Skip to content

docs: v1.0 parity sweep — gttsize deprecation, model-store note, cognee retirement, memory-delete gate - #46

Open
thinmintdev wants to merge 1 commit into
masterfrom
docs/v1-parity-sweep
Open

docs: v1.0 parity sweep — gttsize deprecation, model-store note, cognee retirement, memory-delete gate#46
thinmintdev wants to merge 1 commit into
masterfrom
docs/v1-parity-sweep

Conversation

@thinmintdev

Copy link
Copy Markdown
Contributor

Upstream half of the docs work in Hal0ai/hal0 PR #1364. docs/** there is a mirror of this repo, so these edits must land here or the next mirror-docs.yml sync clobbers them.

Path mapping verified against mirror-docs.yml: src/content/docs/docs/** → hal0s docs/** (flat, doubled segment stripped, rsync --delete, excluding internal/ and README.md).

Changes

  • #1081baremetal-ubuntu.mdx, proxmox-lxc.mdx: upstream still had the dual-param GRUB line (amdgpu.gttsize=126976 ttm.pages_limit=32505856) and "keep it equal to gttsize" prose. Now ttm.pages_limit-only, dropped table row, deprecation Aside naming amdttm.pages_limit
  • #1083first-model.mdx: added the "Where the pull lands" Aside ([models].store, pull_root fallback, --storage-dir)
  • #985configure.mdx, enable-memory.mdx, config-schema.mdx: cognee dropped from the engine enum description; removed the "Migrate a legacy store" section documenting a --cognee-dir flag that never existed in this branchs CLI either
  • Cross-lanereference/api/rest-api.mdx, reference/mcp-tools.mdx: document the /mcp/memory bulk-delete approval gate, which hal0 PR #1359 makes real. Upstreams pre-edit text was confirmed to match that branchs diff context verbatim before applying

Not ported, deliberately

#1084 cannot land upstream yet. This repos first-model.mdx still says the installer seeded the registry — #1084s "fresh box has zero models" premise contradicts that sentence directly. Both #1083 and #1084 are filed as blocked by #1082, and #1082s rewrite was never ported here. #1082 is the actual blocker.

Also skipped: the honcho-memory.mdx cross-link added in hal0 (that page does not exist here — would dangle).

Pre-existing divergence found, not fixed

This repos master has drifted well behind hal0s docs — slot role/chat naming vs agent/utility, ADR-0014 vs ADR-0023, missing proxmox gid= passthrough values, missing "Point hal0 at your model storage" and "Troubleshooting" sections. That is a much larger reconciliation than these four issues and is noted, not attempted.

Verification

npx astro check -> 5 errors, all pre-existing (stash-tested at branch HEAD), none in the 8 changed files
npm run build   -> 57 pages, 0 errors; all 8 changed pages present in output

Zero overlap with fix/web-install-parity-and-manifests (diffed against master directly).

🤖 Generated with Claude Code

…ination

Ports the hal0-side docs/v1-parity-sweep fix (Hal0ai/hal0, same branch
name) forward into this repo's mirrored copies, so the next mirror-docs
sync doesn't clobber it. Scoped to only the four target issues' own
facts — does not sync unrelated drift between this branch's base
(upstream/master@e1cef95) and hal0's current docs; see PR description /
agent report for the divergence inventory.

- #1081 (getting-started/baremetal-ubuntu.mdx, getting-started/proxmox-lxc.mdx):
  GTT sizing guidance drops amdgpu.gttsize from the GRUB example/table,
  uses ttm.pages_limit=32505856 as the sole knob, and adds a deprecation
  Aside covering the amdttm.pages_limit out-of-tree-TTM-module equivalent
  and the modinfo check to tell which applies.
- #985 (guides/configure.mdx, guides/enable-memory.mdx, reference/config-schema.mdx):
  the 'cognee' memory-engine value is retired (schema now rejects it,
  docs updated to match); removed enable-memory.mdx's "Migrate a legacy
  store" section, which documented a `hal0 memory migrate --cognee-dir`
  flag that doesn't exist in the current CLI.
- #1083 (getting-started/first-model.mdx): added a "Where the pull lands"
  note covering [models].store / the legacy pull_root fallback /
  --storage-dir, inserted standalone since this branch's first-model.mdx
  still describes the pre-#1082 bundle-picker first-run flow (see report
  for why #1084's fresh-box-state content was NOT ported here).

Also ports two files from Hal0ai/hal0 branch fix/memory-delete-hardening
(a different, concurrent lane) since they document a real, just-landed
API behavior change and belong in the same coherent docs PR:
- reference/api/rest-api.mdx: /api/memory is ADMIN-classed end-to-end;
  DESTRUCTIVE_MEMORY_ROUTES pins the irreversible subset above the
  generic prefix rule.
- reference/mcp-tools.mdx: the memory_delete bulk-delete arity gate
  applies identically on both the /mcp/memory and /mcp/admin mounts.

Verified: npx astro check — 5 pre-existing errors in src/content.config.ts
(zod version mismatch between astro and starlight-blog) and
src/pages/changelog.astro (implicit-any), confirmed present at this
branch's HEAD before these changes too (stash-tested) — none touch the
8 files changed here. npm run build — 57 pages built clean, all 8 changed
pages present in the output, 0 errors.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
@vercel

vercel Bot commented Jul 27, 2026

Copy link
Copy Markdown

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

Project Deployment Actions Updated (UTC)
hal0-web Ready Ready Preview, Comment Jul 27, 2026 2:37pm

@cloudflare-workers-and-pages

Copy link
Copy Markdown

Deploying hal0-web with  Cloudflare Pages  Cloudflare Pages

Latest commit: bbefaf4
Status: ✅  Deploy successful!
Preview URL: https://7fb7f9e7.hal0-web.pages.dev
Branch Preview URL: https://docs-v1-parity-sweep.hal0-web.pages.dev

View logs

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