molly-sourcing

Sourcing

Triggered by the daily scheduled (cron) run once a role reaches Sourcing status (or when a recruiter asks Molly to run sourcing). For every role in Sourcing, Molly first folds in the talent team's feedback from the previous cycle — left as notes on yesterday's GEM candidates — updates the Profile Spec accordingly, then sources fresh LinkedIn candidates via Exa (MOLLY_DAILY_SOURCING_TARGET, default 25), submits each into that role's GEM project, and posts one daily digest to the sourcing channel.

The notes feedback loop is what makes sourcing sharpen day over day — it's the lightweight, ongoing analog of calibration. The team drops quick notes on the candidates they liked or didn't (right in GEM, where they're already reviewing), and the next run honors that signal. The digest closes the loop by reminding the team to leave feedback within the feedback window (MOLLY_FEEDBACK_WINDOW_HOURS, default 24h), before the next run.

GEM dedupes on LinkedIn handle, so re-running daily is safe — candidates already in GEM are matched, not duplicated.

Inputs

• Env: MOLLY_NOTION_API_KEY, MOLLY_EXA_API_KEY, MOLLY_GEM_API_KEY, MOLLY_SLACK_BOT_TOKEN, SOURCING_SLACK_CHANNEL_ID, MOLLY_TRACKER_DB_ID
• Molly Tracker database: $MOLLY_TRACKER_DB_ID
• Each Sourcing role row must have a GEM Project ID (the GEM project candidates are filed into). Skip and flag any role missing it.

Steps

1. Find roles in Sourcing

Query the Molly Tracker, filtering on the Molly Status status property:

curl -s -X POST "https://api.notion.com/v1/databases/$MOLLY_TRACKER_DB_ID/query" \
  -H "Authorization: Bearer $MOLLY_NOTION_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Notion-Version: 2022-06-28" \
  -d '{"filter": {"property": "Molly Status", "status": {"equals": "Sourcing"}}}'

For each returned row, read: Role name (title), JD Link (url), Profile Spec Link (url), GEM Project ID (rich text). The JD/Spec page IDs are the last 32 hex characters of each URL.