ollama-cloud-usage

name	ollama-cloud-usage
description	Check your Ollama Cloud Pro/Max subscription usage — session and weekly limits, plan tier, and reset timers.
version	2.1.0
author	Rabil
license	MIT
metadata	{"hermes":{"tags":["ollama","cloud","usage","monitor"],"related_skills":[]}}

Ollama Cloud Usage

Check your Ollama Cloud Pro/Max subscription usage directly from Hermes.

Since Ollama Cloud has no public usage API (see ollama/ollama#12532), this skill scrapes the dashboard at https://ollama.com/settings using your authenticated session cookie.

When to Use

• User asks about their Ollama Cloud usage, limits, or remaining quota
• User wants to know their plan tier (Pro/Max) or when their session/weekly usage resets
• User asks "how much Ollama Cloud do I have left?"

Prerequisites

Requires the OLLAMA_CLOUD_COOKIE environment variable in your Hermes env config.

Getting your cookie

1. Open a browser and log in to https://ollama.com/settings
2. Open DevTools → Application (or Storage on Firefox) → Cookies
3. Copy all cookies for ollama.com as a single string (e.g. auth=xxx; other=yyy)
4. Add to your Hermes env config file (the .env file in the Hermes data directory):

   OLLAMA_CLOUD_COOKIE="auth=xxx; other=yyy"
   

5. Restart Hermes or run /reset if mid-session

If the cookie expires, repeat steps 2–4.

Cookie File Locations

The Python script checks these paths in order:

1. OLLAMA_CLOUD_COOKIE env var (if exported to subprocess)
2. /opt/data/.env (Hermes Docker container mount, primary)
3. ~/.hermes/.env (local bare-metal fallback)

How to Check Usage

IMPORTANT: Use terminal with python3 -c "..." — NOT execute_code. execute_code does NOT inherit Hermes env vars, so OLLAMA_CLOUD_COOKIE is always empty there. The terminal tool DOES inherit the environment and the cookie is found via os.environ.

python3 /opt/data/skills/ollama-cloud-usage/scripts/ollama_cloud_usage.py

Presenting Results

Format the JSON output for the user in a readable way, e.g.:

• Plan: Pro
• Session: 0.5% used, resets in 4 hours ▏█░░░░░░░░░░░░░░░░░░░
• Weekly: 38.9% used, resets in 4 days ▏████████░░░░░░░░░░░░░

If session_models or weekly_models are present, show the model-wise breakdown:

Session models:

• glm-5.1: 16 requests
• deepseek-v4-flash: 2 requests

Weekly models:

• glm-5.1: 1,313 requests
• deepseek-v4-pro: 268 requests
• deepseek-v4-flash: 17 requests

Common Pitfalls

1. execute_code does not inherit env vars. Always use terminal for this skill — execute_code runs in an isolated subprocess that won't have OLLAMA_CLOUD_COOKIE.
2. Cookie expires. Session cookies are short-lived. If parsing fails, re-extract the cookie and update the env var.
3. No usage data found. Either the cookie is invalid/expired, or Ollama changed their dashboard HTML. Check the cookie first.
4. Plan tier shows "Unknown". The badge markup varies. The script tries multiple patterns — if it fails, inspect the raw HTML around "Pro"/"Max" keywords.
5. Cookie not in environment. The script reads from os.environ directly (available in terminal). If missing, ensure the env var is set in the Hermes .env file.
6. Do NOT parse width: CSS values as usage percentages. The usage-meter__fill width matches the total, but usage-meter__segment widths (per-model breakdowns like 99.8% for one model) are NOT the total usage. Always parse from aria-label attributes or the "X% used" text labels instead.