linkedin-sourcer（领英来源）

LinkedIn Sourcer

Source candidates from LinkedIn, analyze their profiles, and evaluate fit against role requirements using the linkedin_scraper library (v3.0+, Playwright-based, async).

Prerequisites

Ensure dependencies are installed before any scraping:

pip install linkedin-scraper
playwright install chromium

An authenticated session file (session.json) is required. If one does not exist, create one:

Programmatic login (using credentials):

python3 scripts/create_session.py --email USER@EXAMPLE.COM --password PASS

Or via environment variables:

export LINKEDIN_EMAIL=user@example.com
export LINKEDIN_PASSWORD=mypassword
python3 scripts/create_session.py

Manual login (opens a browser window — use when programmatic login fails due to CAPTCHA/2FA):

python3 scripts/create_session.py

The session file is reusable until LinkedIn expires it. See references/linkedin_scraper_api.md for browser configuration options.

Workflow Decision Tree

Determine the task type:

1. "Scrape this profile / these profiles" → Profile Scraping
2. "Find candidates for this role" → Candidate Search
3. "Evaluate this candidate for this role" → Candidate Evaluation
4. "Compare these candidates" → Candidate Comparison

1. Profile Scraping

Run scripts/scrape_profile.py to extract structured profile data:

python3 scripts/scrape_profile.py "https://linkedin.com/in/username" --session session.json

For multiple profiles:

python3 scripts/scrape_profile.py URL1 URL2 URL3 --delay 2 --output profiles.json

Output is JSON with: name, headline, location, about, experiences, educations, skills.

For inline scraping within custom code, see references/linkedin_scraper_api.md → PersonScraper.

2. Candidate Search

Generate boolean search queries the user can paste into LinkedIn or Google to find candidates. See references/sourcing_workflows.md → Boolean Search String Patterns for templates and examples. Tailor the boolean string to the specific role requirements provided.

3. Candidate Evaluation

After scraping profile(s), evaluate fit against a job description:

1. Scrape the candidate's profile
2. Apply the scorecard template from references/sourcing_workflows.md → Candidate Scorecard Template
3. Rate each criterion (1-5) with notes based on the scraped data
4. Assign an overall fit rating: STRONG_FIT, GOOD_FIT, PARTIAL_FIT, or WEAK_FIT
5. Identify strengths, concerns, and key questions for outreach

Use the evaluation heuristics in references/sourcing_workflows.md → Evaluation Heuristics to guide ratings.

For quick single-candidate output, use the Candidate Summary Template instead.

4. Candidate Comparison

When evaluating multiple candidates for the same role:

1. Scrape all candidate profiles
2. Apply the comparison table from references/sourcing_workflows.md → Candidate Comparison Table
3. Rank candidates with rationale

Error Handling

• AuthenticationError → Session expired. Re-run scripts/create_session.py with credentials or manual login
• RateLimitError → Wait and retry. Increase --delay between requests
• ProfileNotFoundError → Profile is private or URL is invalid

See references/linkedin_scraper_api.md → Error Handling for try/except patterns.

Rate Limiting

Always use delays between requests (default 2s in scripts). For large batches, increase to 3-5s. Never scrape aggressively — respect LinkedIn's rate limits.