docs(readme): refresh stats + v0.9.26 callout + corrected P@5

[rohitg00]

Summary

README + website meta were both behind v0.9.25/v0.9.26.

README

• "New in" callout: v0.9.22 → v0.9.26 + v0.9.25 summary
• coding-agent-life-v1 table (line 247): P@5 0.578 / 0.267 → 0.240 / 0.227. The pre-feat(eval): pluggable benchmark harness with in-house coding-agent corpus #562 numbers were mathematically impossible under the current hits / k formula (math ceiling at K=5 = 0.240 on this corpus). Reframed lift as recall + temporal not aggregate precision. Same root cause as PR docs(benchmarks): correct coding-agent-life-v1 P@5 numbers (#796) #805 (scorecard correction).
• Stats line (line 1144): 118 files / 21,800 LOC / 950+ tests / 123 functions / 34 KV scopes → 174 / 37,800 / 1,390+ / 258 / 44.
• npm test comment (line 1468): 950+ → 1,390+.
• stat-tests.svg + dark variant: 1171+ → 1390+.

Website

• website/lib/generated-meta.json: regen via node scripts/gen-meta.mjs. Bumped v0.9.20 → v0.9.26, tests 1055 → 1394, REST endpoints 124 → 126.
• The live site at https://www.agent-memory.dev/ already auto-derives the right numbers at Vercel prebuild — this only refreshes the committed local-dev snapshot.

Out of scope

• 95.2% R@5 LongMemEval still backed by benchmark/LONGMEMEVAL.md — unchanged.
• stat-tools (53), stat-hooks (12), stat-deps (0), stat-recall, stat-tokens verified against current source — unchanged.
• No code changes. Doc/asset-only.

Test plan

• Doc-only; no source or test files touched
• Numbers verified against current src/, plugin/hooks/hooks.json, benchmark/LONGMEMEVAL.md
• node website/scripts/gen-meta.mjs writes the committed JSON
• P@5 numbers consistent with merged scorecard correction in PR docs(benchmarks): correct coding-agent-life-v1 P@5 numbers (#796) #805

Summary by CodeRabbit

• Documentation
  • Updated test coverage metrics to show 1,390+ passing tests.
  • Expanded release notes with v0.9.26 hotfix details and added v0.9.25 regression/feature summary.
  • Refreshed benchmark results and explanations for the coding-agent-life-v1 agent.
  • Updated public feature statistics (sources, LOC, tests, functions, KV scopes) and build metadata/version.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
agentmemory	Ready	Preview, Comment	Jun 3, 2026 10:14am

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: b66a4323-cb33-49d2-a816-dfd3318628f1

📥 Commits

Reviewing files that changed from the base of the PR and between 22b9502 and f21ccf3.

📒 Files selected for processing (1)

• website/lib/generated-meta.json

✅ Files skipped from review due to trivial changes (1)

• website/lib/generated-meta.json

---

📝 Walkthrough

Walkthrough

This PR updates README content (test badge, release notes for v0.9.26/v0.9.25, coding-agent-life-v1 benchmark numbers, "Powered by iii" stats, and adds an npm test line) and refreshes generated website metadata (version, restEndpoints, testsPassing, generatedAt).

Changes

README Metrics and Release Notes

Layer / File(s)	Summary
Test count and dev command updates README.md	Updated test passing badge text to "1,390+ tests passing" and added an npm test development command referencing the new count.
Release notes update README.md	Replaced the "New in" changelog snippet to include v0.9.26 hotfix and a v0.9.25 regression/feature summary.
Benchmark and project stats edits README.md	Reworked coding-agent-life-v1 benchmark table values and narrative; updated "Powered by iii" statistics to higher totals.

Generated Metadata

Layer / File(s)	Summary
website metadata bump website/lib/generated-meta.json	Bumped version to 0.9.26, increased restEndpoints and testsPassing, and updated generatedAt timestamp.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

Possibly related PRs

• rohitg00/agentmemory#660: Also updates README metrics/version figures and overlaps with the README counts/version area.

Poem

A rabbit hops through README bright,
Counts grow taller in the light.
Tests leap up to a higher tune,
Benchmarks shimmer, stats festoon.
npm runs, the badges cheer—hooray! 🐇✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Linked Issues check	⚠️ Warning	The PR is a documentation-only update that refreshes README and site stats. Linked issue #797 reports a runtime error in index persistence; the PR makes no code changes to address this.	This PR appears to be a documentation refresh unrelated to the runtime bug in #797. Clarify the relationship between this PR and the linked issue, or link to the correct issue if this documentation change addresses a specific requirement.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely summarizes the main changes: documentation refresh with updated statistics, v0.9.26 callout, and corrected P@5 metrics.
Out of Scope Changes check	✅ Passed	The PR contains only documentation updates to README and generated metadata, which align with the stated objective of refreshing stats and callouts for releases v0.9.22 through v0.9.26.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch docs/readme-refresh-v0.9.26

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}