Revert "feat: force model to react with lat and lon for guessing"

feat: retying for loading page

feat: force model to react with lat and lon for guessing

fix: openrouter model sometimes has a dash

fix: do not fail silently

fix: add missing dash in qwen model id

Merge pull request #2 from yichuan520030910320/qwen

fix: also support openrouter in app.py

Merge remote-tracking branch 'origin/main' into qwen

udd mapcrunch version

update title

small about comment and ready to use

small fix

add a package

fix app.py and now it is a good version

fix app.py and run agent loop

fix guess

elegeant copy

fix some guess logic

add raw online mode and give an example and allow paste to the empty bar

add raw online mode and give an example

add raw online mode

add address collect

change ui for user-friendly

change dataset name

add default

feat: support qwen and openrouters

fix: streamlit compatible old-ver

fix: history ui

add clean

add temp to user space

fix agent meory stuff

fix: shorten title

fix: update app title to Omniscient branding

add new title

upd link

fix: the decisions sometimes are not followed

fix: qwen 2.5 not supported

revert: better ui

fix: keep simple, and use hf_token for qwen

fix: unique streamlit key

feat: more models, including qwen

fix: headless mode, do some hacks

feat: let users choose diff datasets

feat: persistent dataset by name

chore: upgrade streamlit dep

chore: remove macos specific packages

fix: using requirements.txt as per HF docs

chore: rename requirements for hf adpat

docs: README