refactor(executor): split providers and preserve legacy entrypoints

feat: passthrough upstream response headers to clients

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

refactor(executor): optimize payload cloning and streamline SDK translator usage

feat(openai): add responses/compact support

feat(executor): apply payload rules using requested model

feat(registry): support provider-specific model info lookup

refactor(thinking): pass source and target formats to ApplyThinking for cross-format validation

fix(executor): properly handle thinking application errors

refactor: improve thinking logic

feat(executor): add HttpRequest support across executors for better http request handling

feat(executor): enhance payload translation with original request context

refactor(executor): remove redundant upstream model parameter from translateRequest

refactor(executor): resolve upstream model at conductor level before execution

fix(executor): use upstream model for thinking config and payload translation

fix(gemini): add optional skip for gemini3 thinking conversion

fix: apply thinkingLevel from model suffix metadata for Gemini 3

feat: use thinkingLevel for Gemini 3 models per Google documentation

refactor(thinking): export thinking helpers

Fixed: #534

refactor(executor): clarify executor comments and oauth names

refactor(executor): clarify providers and streams

fix(executor): centralize default thinking config

feat(executor): normalize thinking budget across all Gemini executors

feat(aistudio): normalize thinking budget in request translation

refactor(executor): dedupe thinking metadata helpers across Gemini executors

fix(aistudio): strip Gemini generation config overrides

Fixed: #302

feat(executor): share SSE usage filtering across streams

Fixed: #291

**feat(runtime): add payload configuration support for executors**

fix(executor): remove `safetySettings` from payload in token counting request

fix(executor): only apply thinking config to supported models

fix(executor): pass authID to relay instead of provider

feat(registry): unify Gemini models and add AI Studio set

fix(aistudio): ensure colon-spaced JSON in responses

refactor(wsrelay): rename RoundTrip to NonStream

fix(aistudio): remove generationConfig and tools when action is countTokens

fix(aistudio): strip usage metadata from non-final stream chunks

feat(aistudio): track Gemini usage and improve stream errors

feat: add websocket routing and executor unregister API