docs(readme): drop content duplicated by docs.erigon.tech

[bloxster]

What

The root README.md had grown into a second, slower-moving copy of the documentation site. This cuts it down to what only the repository can answer, and points everything else at the page that owns it.

README.md: 731 → 220 lines (+99 / −610).

Three commits, each reviewable on its own:

1. Move operator content to the site and add the developer sections that were missing.
2. Drop the sections the site already covers.
3. Link out Hetzner and Windows; fix the torrent-log wording.

What moved out, and where it was verified

Every section below was read against the page that owns it before being removed — not assumed.

Removed from README	Verified against
System Requirements	get-started/hardware-requirements — disk figures for all five networks, per-mode RAM, kernel > v4, SSD/NVMe-over-HDD and the capacity-degradation caveat
Sync Times	get-started/hardware-requirements, where the table now sits with a bandwidth caveat the README lacked
Usage / Getting Started	get-started/installation — clone, tag checkout, make erigon, make -j<n>, ./build/bin/erigon
Datadir structure	fundamentals/database → "The datadir at a glance" — same tree, clearer per-directory comments
Erigon3 datadir size	fundamentals/database → "What does it cost on disk?" — identical measured figures for all five networks
Caplin / Caplin's Usage	fundamentals/caplin + staking/caplin; --externalcl in fundamentals/architecture and staking/external-consensus-client-as-validator
Hetzner strict-firewall list	help-center/troubleshooting → Hetzner firewall note
Windows	get-started/installation — native compilation (Chocolatey, MinGW, the Win64/Kryptic.CIS anti-virus false positive), Docker, and WSL2

The earlier scope note in this PR — that it avoided everything #22866 and #22867 touched — no longer applies: both are merged to main, so the disk-size and network content they rewrote is settled and safe to defer to.

What is deliberately kept

• Why Caplin exists instead of the Engine API. The site does not have this. fundamentals/architecture only notes that the Engine-API path matches Geth/Besu/Reth; it does not explain that one-block-at-a-time delivery is what conflicts with Erigon's bulk model. Condensed to two sentences.
• Build prerequisites, moved out of the operator preamble into a new Building section under For developers — where a repository README should state its own toolchain. Trimmed to the toolchain plus a pointer to the installation page.
• Everything a checkout can answer but the site cannot: executables table, testing, go get as a library, repository-doc index, devnet, diagnostics.

Two corrections, not just deletions

• The Hetzner list was corrupted. It carried 127.16.0.0/12 "Private-Use Networks RFC 1918" — a mangling of loopback and RFC 1918, with 172.16.0.0/12 already listed separately — plus a Network Interconnect row and a Limited Broadcast row each split across two lines by the code fence. The linked version is correct.
• The torrent-log level read backwards. The README said logs/torrent.log is written at whichever is lower of --torrent.verbosity and WARN. erigonToSlogLevel is 12 - 4*lvl (slogger.go:13), so a higher Erigon verbosity maps to a numerically lower slog.Level, and the min(erigonToSlogLevel(verbosity), slog.LevelWarn) at downloadercfg.go:249 selects the more verbose threshold. Reworded to match the behaviour operators actually see.

Ordering

Two of these cuts land content that reaches readers only once #22962 deploys — the Sync Times table and the corrected Hetzner blocklist are added to the site there, against release/3.5 (the branch docs-deploy.yml publishes from). Merge this after #22962 has deployed, otherwise those two links point at pages that do not yet carry the content.

Known gaps, for a follow-up docs PR against release/3.5

Three details existed only in the README and are not on the site. They are not blockers for the cut, but they should be added:

• caplin/ is missing from the datadir tree in fundamentals/database.
• The six domains are named in the README (account, storage, code, commitment, receipt, rcache); database.md only counts them as "4 state domains + 2 receipt domains".
• The off-by-default size table had per-network columns for --prune.include-receipts, --prune.include-commitment-history and the Caplin archive; database.md gives mainnet only, as prose.

Verification

• Every relative repo link resolves. This also fixes one that was already broken on main: cmd/downloader/README.md does not exist — the file is cmd/downloader/readme.md, now listed under Repository docs.
• Both remaining internal anchors resolve; no dangling anchors left by the removed sections, and nothing in the repo or on the docs site links to any removed README anchor.
• Every docs.erigon.tech target checked to exist in docs/site/, including the anchors #install-native, #install-wsl and the Hetzner firewall note.
• docs/site: npm ci && npm run build clean; generate-llms.py --check OK (4 files, 74 pages).
• Markdown only; no Go changed.

[Copilot]

Pull request overview

This PR reduces duplication between the root README.md and the docs site (docs.erigon.tech) by removing large operator-focused sections that already exist in docs/site/docs/, and replacing them with curated links to the canonical documentation while keeping repository/developer-focused guidance.

Changes:

• Replaces the large README table-of-contents + duplicated operator docs with a “Documentation” section linking to key docs.erigon.tech entry points.
• Trims/redirects several previously inlined sections (logging, beacon-chain/JWT/CL setup, ports/firewalls, rpcdaemon, etc.) to the docs site.
• Adds a “For developers” section with executables, testing targets, and a “Repository docs” link list to in-repo documentation.

---

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