LLM Foundry deliverables

Local files and screenshots for the repo update.

What changed

• Semantic retrieval now uses an embedding index.
• Multi-provider support includes OpenAI-compatible and Anthropic endpoints.
• Docs now say orchestration helps, but does not magically upgrade a weak model.

Validation

benchmark: 4 cases
harnesses: reasoning 60%, coding 100%, tool_use 100%, memory 100%
commit pushed to GitHub