chore(llm): Removing Claude Haiku 3.5

[justin-tahara]

Description

Removing to stay ahead

How Has This Been Tested?

Ran updated tests

Additional Options

• [Optional] Override Linear Check

---

Summary by cubic

Removed all Claude Haiku 3.5 model aliases from the LLM registry and updated the Anthropic prompt-caching test to a supported Haiku model. This removes a deprecated option and keeps caching metrics coverage.

• Migration
  • Update any configs using claude-3-5-haiku* IDs (incl. Bedrock/OpenRouter/Vertex AI variants) to claude-haiku-4-5-20251001 or another supported Anthropic model.

^{Written for commit 199b0d7. Summary will update on new commits.}

[greptile-apps]

Greptile Summary

This PR removes the deprecated claude-3-5-haiku model from the codebase and updates tests to use the newer claude-haiku-4-5 model naming convention.

• Removed all claude-3-5-haiku variants from model metadata (direct Anthropic, Bedrock, OpenRouter, Vertex AI)
• Updated prompt caching test to use claude-haiku-4-5-20251001 instead of claude-3-5-haiku-20241022
• Changes align with existing model naming in constants.py which already uses claude-haiku-4-5

Confidence Score: 5/5

• This PR is safe to merge - it's a clean model removal with consistent updates across all references.
• Simple metadata-only changes removing a deprecated model. All references have been consistently updated. The replacement model exists in the metadata file. No code logic changes.
• No files require special attention.

Important Files Changed

Filename	Overview
backend/onyx/llm/model_metadata_enrichments.json	Removed all claude-3-5-haiku model entries across multiple providers (Anthropic, Bedrock, OpenRouter, Vertex AI). Consistent removal with no residual references.
backend/tests/external_dependency_unit/llm/test_prompt_caching.py	Updated test to use claude-haiku-4-5-20251001 instead of claude-3-5-haiku-20241022. The new model exists in model_metadata_enrichments.json.

Sequence Diagram

sequenceDiagram
    participant User
    participant LitellmLLM
    participant Anthropic API
    
    Note over User,Anthropic API: Before PR: claude-3-5-haiku-20241022
    User->>LitellmLLM: Request with model name
    LitellmLLM->>Anthropic API: Call claude-3-5-haiku
    Anthropic API-->>LitellmLLM: Response
    LitellmLLM-->>User: Model response
    
    Note over User,Anthropic API: After PR: claude-haiku-4-5-20251001
    User->>LitellmLLM: Request with model name
    LitellmLLM->>Anthropic API: Call claude-haiku-4-5
    Anthropic API-->>LitellmLLM: Response
    LitellmLLM-->>User: Model response

Loading

[cubic-dev-ai]

No issues found across 2 files