Turn the tool from “a table you watch” into “a monitor that notifies you when something changes”.
What it is
A small rule engine that evaluates conditions continuously and triggers actions.
Rule format (conceptually):
• Condition: something becomes true for long enough
• Duration: avoid noise (e.g. must be true for 30–120s)
• Action: notify/log/webhook
Examples of useful rules
• Favorite model down
• “Alert me if nvidia/deepseek-v3.x uptime drops below 95% for 2 minutes”
• New best model
• “Notify when any S/S+ model becomes rank #1 for 60s”
• Provider degraded
• “Alert when provider Groq has >30% timeouts in last 5 minutes”
• Keyless becomes worth getting a key
• “If a NO KEY model is top 3 stable for 5 minutes, ping me”
Delivery channels (choose 2–3)
• Terminal alert: beep + highlight + write a one-line summary
• Desktop notification (optional): macOS/Windows/Linux
• Webhook: Discord/Slack/custom endpoint
• Log file: append events for later debugging (~/.free-coding-models/events.log)
Why this is a big upgrade
Users stop babysitting the TUI. They can:
• keep it running while working
• get notified only when it’s worth switching models/providers
• use it like real uptime/latency monitoring for free tiers
Pitfall to avoid ⚠️
Don’t fire alerts on every fluctuation. Use:
• sliding windows (last N pings / last 5 minutes)
• “must hold for X seconds”
• cooldown per rule (e.g. no more than once every 10 minutes)
Turn the tool from “a table you watch” into “a monitor that notifies you when something changes”.
What it is
A small rule engine that evaluates conditions continuously and triggers actions.
Rule format (conceptually):
• Condition: something becomes true for long enough
• Duration: avoid noise (e.g. must be true for 30–120s)
• Action: notify/log/webhook
Examples of useful rules
• Favorite model down
• “Alert me if nvidia/deepseek-v3.x uptime drops below 95% for 2 minutes”
• New best model
• “Notify when any S/S+ model becomes rank #1 for 60s”
• Provider degraded
• “Alert when provider Groq has >30% timeouts in last 5 minutes”
• Keyless becomes worth getting a key
• “If a NO KEY model is top 3 stable for 5 minutes, ping me”
Delivery channels (choose 2–3)
• Terminal alert: beep + highlight + write a one-line summary
• Desktop notification (optional): macOS/Windows/Linux
• Webhook: Discord/Slack/custom endpoint
• Log file: append events for later debugging (~/.free-coding-models/events.log)
Why this is a big upgrade
Users stop babysitting the TUI. They can:
• keep it running while working
• get notified only when it’s worth switching models/providers
• use it like real uptime/latency monitoring for free tiers
Pitfall to avoid⚠️
Don’t fire alerts on every fluctuation. Use:
• sliding windows (last N pings / last 5 minutes)
• “must hold for X seconds”
• cooldown per rule (e.g. no more than once every 10 minutes)