🔧 refactor: Improve Params Handling, Remove Legacy Items, & Update Configs

[danny-avila]

Summary

• Refactored stream handling in AnthropicClient by introducing a custom SplitStreamHandler that extracts delta text and reasoning smoothly, without needing to create new objects mid-stream
• Removed legacy logic and environment references for OPENROUTER_API_KEY from .env.example, OpenAIClient, ModelService, and related tests.
  • Closes [Bug]: Using OPENROUTER_API_KEY breaks all librechat.yaml custom endpoints #2097
• Added comprehensive unit tests for getLLMConfig to verify proxy settings, reverse proxy URLs, and conditional inclusion of topK/topP parameters.
• Restructured parameter handling in AnthropicClient so that top_k and top_p parameters are only included for non-claude-3.7 models, and I validated this behavior with targeted tests.
• Updated maxOutputTokens handling and tests to enforce appropriate token cap limits across various Claude model variants.
• Skipped index fixing in CI, development, and test environments within the Token model to avoid unnecessary index operations.
• Adjusted default token values and schema transformations (including simplifying the googleSchema transformation) to better support model-specific constraints.
• Excluded image files from service worker caching and revised client configuration (Vite config, tsconfig, and ESLint settings) to pre-cache all assets, in hopes of resolving offline usage.
• Add frontend/backend feature for Resetting parameters

Change Type

• Bug fix (non-breaking change which fixes an issue)
• Refactor (non-breaking change that improves code quality)
• CI/Chore (updates to development workflows and configurations)

Checklist

• My code adheres to this project's style guidelines
• I have performed a self-review of my code
• I have commented in any complex areas of my code
• I have made pertinent documentation changes
• My changes do not introduce new warnings
• I have written tests demonstrating that my changes are effective
• Local unit tests pass with my changes
• Any changes dependent on mine have been merged and published in downstream modules

[KiGamji]

@danny-avila the top_p and top_k parameters need to be dropped only when thinking is enabled, otherwise they are supported for the Claude 3.7 Sonnet model

[danny-avila]

@KiGamji ok thanks