docs(changelog): catalogue unreleased post-v0.8.4b1 work (incl. #1045)#1047
Merged
Conversation
Fills the empty [Unreleased] section with everything landed on main since the v0.8.4b1 cut — GPU generalization (experimental CUDA + multi-GPU), companion-service management (#1037), settings phases 3-5 + Advanced section, catalog UX (#1042), /api/meta/enums taxonomy, remote- model identification (#1035), board + slot-pipeline + settings fixes, and the README/docs truth pass (#1044, #1046) — plus the #1045 profile consolidation (virtual seeds, embed/rerank lanes, MTP tri-state) that is merging alongside this. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LsqZewS75G2Grf9AjUQmun
…ite-sync-1r1b6c # Conflicts: # CHANGELOG.md
thinmintdev
pushed a commit
that referenced
this pull request
Jul 4, 2026
Convert [Unreleased] (catalogued in #1047) into the v0.8.5b1 section with the upgrade notes, and touch up the #1045 bullets per #1047's own caveat: the PR landed with bench-matrix tooling (profile-matrix.sh + server_ab.py + on-box runbook), the data-safe prune migration (backup + -custom rescue), and the null-clears-override config-write semantics — all now reflected. Bump pyproject to 0.8.5b1 (the only version pin; release-check gate 7). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ky49xV79kYjmiYMp4uzHMw
thinmintdev
added a commit
that referenced
this pull request
Jul 4, 2026
* release: v0.8.5b1 — version bump + changelog cut Convert [Unreleased] (catalogued in #1047) into the v0.8.5b1 section with the upgrade notes, and touch up the #1045 bullets per #1047's own caveat: the PR landed with bench-matrix tooling (profile-matrix.sh + server_ab.py + on-box runbook), the data-safe prune migration (backup + -custom rescue), and the null-clears-override config-write semantics — all now reflected. Bump pyproject to 0.8.5b1 (the only version pin; release-check gate 7). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ky49xV79kYjmiYMp4uzHMw * release: sync uv.lock to 0.8.5b1 The lockfile pins the project's own version; uv re-locked it after the pyproject bump. Without this, the first uv sync on main dirties the tree. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ky49xV79kYjmiYMp4uzHMw * providers: gate the MTP auto-off breadcrumb to the launch path mtp.auto_off_model_ineligible was emitted inside _effective_mtp, which runs in _resolve_llama_scalars — the SINGLE SOURCE for launch AND preview. The preview path is hit by every dashboard GET /api/slots poll, so the once-per-launch hint fired once per status render per polling client (measured 1:1 — 43 logs to 43 status GETs in 120s with zero launches) for any AUTO slot pairing an MTP profile with a non-MTP model. _effective_mtp grows a keyword-only log_ineligible flag (default False); _resolve_llama_scalars threads it as for_launch, set True only by container_spec. Resolved values are identical on both paths — the flag gates the side-effect log only, so launch/preview argv parity is untouched. Regression test pins preview-silent / launch-logs-once; CHANGELOG entry added to the v0.8.5b1 section. Reported-by: operator (log-volume analysis on the live box) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ky49xV79kYjmiYMp4uzHMw * release: v0.8.5b1 upgrade note — slot units are not auto-re-rendered on update Operational finding from the live box: a container slot's systemd unit bakes the launch argv at load time, so after an update re-runs the stale ExecStart; only a hal0-level restart (unload→load) re-renders the unit through the new code. Document the operator action for this release; automatic unit re-render on update (decoupled from restart) is the planned follow-up. Reported-by: operator (stale hal0-slot@agent.service, zero spec-type tokens post-update) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ky49xV79kYjmiYMp4uzHMw --------- Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fills
CHANGELOG.md's empty[Unreleased]section with everything landed onmainsince the v0.8.4b1 release commit, so the next cut starts from an accurate ledger.Added: GPU generalization (experimental
cudaseed profile,gpu_indexmulti-GPU pinning, multi-file pulls), companion-service management (#1037: registry,/api/services, mDNS, dashboard Services page), settings phases 3–5 (#1038/#1040/#1043), Advanced settings section (fullhal0.tomlparity, memory-graph panel, API restart, AWS preset, reload-from-disk), dedicatedembed/rerankGPU profiles (#1045), catalog UX finish (#1042),GET /api/meta/enumstaxonomy.Changed: virtual seed profiles with prune migration (#1045 PS-2); MTP as a model × profile × slot decision with tri-state slot override (#1045).
Fixed: remote-model identification (#1035), Operator Board Hermes-contract repairs (#1032), slot-pipeline hardening, settings polish.
Docs: README truth pass + Discord link (#1044/#1046), hal0.dev docs mirror refresh incl. the auth-docs correction, three handoffs (#1010/#1034/#1041).
Note: the #1045 entries describe the PR that's merging concurrently — if it lands materially changed, the three bullets referencing it should be touched up before the next release cut.
🤖 Generated with Claude Code
https://claude.ai/code/session_01LsqZewS75G2Grf9AjUQmun
Generated by Claude Code