feat(iflow): add support for new models (glm-5, minimax-m2.5, qwen3-32b, tstars2.0, iflow-rome-30ba3b)

[MasuRii]

Summary

The competitor repository (router-for-me/CLIProxyAPI) recently added support for new iFlow models. We need to add these models to maintain feature parity.

Models to Add

• glm-5 — Zhipu GLM 5 (with thinking support, GLM-style handling)
• minimax-m2.5 — MiniMax M2.5 (with thinking support, uses reasoning_split)
• qwen3-32b — Qwen3 32B (with thinking support)
• tstars2.0 — TStars-2.0 multimodal assistant
• iflow-rome-30ba3b — iFlow Rome model

Requirements

• Add all 5 models to HARDCODED_MODELS list
• Add glm-5 to ENABLE_THINKING_MODELS and GLM_MODELS sets
• Add minimax-m2.5 to REASONING_SPLIT_MODELS set
• Add qwen3-32b to ENABLE_THINKING_MODELS set
• Update REASONING_PRESERVATION_MODELS_PREFIXES with appropriate prefixes
• Maintain backward compatibility
• Ensure proper error handling

References

• Competitor: https://github.com/router-for-me/CLIProxyAPI/

[mirrobot-agent]

@MasuRii Thank you for submitting this issue. I am now beginning my analysis and will report back shortly.

[mirrobot-agent]

Initial Analysis Report

Summary: The request is to add 5 new models to the iFlow provider and implement support for their specific "thinking" and "reasoning" features by updating various configuration sets.
Issue Validation: Confirmed. The requested models are currently missing from the HARDCODED_MODELS list in src/rotator_library/providers/iflow_provider.py.
Reproducibility Assessment: Reproducible. The absence of these models can be verified by inspecting the provider's hardcoded model list or by attempting to use them via the proxy.
Root Cause Analysis: The iflow_provider.py file has a static HARDCODED_MODELS list that has not been updated with the latest models (glm-5, minimax-m2.5, etc.). Furthermore, the proxy does not yet have the specific logic or configuration sets (ENABLE_THINKING_MODELS, REASONING_SPLIT_MODELS, etc.) requested for handling the advanced features of these models, which appear to be based on the architecture of the referenced competitor repository.
Suggested Labels: enhancement, provider:iflow. These labels are justified as this is a feature request to expand support for a specific provider.
Proposed Next Steps:

1. Update Model List: Add the 5 new models to the HARDCODED_MODELS list in src/rotator_library/providers/iflow_provider.py.
2. Infrastructure Implementation: Since the sets ENABLE_THINKING_MODELS, GLM_MODELS, REASONING_SPLIT_MODELS, and REASONING_PRESERVATION_MODELS_PREFIXES do not currently exist in the codebase, they must be introduced.
3. Feature Logic: Implement the "GLM-style handling" for glm-5 and "reasoning_split" logic for minimax-m2.5 within the iflow_provider.py completion logic.
4. Verification: Verify that the new models are correctly identified and that thinking content is properly aggregated or split according to the new configurations.
   Missing Information (if any):

• The specific definitions of the requested sets (ENABLE_THINKING_MODELS, etc.) are missing from the current repository. Clarification is needed on whether these should be implemented as new global constants in iflow_provider.py or if there is a preferred central configuration file.
• The "appropriate prefixes" for REASONING_PRESERVATION_MODELS_PREFIXES should be specified (e.g., glm-, minimax-, etc.).

This analysis was generated by an AI assistant.