Eve Retriever ore-hold watcher (runs on Goliath)

Pings me on my phone (ntfy) and with an on-screen toast (Goliath) when my Retriever's ore hold is full, so I stop AFK-mining and unload/compress.

This runs entirely on Goliath (the Windows gaming PC) — Rocky isn't involved, it just reuses the same ntfy push app already on my phone.

Why not just read Eve's log files?

Eve writes (combat) / (notify) lines to Documents\EVE\logs\Gamelogs, but there is no reliable log line for "ore hold full" — mining-laser auto-deactivation is a client-side event the established mining bots detect by reading the in-game UI, not the logs. So this watcher reads the Ore Hold inventory window (OCR mode), or runs a yield-based countdown (timer mode).

Pick a mode

	OCR mode	Timer mode
Accuracy	Reads real fill %	Estimate from yield
Setup	Tesseract + one-time --snip	Just enter your m³/min
Fragility	Breaks if you move/close the Ore Hold window or change UI scale	Immune to UI; misses early rock depletion
In-game requirement	Keep the Ore Hold window open in a fixed spot	Press Enter when lasers go hot

Start with timer mode to prove delivery in 2 minutes, switch to OCR for hands-off.

Install (Goliath + Adam's PC)

Prereqs once: git (winget install --id Git.Git -e) and Python 3 (python.org installer, keep "Add to PATH" checked — includes tkinter).

Paste this into PowerShell once — that's the whole install, forever:

irm https://git.armoredarmadillo.com/brockdarnold/eve-watcher/raw/branch/main/install.ps1 | iex

It clones the repo to %USERPROFILE%\eve-watcher, starts the watchers, and registers a Windows Scheduled Task that re-runs update.py at every logon and once a day. That task does git pull + dep refresh and restarts the watchers only when something changed — so you never install or update again, and it never interrupts an active session for nothing. The shared Discord webhook is prefilled; your config.ini (webhook + OCR snip) is gitignored, so auto-updates never overwrite it.

The one remaining manual step (one-time, per PC) is the OCR snip for hold alerts — it's a GUI action so it can't be scripted:

cd ~\eve-watcher ; python eve_orehold_watcher.py --snip

(Combat + chat watchers need no snip — they're fully automatic.)

Manual config (optional)

config.ini is created for you, but you can edit it:

• [ntfy] topic → pick a private string (e.g. eve-retriever-7h2k9). On your phone, open the ntfy app → Add subscription → enter that same topic.
• For timer mode: set [watcher] mode = timer and [timer] yield_m3_per_min to your effective yield (Retriever ≈ 1,000–1,600 m³/min depending on fit/skills).
• For OCR mode: install Tesseract (winget install --id UB-Mannheim.TesseractOCR), then run the region picker (below).

Test delivery

python eve_orehold_watcher.py --test

You should get a phone push and a Windows toast. If the toast fails, check winotify installed; if the push fails, check the topic matches your phone subscription.

OCR calibration (one time)

In-game, open your Ore Hold so the 12,345 / 22,000 m³ text is visible and parked in a fixed spot. Then:

python eve_orehold_watcher.py --snip

Drag a tight box around just the current / capacity m³ numbers; it saves the region to config.ini. (Esc cancels.)

Run it

python eve_orehold_watcher.py

• OCR mode: leave it running; it re-arms after you unload (hold drops below reset_pct).
• Timer mode: press Enter each time you start a fresh hold.

Autostart (optional)

Task Scheduler → Create Task → Trigger At log on → Action: pythonw.exe with argument <full path>\eve_orehold_watcher.py → Start in: this folder. pythonw runs it without a console window.

Tuning

All in [watcher] of config.ini: alert_pct (default 95), poll_secs (10), reset_pct (50), cooldown_secs (120).

Notify a fleetmate (e.g. Adam) → Discord

"Ping Adam to compress when his hold is nearly full" can't be done from ESI — cargo/ore-hold fill is not exposed by any API. The detection has to run on Adam's own PC. So:

1. Adam installs this watcher on his machine (setup.ps1 / pip install -r requirements.txt).
2. In his config.ini set his ship's hold (Venture ore/gas hold = 5,000 m³), alert_pct = 90, and a [discord] webhook (a webhook for your shared channel) with mention = <@his-id>.
3. OCR mode: --snip his Venture's Ore Hold window. Timer mode: set his yield_m3_per_min.
4. When his hold passes 90%, the watcher posts "compress now" to your Discord channel (and/or his phone if he also sets [ntfy] topic).

• The watcher fans out to ntfy (phone) + Windows toast + Discord — set any/all.

Combat watcher — rats spawned / being tackled (eve_combat_watcher.py)

EVE has no API for in-space NPCs or being tackled — but it logs combat to disk. This companion tails Documents/EVE/logs/Gamelogs and alerts (ntfy + toast + Discord) on:

• Incoming damage ("rats?") — (combat) … from … lines.
• ⚠ Tackled / EWAR'd — (notify) warp-scramble/disrupt/jam/web/neut — top priority, no cooldown (if you can't warp, you need to know instantly).

Run: python eve_combat_watcher.py (--test fires one alert). Reads the same config.ini ([combat] section). Reading the log is legit — it never touches the game. Pair with in-game Drone Settings = Aggressive so drones auto-engage rats.

Other holds (e.g. Pioneer gas hold)

This watcher is hold-agnostic — it works for any "current / capacity m³" readout.

• Pioneer gas/specialized hold = 8,000 m³. For timer mode, set [timer] ore_hold_m3 = 8000 and your gas yield (Gas Cloud Scoop I = 10 m³ / 30 s = 20 m³/min per scoop; ×2 with Gas Cloud Harvesting II → ~40 m³/min). For OCR mode, just --snip the Pioneer's Gas Hold window instead of the ore hold.
• Keep separate config.ini copies (one per ship) if you switch between the Retriever and the Pioneer.