docs: rewrite public README #8

Merged

fwazb merged 1 commit from ship/readme-public-rewrite into main

2026-05-13 03:20:03 +00:00

m3taversal commented

2026-04-28 09:19:41 +00:00

Owner

Replaces directory-listing README with one that explains pipeline architecture and shows production scale.

Key changes

Lede: Lead with what the pipeline actually does + verified production metrics (1,546 claims, 13 domains, 1,975 merged PRs, ~73 merges/day, 94% approval, ~$0.10/merged claim).

Two Mermaid diagrams: pipeline flow (Inbox→Extract→Validate→Evaluate→Merge→Effects) and review tier matrix (Light/Standard/Deep). Both verified to render on GitHub and Forgejo 9/Gitea 1.22.

Sensitive details removed (per Epimetheus review):

VPS public IP
as user teleo username
/opt/teleo-eval/ install paths
Hetzner CAX31 machine size

These move to the private teleo-ops repo (Cory creating).

Numbers verified live against production:

Claim count: grep -rl '^type: claim' domains/ core/ foundations/ | wc -l = 1,546
Domain count: ls domains/ | wc -l = 13
Merged PRs: SELECT COUNT(*) FROM prs WHERE status='merged' = 1,975
7d throughput: WHERE merged_at > datetime('now','-7 days') = 508
Approval rate: 1608 approved / 1711 total = 94.0% exact
Cost/claim: 30d sum / 30d merges = $0.0971 → rounded to $0.10
Tier mix: STANDARD 76%, LIGHT 21%, DEEP 2%

Models corrected from memory (this is what 'check your work' caught):

Extract was 'Sonnet 4.6' → actually Sonnet 4.5 via OpenRouter
Domain review was 'GPT-4o' → actually Gemini 2.5 Flash (GPT-4o is legacy)
Leo standard was 'Sonnet 4.6' → actually Sonnet 4.5
Verified by reading lib/config.py + lib/extract.py + lib/evaluate.py

Ownership simplification: dropped per-directory table (a duplicate source of truth that drifts). CODEOWNERS is the only enforced ownership. README now keeps a 5-line high-level role map for orientation, points at CODEOWNERS for the actual rules.

Review focus

Are the production metrics correct as you read them?
Mermaid renders correctly on Forgejo PR view?
Anything still sensitive that should move to teleo-ops?
Tone — too boastful? Too understated?

Replaces directory-listing README with one that explains pipeline architecture and shows production scale. ## Key changes **Lede:** Lead with what the pipeline actually does + verified production metrics (1,546 claims, 13 domains, 1,975 merged PRs, ~73 merges/day, 94% approval, ~\$0.10/merged claim). **Two Mermaid diagrams:** pipeline flow (Inbox→Extract→Validate→Evaluate→Merge→Effects) and review tier matrix (Light/Standard/Deep). Both verified to render on GitHub and Forgejo 9/Gitea 1.22. **Sensitive details removed** (per Epimetheus review): - VPS public IP - `as user teleo` username - `/opt/teleo-eval/` install paths - Hetzner CAX31 machine size These move to the private `teleo-ops` repo (Cory creating). **Numbers verified live against production:** - Claim count: `grep -rl '^type: claim' domains/ core/ foundations/ | wc -l` = 1,546 - Domain count: `ls domains/ | wc -l` = 13 - Merged PRs: `SELECT COUNT(*) FROM prs WHERE status='merged'` = 1,975 - 7d throughput: `WHERE merged_at > datetime('now','-7 days')` = 508 - Approval rate: 1608 approved / 1711 total = 94.0% exact - Cost/claim: 30d sum / 30d merges = \$0.0971 → rounded to \$0.10 - Tier mix: STANDARD 76%, LIGHT 21%, DEEP 2% **Models corrected from memory** (this is what 'check your work' caught): - Extract was 'Sonnet 4.6' → actually Sonnet 4.5 via OpenRouter - Domain review was 'GPT-4o' → actually Gemini 2.5 Flash (GPT-4o is legacy) - Leo standard was 'Sonnet 4.6' → actually Sonnet 4.5 - Verified by reading lib/config.py + lib/extract.py + lib/evaluate.py **Ownership simplification:** dropped per-directory table (a duplicate source of truth that drifts). CODEOWNERS is the only enforced ownership. README now keeps a 5-line high-level role map for orientation, points at CODEOWNERS for the actual rules. ## Review focus - Are the production metrics correct as you read them? - Mermaid renders correctly on Forgejo PR view? - Anything still sensitive that should move to teleo-ops? - Tone — too boastful? Too understated?

m3taversal added 1 commit 2026-04-28 09:19:41 +00:00

docs: rewrite public README

CI / lint-and-test (pull_request) Has been cancelled

Details

b8504c1b60

Replaces the directory-listing format with one that explains what the
pipeline does and shows production scale. Verified all numbers against
production (1,546 claims, 13 domains, 1,975 merged PRs, 508 last-7d
throughput, 94% approval, ~\$0.10/merged claim incl. all stages).

Removes the VPS layout section (IP + paths + username) per Epimetheus
review — that detail moves to the private teleo-ops repo. Generalizes
deploy targets without naming the host.

Adds two Mermaid diagrams (pipeline flow + review tier matrix), both
syntactically safe across GitHub and Forgejo 9 / Gitea 1.22.

Drops the per-directory ownership table — CODEOWNERS is the single
source of truth on review authority. Keeps the high-level role map
for orientation only.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>