feat: add support of gpt2giga

[doppoluv]

📑 Description

Added support for the GPT2Giga proxy provider to integrate with the GigaChat API via OpenAI/Anthropic-compatible interfaces. Key changes include:

1. Created gpt2giga.go with the client implementation:
   • Acts as a proxy to forward requests to the GigaChat API using OpenAI format
   • Configure() method supports setting proxy endpoints and model parameters
   • GetCompletion() sends requests to the proxy endpoint and processes responses
2. Added the Gpt2GigaClient to the list of available providers in iai.go
3. Updated backend and passwordless provider lists

✅ Checks

• My pull request adheres to the code style of this project
• My code requires changes to the documentation
• I have updated the documentation as required
• All the tests have passed

ℹ Additional Information

[linux-foundation-easycla]

• ✅ login: doppoluv / name: Semyon Inokov (380987c, 72c126b)
• ❌ The email address for the commit (150685b) is not linked to the GitHub account, preventing the EasyCLA check. Consult this Help Article and GitHub Help to resolve. (To view the commit's email address, add .patch at the end of this PR page's URL.) For further assistance with EasyCLA, please submit a support request ticket.

[AlexsJones]

It would be cleaner to reuse the existing OpenAIClient with a custom base URL instead of adding a new gpt2giga provider, as it avoids duplicated logic and maintenance overhead. If a distinct client is required, consider fixing the variable shadowing, adding tests, and updating docs.

[three-foxes-in-a-trenchcoat]

New provider support. The OpenAI-compatible proxy pattern is well-implemented.

Notes:

1. Default baseURL is localhost:8090 — consider documenting this as a proxy endpoint that users need to set up themselves.
2. No error handling for non-200 response bodies (just returns status code). Consider including the response body in the error for debugging.
3. No support for system messages or conversation history (single-turn only). Document this limitation.

LGTM for a new provider.

[codecov]

Codecov Report

❌ Patch coverage is 0% with 50 lines in your changes missing coverage. Please review.
✅ Project coverage is 43.25%. Comparing base (173e4dc) to head (72c126b).
⚠️ Report is 253 commits behind head on main.

Files with missing lines	Patch %	Lines
pkg/ai/gpt2giga.go	0.00%	50 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1635      +/-   ##
==========================================
+ Coverage   34.76%   43.25%   +8.48%     
==========================================
  Files          94      120      +26     
  Lines        6342     7431    +1089     
==========================================
+ Hits         2205     3214    +1009     
+ Misses       4046     4005      -41     
- Partials       91      212     +121     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[AlexsJones]

Hi @doppoluv — steward check-in here. Alex left design feedback back in April that's still open:

It would be cleaner to reuse the existing OpenAIClient with a custom base URL instead of adding a new gpt2giga provider… If a distinct client is required, consider fixing the variable shadowing, adding tests, and updating docs.

Since then the branch has only picked up merges from main — nothing addressing that feedback — so it's stalled. Two ways forward:

1. Preferred (Alex's suggestion): drop the dedicated Gpt2GigaClient and just point the existing openai backend at the proxy (--baseurl http://localhost:8090). gpt2giga already exposes the OpenAI /chat/completions shape your client targets, so this avoids a duplicated backend and its maintenance.
2. Keep the dedicated client, but then please: (a) rename the url := fmt.Sprintf(...) local in GetCompletion — it shadows the imported net/url package (a likely golangci-lint trigger, which is currently red); (b) add a unit test — the other backends in pkg/ai use an httptest.Server to cover the happy path plus the non-200 and empty-choices error branches — so codecov/patch goes green; and (c) add a short docs note for the new backend.

Separately, EasyCLA is still unsigned, which is a hard merge blocker regardless of the approach above — the sign link is in the failing check.

Happy to re-review as soon as you've picked a direction and CI is green. If we don't hear back, we'll likely close this as stale in a couple of weeks — just drop a note to keep it open.

🤝 Handled by Alex's Repo Steward — replies reviewed by @AlexsJones.

Learn more or run your own: https://github.com/AlexsJones/repo-steward