Fix: Do not strip spaces from streaming chunks to prevent words from merging

Fix socket leaks on shutdown, remove hardcoded fake thinking block, and add debug logs for requests and tools

Update think block extraction and format inline code execution blocks

Reapply "fix: remove duplicate thinking block extraction as payload[4] already contains full formatted reasoning"

Revert "fix: remove duplicate thinking block extraction as payload[4] already contains full formatted reasoning"

fix: remove duplicate thinking block extraction as payload[4] already contains full formatted reasoning

feat: dynamically extract gemini_bl from homepage on startup

feat: extract and format reasoning block for thinking models

fix: 500 error on chat completions due to unhandled chunked encoding and missing JSON error handling

fix: thinking mode was hardcoded to always-on for all models

v2.0: curl_cffi Chrome TLS fingerprinting + upstream model migration

Update API authentication to use HF secrets

Create Dockerfile

Create gemini_web2api.py

initial commit