Birds.Quest – My Effortless Bird Watching Auto Blogging Buddy

Description and use

An automated bird‑watching blog generator that turns GPS tracks into draft posts, preserving privacy and allowing user approval.

Problem

Manual entry of bird watching walks into eBird became tedious, turning a fun activity into homework.

Outcome

The system automatically turns a GPS track into a draft blog post about birds that could be seen on the route, preserving privacy by omitting location data and allowing user approval before publishing.

Workflow

• Record GPS positions with Traccar Client
• Traccar server stores positions in MariaDB
• Read‑only worker discovers trips from positions
• Sample route points and determine week number
• Run BirdNET GeoModel to predict species probabilities
• Filter predictions by threshold and top‑20 cap
• Fetch species summaries and licensed images from Wikipedia
• Generate draft post with Gemma 4 LLM
• Open /admin to review, tick seen species, regenerate if needed
• Publish draft to public blog

Evidence

Location never leaves the server. The blog and the LLM prompt contain no coordinates and no place names.

Source

The pipeline processed it like this: Discover, Sample, Week, Predict, Filter

Source

The GeoModel ran once, as a single batch of 24 [lat, lon, week] inputs, on a plain CPU. Each species keeps its best score along the route.

Source