Pattern Proposal: Model-Agnostic Provider Indirection

Summary

Abstract LLM interactions behind a unified client adapter with interchangeable mock fallbacks.

Candidate Pattern

• ID: model-agnostic-provider-indirection
• Classification: intelligence
• Maturity: established
• Type: implemented

Observed Signals

No specific signals listed.

Evidence

Supporting Evidence

• src/cas_runner/llm/litellm_client.py (LiteLLMClient) - Production LLM abstraction layer using LiteLLM
• src/cas_runner/llm/mock_client.py (MockLLMClient) - Mock client supporting offline and deterministic agent runs

Scope Assessment

• Evidence Strength: high
• Confidence: high
• Applies to: ideas-to-life-cas

Rationale

• Intent: Shield agent reasoning steps from provider-specific APIs and allow deterministic offline testing.
• Context: Agent execution flows require flexible LLM routing and dry-run execution without network overhead.
• Force: Vendor lock-in with proprietary SDKs
• Force: Need for reproducible unit and end-to-end integration tests without live API keys

Traceability

Artifact source: proposed-patterns