fix: update Gemini base URL to the correct endpoint

[MarcoMuellner]

This pull request updates the base URL for the Gemini provider across multiple files to ensure consistency and accuracy. The changes replace the outdated URL with the correct one.

Updates to Gemini provider configuration:

• README.md: Updated the example configuration to use the new base URL https://generativelanguage.googleapis.com/v1beta/models for the Gemini provider.
• codex-cli/src/utils/providers.ts: Updated the providers record to reflect the new base URL for the Gemini provider.
• codex-rs/core/src/model_provider_info.rs: Updated the built_in_model_providers function to use the new base URL for the Gemini provider.

Should fix #880

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[MarcoMuellner]

I have read the CLA Document and I hereby sign the CLA

[MarcoMuellner]

This doesn't a 100% fix the full issue - If you run this you won't get the 400 anymore, but rather you will get this:

OpenAI:DEBUG:request https://generativelanguage.googleapis.com/v1beta/models/models { method: 'get', path: '/models' } {
  accept: 'application/json',
  'content-type': 'application/json',
  'user-agent': 'OpenAI/JS 4.95.1',
  'x-stainless-lang': 'js',
  'x-stainless-package-version': '4.95.1',
  'x-stainless-os': 'MacOS',
  'x-stainless-arch': 'arm64',
  'x-stainless-runtime': 'node',
  'x-stainless-runtime-version': 'v22.14.0',
  authorization: 'REDACTED',
  'x-stainless-retry-count': '0',
  'x-stainless-timeout': '600'
}
OpenAI:DEBUG:response (error; (error; no more retries left)) 401 https://generativelanguage.googleapis.com/v1beta/models/models {
  'alt-svc': 'h3=":443"; ma=2592000,h3-29=":443"; ma=2592000',
  'content-encoding': 'gzip',
  'content-type': 'application/json; charset=UTF-8',
  date: 'Fri, 09 May 2025 18:42:44 GMT',
  server: 'scaffolding on HTTPServer2',
  'server-timing': 'gfet4t7; dur=27',
  'transfer-encoding': 'chunked',
  vary: 'Origin, X-Origin, Referer',
  'www-authenticate': 'Bearer realm="https://accounts.google.com/", error="invalid_token"',
  'x-content-type-options': 'nosniff',
  'x-frame-options': 'SAMEORIGIN',
  'x-xss-protection': '0'
} 

I assume the reason behind this is because gemini expects the key to be passed differently. As far as i understand it right now in the code, it assumes the provider works the same way as the OpenAI API does - and the curl script clearly tells a different story from google:

curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?key=GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-X POST \
-d '{
  "contents": [{
    "parts":[{"text": "Explain how AI works"}]
    }]
   }'

I assume this would probably require a deeper take on the whole thing? Not sure if this ever worked correctly?

[MarcoMuellner]

Probably also touches #731