Skip to content

feat(migration): haloai → hal0 model migration script#22

Merged
thinmintdev merged 1 commit into
mainfrom
feat/haloai-migration-script-v2
May 16, 2026
Merged

feat(migration): haloai → hal0 model migration script#22
thinmintdev merged 1 commit into
mainfrom
feat/haloai-migration-script-v2

Conversation

@thinmintdev
Copy link
Copy Markdown
Contributor

Closes #27. 14-entry curated allow-list, 19 tests pass, docs/migration.md covers full 5-phase cutover.

PLAN §11 step 2. Reads a curated 14-entry allow-list of model dirs
from the shared HF cache (default /mnt/ai-models/huggingface/hub) and
writes a hal0 registry.toml describing the resolved entries. Models
stay on NFS — only the registry file is emitted.

Scope is intentionally narrow per session decision: models only.
Slots, providers, upstreams, and OpenWebUI state all start fresh —
operator runs FirstRun wizard post-install to bind models to slots.

Allow-list covers the daily-driver chat/coder models on haloai:
Qwen3-Coder-Next (GGUF + AMD MXFP4), Qwen3.6-27B/35B-A3B,
Qwen3.6-27B-Heretic-NEO-CODE, Qwen3-Coder-REAP-25B, Qwen3-Zero-
Coder-Reasoning-0.8B-NEO, Qwen3-Next-80B-Thinking, Qwen3.5 family
(0.8B/4B/9B/35B-A3B), kappa-20b (MXFP4 + i1 GGUF).

Llama 4 / Llama 4 Scout / Nemotron 115B requested but not on disk —
documented as post-cutover `hal0 model pull` in docs/migration.md.

19 hermetic pytest tests, no real /mnt/ai-models touched. Every
emitted entry re-validates via the hal0.registry.Model pydantic
schema. Closes task #27.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@thinmintdev thinmintdev merged commit ce61231 into main May 16, 2026
@thinmintdev thinmintdev deleted the feat/haloai-migration-script-v2 branch May 21, 2026 20:11
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