Skip to content

docs(site): document phase 2 features - suites, A/B, plugins, benchmarks - #16

Merged
lostmartian merged 1 commit into
mainfrom
docs/site-phase2-sync
Aug 22, 2026
Merged

lostmartian merged 1 commit into
mainfrom
docs/site-phase2-sync

Conversation

@lostmartian

Copy link
Copy Markdown
Collaborator

What

Syncs the docs site with all Phase 2 feature work (PRs #9-#15): adapter registry/plugins, LangGraph + CrewAI direct ingestion, scenario suites, A/B benchmark mode, and the performance benchmark suite.

Why

Features shipped without site documentation; the agreed pre-0.3.0 sequence is benchmarks -> docs sync -> owner's website changes -> release.

How

  • Ingestion Adapters: LangGraph + CrewAI sections added to the table and detail; new "Custom adapters & plugins" section with runtime registration + entry-point examples.
  • New guide: Scenario Suites (run_scenarios, per-scenario GateThresholds, parallel workers, summary output).
  • New guide: A/B Benchmark Mode (run_benchmark, dimension-scoring semantics, explicit non-goal note on semantic judgment).
  • New Development pages: Custom Adapters (registry rules, plugin packaging, adapter quality bar) and Performance Benchmarks (make bench, what's measured, baseline findings incl. the alignment-dominates-scaling evidence that justified descoping streaming ingestion).

Testing

  • pnpm build compiles; static /docs output contains the new pages
  • Product untouched: lint clean, 300 tests green

Checklist

  • Exactly one logical change in this PR
  • make lint passes
  • uv run python -m pytest is green (300 passed)
  • Docs updated if user-facing
  • CHANGELOG entry - n/a (site content)
  • Conventional title

@lostmartian
lostmartian merged commit 6174517 into main Aug 22, 2026
@lostmartian
lostmartian deleted the docs/site-phase2-sync branch August 22, 2026 20:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant