puget-sound-salmon-report

Scrape WDFW Puget Sound creel data into salmon CPUE reports: daily digest, weekly top-launch email with chart, and proactive hot-bite alerts.

Calvin Tsai@calvin-tsai

Install

openclaw skills install @calvin-tsai/puget-sound-salmon-report

Puget Sound Salmon Creel Report

Turn WDFW's Puget Sound creel reports into actionable salmon fishing intel: per-area catch-per-angler (CPUE), trends, the best boat launches this week, a chart, an emailable digest, and optional proactive "hot bite" alerts. Pick any of the 5 Pacific salmon and any marine areas.

When to use

Use when a user wants to track recreational salmon catch/effort for Puget Sound marine areas — "how's the coho bite in Area 10", "email me the best launches Friday", "alert me when fishing spikes". The quantitative core is deterministic Python (trustworthy numbers); there is no LLM in the data path.

Setup

bash

pip install --user matplotlib   # only needed for the weekly chart; everything else is stdlib

The script is scripts/creel_report.py. Copy templates/config.json next to it (or pass --config PATH) to choose species/areas/thresholds. CLI flags override the config file.

Configure (`config.json`)

areas — marine area codes as shown on WDFW (e.g. "8-2","9","10","11").
species — any of chinook, coho, chum, pink, sockeye; multiple are combined into one CPUE.
trailing_days — days re-scraped each run (WDFW revises data after QA/QC).
top_launches, min_week_anglers — weekly launch ranking size + sample floor.
until — "YYYY-MM-DD"; --email stops on/after this date (for time-boxed seasons).
alerts — proactive hot-bite watcher (see below).

Usage

bash

python3 creel_report.py                       # daily per-area digest -> stdout
python3 creel_report.py --weekly              # weekly top-launch report + chart -> stdout
python3 creel_report.py --weekly --email      # HTML email with embedded chart
python3 creel_report.py --watch               # proactive hot-bite alert, or "NO_ALERT"
python3 creel_report.py --species chum,coho --areas 9,10 --weekly
python3 creel_report.py --no-fetch ...        # use stored data, skip scraping

Scheduling

Frequency = how often you schedule the mode. Any scheduler works (cron, launchd, or an OpenClaw cron job). Example crontab:

cron

30 8 * * *   python3 creel_report.py                                  # daily digest
0  9 * * 5   python3 creel_report.py --weekly --email --until 2026-09-25
0  12,17 * * * python3 creel_report.py --watch                        # hot-bite checks

Email (optional)

--email reads SMTP creds from ~/.openclaw/creel_email.json (never commit it):

json

{
  "smtp_host": "smtp.gmail.com", "smtp_port": 465,
  "smtp_user": "you@gmail.com", "smtp_pass": "APP_PASSWORD",
  "from": "you@gmail.com",
  "to": ["you@example.com"],
  "bcc": ["friend@example.com"],
  "subscribers_url": "",        // published Google Sheet CSV (Google Form signups) -> merged into bcc
  "unsubscribe_url": "",        // published CSV of opt-outs -> removed from every send
  "subscribe_form_url": "", "unsubscribe_form_url": ""   // footer links
}

For Gmail use an App Password. to/cc/bcc accept a string, comma-string, or list. A published Google Form → Sheet gives a self-serve signup UI: put the CSV link in subscribers_url and every email-looking cell is merged into BCC (de-duped); unsubscribe_url removes opt-outs (your own to is always kept). The chart embeds via a cid: attachment so it renders in Outlook.

Proactive hot-bite alerts (`--watch`)

--watch flags when the latest day's CPUE in an area spikes vs its recent baseline, else prints NO_ALERT. Toggle with alerts.enabled. It's self-limiting — a state file (memory/creel/alert_state.json) plus a cooldown means an area won't re-alert until the cooldown passes or the bite gets meaningfully hotter.

json

"alerts": {
  "enabled": true, "spike_mult": 1.8, "min_anglers": 30,
  "min_cpue": 0.4, "cooldown_days": 2, "min_baseline_days": 2
}

Wire delivery so a bare NO_ALERT posts nothing. This pairs naturally with an agent runtime (e.g. OpenClaw heartbeat/cron) that can message you proactively; with plain cron, gate on the output: out=$(python3 creel_report.py --watch); [ "$out" = NO_ALERT ] || notify "$out".

Data / output

Written under memory/creel/ relative to the script: daily.json (ramp-level history, all species), cpue_areas.png, latest_digest.md, weekly_report.md, alert_state.json.

Notes / limitations

WDFW publishes no public API for the marine areas; this scrapes the server-rendered HTML table, so a major page redesign could require updating the parser.
Same-day data is partial until WDFW fills it in overnight; the digest anchors on the latest substantially-sampled day.
Quotas/closures are not tracked (Coho isn't quota-managed); check WDFW regs before fishing.