6a3683

# JSON output
opencli indeed search "data engineer" -f json

Output Columns

Command	Columns
`search`	`rank, id, title, company, location, salary, tags, url`
`job`	`id, title, company, location, salary, job_type, description, url`

id is the Indeed job key (jk) — a 16-character lowercase hex identifier. Pipe it into opencli indeed job <jk> to drill into the full posting (per the listing↔detail ID pairing convention).

Args

`search`

Arg	Type	Default	Notes
`query` (positional, required)	string	—	Job title / skill / company
`--location`	string	(none)	E.g. `"Remote"`, `"New York, NY"`
`--fromage`	string	(none)	Recency filter, days back: `1` / `3` / `7` / `14`
`--sort`	string	`relevance`	`relevance` or `date`
`--start`	int	`0`	Pagination offset (multiple of 10, 0-based)
`--limit`	int	`15`	Max rows to return (1–25, capped to one page)

`job`

Arg	Type	Default	Notes
`id` (positional, required)	string	—	Job key (16-char lowercase hex from `search`)

Prerequisites

Indeed protects the site with Cloudflare. The first run in a fresh browser session may surface the Just a moment… interstitial, in which case the adapter throws:

Indeed served a Cloudflare challenge page
hint: Open https://www.indeed.com in the connected browser and clear the
challenge, then retry.

Open the connected browser, complete the human check on indeed.com once, and the cookies will carry forward into subsequent adapter calls (Strategy.COOKIE).

Limitations

US site (www.indeed.com) only. Indeed runs region-specific subdomains (uk.indeed.com, de.indeed.com, ...) — adding them is mostly a question of swapping the origin and re-checking the DOM selectors.
Salary parsing pulls the visible text without normalizing currencies or ranges. Pipe through your own normalizer if you need structured values.
Tag column is a ·-joined free-text list (job type, schedule, etc.); Indeed's metadata pills change wording per A/B bucket so don't expect a fixed taxonomy.

3.1 KiB Raw Permalink Blame History Unescape Escape