Restore configs from 5743f66

Fix: default model config

Fix config default model to openrouter format

Fix: model_name format for litellm OpenRouter = openrouter/owl-alpha

Restore 3d39bbc: frontend config

Config default to deepseek-ai/DeepSeek-V4-Pro

Try openai/ prefix with api_base for OpenRouter owl-alpha

Fix: correct litellm model ID for OpenRouter = openrouter/openrouter/owl-alpha

Set default model to openrouter/owl-alpha

Update default model to DeepSeek-V4-Pro

Restore 0c286f8: configs + patches

Fix: Use deepseek/ provider directly (litellm native support, no openrouter prefix needed)

Fix models: DeepSeek V3 via OpenRouter as default + fix model IDs

Restore to f15d8ed

Restore config from 4bac0e2

Remove all custom files to restore 4bac0e2

Fix: Try direct deepseek model without openai prefix - let llm_params handle HF router format

Fix: Override available models to use OpenRouter free models + correct model IDs

Fix: correct Gemini 2.5 Flash model ID on OpenRouter

Try simpler model name format for OpenRouter

Switch to Gemini 2.5 Flash (cheap, fast, good tool calling)

Fix: correct model ID format for OpenRouter free model

Switch to DeepSeek V3 free via OpenRouter (has tool calling)

Fix: correct OpenRouter model name format

Fix: Use HF router format with OpenRouter credentials passed as HF token trick

Fix: Use OpenAI format with OpenRouter base URL for Claude Opus 4

Fix: Use Claude Opus via OpenRouter by setting Anthropic base URL to OpenRouter

Fix: correct OpenRouter model ID for Claude Opus 4

Switch to Claude Opus 4.8 via OpenRouter

Switch to Claude via OpenRouter

Switch to OpenAI GPT-4o

Fix: Use novita provider for Qwen (works with standard HF token)

Use Qwen3-235B via HF Inference (free with Pro) - supports tool calling

Fix: Override model to use Anthropic API directly (not Bedrock)