Fix usage convertation from gemini response to openai format

feat(translator): support Claude thinking type adaptive

fix(translator): correct gemini-cli log prefix

fix(translator): normalize and restrict `stop_reason`/`finish_reason` usage

feat(translator): capture cached token count in usage metadata and handle prompt caching

refactor(translator, executor): remove redundant `bytes.Clone` calls for improved performance

feat(gemini-cli): support image content in Claude request conversion

fix(translator): handle stop_reason and MAX_TOKENS for Claude responses

feat(translator): add code_execution and url_context tool passthrough

Fixed: #1077

Merge branch 'main' into dev (PR #961)

refactor(translator): remove registry model lookups from thinking config conversions

Fixed: #1064

refactor(thinking): remove support for non-standard thinking configurations

fix(thinking): use LookupModelInfo for model data

fix(thinking): map reasoning_effort to thinkingConfig

refactor: improve thinking logic

Add candidate count handling in OpenAI request

Merge pull request #850 from can1357/main

feat(translator): add `thoughtSignature` to node parts for Gemini and Antigravity requests

feat(translator): add developer role support for Gemini translators

fix(translators): ensure correct handling and output of multimodal assistant content across request handlers

refactor(image): avoid using json.Marshal

fix(openai): add index field to image response for LiteLLM compatibility

fix(translators): update role from `tool` to `user` in Gemini and Gemini-CLI requests

fix(translators): handle string system instructions in Claude translators

Fixed: #654

refactor: replace `json.Marshal` and `json.Unmarshal` with `sjson` and `gjson`

Fixed: #642

Fixed: #551

fix(thinking): centralize reasoning_effort mapping

fix(thinking): map budgets to effort for level models

fix(claude): prevent final events when no content streamed

feat(responses): add unique identifiers for responses, function calls, and tool uses

fix(executor): centralize default thinking config

refactor(translator): remove thinking budget normalization across all translators

Fixed: #433

**fix(translator): ensure partial content is retained while skipping encrypted thoughtSignature**

**fix(translator): reintroduce `thoughtSignature` bypass logic for model parts**

fix: handle empty and non-JSON SSE chunks safely

fix(translator): ignore empty JSON chunks in OpenAI responses

fix(translator): include empty text chunks in responses

feat(translator): support image size and googleSearch tools

fix(gemini-cli): ignore thoughtSignature and empty parts

**feat(auth, executor, cmd): add Antigravity provider integration**

feat(translator): add default thinkingConfig for gemini-3-pro-preview

**fix(translator): handle missing parameters in Gemini tool schema gracefully**

**feat(translator): add ThoughtSignature handling in Gemini request transformations**

Fixed: #263

fix(translator): accept camelCase thinking config in OpenAI→Gemini