de8a7e7c66
completed the portland_wildlife_dex_entries.csv added a method to skip species that don't have a wiki page
Suppai_Neko2026-03-26 15:45:28 -07:00
8978130b7c
Created boilerplate setup for wilddex.db file. .db file as 5 tables: lookalike, sightings, species species_detail and sqlite_sequence. Example .db file added
Suppai_Neko2026-03-26 15:41:51 -07:00
b229f5816e
Changed get_wikipedia_article to be more robust Will now fall back to wiki if local copy isn't avail
SuppaiNeko2026-03-26 12:17:06 -07:00
f01728adb7
- Added dex_info_writer.py with a full pipeline for generating Pokédex-style species descriptions - Loads species from portland_wildlife.csv (taxon_id, common_name, wikipedia_url) - Fetches Wikipedia articles from a self-hosted Kiwix mirror (WIKI_LINK env var); falls back to live Wikipedia if unset - Spaces in Wikipedia URLs are normalized to underscores before fetching - LiteLLM instance prompted via OpenAI-compatible client (LITELLM_URL, LITELLM_API_KEY, LITELLM_MODEL env vars) - LLM has tool-calling access to fetch_wikipedia_article with an agentic loop - Output descriptions are sanitized — only alphanumerics and . , ! ? ' - : allowed - Writes results to portland_wildlife_dex_entries.csv in append mode with resume support (skips already-processed taxon_ids) - Live [i/total] progress counter printed each iteration along with the generated description
SuppaiNeko2026-03-25 21:02:30 -07:00
7a91a9bed4
Fix wikipedia_url population in INaturalist requester - Fixed wikipedia_url being incorrectly looked up in the ancestors dict (which maps rank→name) — now reads directly from the taxon object - Updated enrich_with_ancestors to capture and store wikipedia_url per taxon alongside ancestor data - Merges wikipedia_url back into species during enrichment pass, with fallback to value captured in the first pass
SuppaiNeko2026-03-25 21:00:10 -07:00
Suppai_Neko
SuppaiNeko