Add reproduction log entries for Foundations of Retrieval lessons 1-3

[niloy-biswas]

Add reproduction log entries for the "Foundations of Retrieval" onboarding path (lessons 1-3): start-here.md, BM25 baseline, and dense retrieval (BGE) on MS MARCO passage ranking.

Setup

• OS: macOS 26.5.2 (Darwin 25.5.0), Apple M4, 16 GB RAM
• Java: OpenJDK 21.0.12 (Homebrew)
• Maven: 3.9.16
• Build: bin/qbuild.sh (quick build, fatjar produced), verified against the CACM smoke test (map 0.3123, P_30 0.1942 — matches expected)

Results

All three lessons reproduced exactly against the documented expected values:

• start-here.md: downloaded/converted MS MARCO passage collection (checksum 31644046b18952c1386cd4564ba2ae69 verified), 9 JSONL files (8,841,823 docs total, last file 841,823 lines — matches), filtered dev queries to 6,980 lines, cross-referenced qrels/collection for a sample query — all as documented.
• experiments-msmarco-passage.md (BM25 baseline): indexed 8,841,823 docs (0 errors, ~3 min, 4.4G index). MRR@10 = 0.18741227770955546 (official MS MARCO eval script), map = 0.1957, recall_1000 = 0.8573 (trec_eval) — exact match.
• experiments-msmarco-passage2.md (dense retrieval): prebuilt BM25 index run gave MRR@10 = 0.1875 (exact match). BGE-base HNSW dense retrieval (26 GB prebuilt index) gave MRR@10 = 0.3521 (exact match), confirming the expected improvement over BM25.

No issues encountered, aside from a transient GitHub API rate-limit (403) on the prebuilt-index metadata fetch, which resolved itself after the standard hourly reset.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 73.40%. Comparing base (02f25de) to head (6356b9a).

Additional details and impacted files

@@            Coverage Diff            @@
##             master    #3359   +/-   ##
=========================================
  Coverage     73.40%   73.40%           
  Complexity     1913     1913           
=========================================
  Files           206      206           
  Lines         13493    13493           
  Branches       1867     1867           
=========================================
  Hits           9904     9904           
  Misses         2843     2843           
  Partials        746      746           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.