Pattern Proposal: Pluggable LLM Provider Abstraction

Summary

Wraps LLM invocations in a standardized client interface enabling transparent switching between production multi-provider models and deterministic test mock clients.

Candidate Pattern

• ID: pluggable-llm-provider-abstraction
• Classification: intelligence
• Maturity: established
• Type: implemented

Observed Signals

No specific signals listed.

Evidence

Supporting Evidence

• src/cas_runner/llm/litellm_client.py (LiteLLMClient) - Production implementation for multi-provider invocation
• src/cas_runner/llm/mock_client.py (MockLLMClient) - Deterministic test mock implementation

Scope Assessment

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

Rationale

• Intent: Decouple orchestration and governance agents from provider-specific APIs while facilitating deterministic local testing.
• Context: Agentic components (merger, compliance engine) require model completions but must remain agnostic to specific LLM endpoints and testable without active API keys.
• Force: Need multi-model flexibility across distinct providers without changing prompt consumption logic.
• Force: Requirement to test agent pipelines offline in CI without consuming API budget.

Traceability

Artifact source: proposed-patterns