Clarify Phase 2b practice RAGAs limitations

[Camier]

Summary

• Reframe the Oct 27 practice report to note that it only validated workflow rehearsal and not retrieval quality
• Link to the Nov 5 validation README and summarize the flaws that invalidated the 10-query dataset (undersized corpus, page-based chunking, circular ground truth, NaN-prone metrics)
• Direct readers to the updated evaluation roadmap and explicitly document what, if anything, from the session remains useful

Testing

• Not run (documentation-only change)

---

Codex Task

[Copilot]

Pull Request Overview

This PR updates the Phase 2b Day 1 practice report to clarify that the October 27 RAGAs validation was only a workflow rehearsal, not a production-quality evaluation. The update acknowledges that the dataset used was later invalidated due to fundamental methodology flaws discovered on November 5, 2025.

• Reframes the practice session as a workflow/infrastructure rehearsal rather than a validation of retrieval quality
• Documents the specific flaws that invalidated the dataset (undersized corpus, page-based chunking, circular ground truth, NaN-prone metrics)
• Clarifies what remains valuable from the session (workflow validation, cost/latency measurements, team readiness) while discarding the metrics

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.