Update readme.md

readme.md

@@ -24,9 +24,10 @@ This repository contains configuration files for a doorbell camera setup using F
 ## Files
 
 - `config.yml` - Frigate NVR configuration
-- `ha_automation.yaml` - Home Assistant automation
+- `doorbell_notification.yaml` - Home Assistant automation
 
 ## Frigate Configuration Highlights
+
 ```yaml
 detectors:
   cpu:
@@ -39,3 +40,62 @@ objects:
       min_area: 2000
       min_score: 0.6
       threshold: 0.5  # 0.5s detection time before triggering
+```
+
+**Detection Parameters:**
+- `min_area: 2000` - Minimum pixel area for person detection
+- `min_score: 0.6` - Confidence threshold (60%)
+- `threshold: 0.5` - Person must be detected for 0.5 seconds
+
+## Home Assistant Automation
+
+The automation triggers on new person detection events and includes:
+- `mode: single` - Prevents concurrent runs
+- 60-second cooldown between notifications
+- Dynamic event snapshot URLs
+- Deep link to camera dashboard
+
+## Setup
+
+### Frigate
+
+1. Copy `config.yml` to your Frigate config directory (typically `/config`)
+2. Update the following values:
+   - MQTT credentials (`host`, `user`, `password`)
+   - Camera RTSP URLs and credentials
+   - Camera IP address in the snapshot URL
+
+### Home Assistant
+
+1. Import `doorbell_notification.yaml` as a new automation
+2. Update `notify.household_phones` to your notification service
+3. Adjust the camera dashboard URL if needed
+
+## Tuning
+
+### Too many notifications?
+- Increase cooldown delay in Home Assistant automation (currently 60s)
+- Increase `threshold` in Frigate config (higher = longer detection time required)
+
+### Missing detections?
+- Lower `min_score` (currently 0.6)
+- Lower `threshold` (currently 0.5)
+- Lower `min_area` (currently 2000)
+
+### Detection too slow?
+- Lower `threshold` to 0.3-0.4 seconds
+- Ensure cooldown in HA automation prevents spam
+
+## Network Configuration
+
+- Frigate Web UI: `http://192.168.0.5:5341`
+- Camera IP: `192.168.0.118`
+- MQTT Broker: `192.168.0.205:1883`
+
+## License
+
+MIT
+
+## Credits
+
+Optimized through iterative testing for the Amcrest AD110 doorbell form factor and typical approach patterns.