installer: pick-free install + auto-resolve Hindsight-compatible Python#1085
Merged
Merged
Conversation
… a raw pip wall The Hindsight memory engine builds its own venv and pulls litellm, which publishes `requires-python >=3.10,<3.14`. On a Python 3.14 host (Ubuntu 25.10+/26.04) the main hal0 venv is fine, but litellm's metadata gate makes `pip install hindsight-api` fail with a wall of "Could not find a version" resolver output before the install could fall back. Resolve the Hindsight interpreter up front in preflight: - prefer the default python when it's already in litellm's 3.11-3.13 band - else pick python3.13→3.11 from PATH - else best-effort auto-install one via the distro package manager (HAL0_HINDSIGHT_AUTOINSTALL=1, set by install.sh) - else fall back to the default interpreter and go STRAIGHT to --ignore-requires-python (litellm runs fine on 3.14 — the classifier was dropped over a since-resolved fastuuid wheel gap, BerriAI/litellm#26343), skipping the doomed first attempt and its resolver noise. Exposes HAL0_HINDSIGHT_PYTHON override and a read-only preflight_hindsight_python wired into preflight_all so `hal0 doctor` reports the decision. The Hindsight install block now consumes HINDSIGHT_PY / HINDSIGHT_PY_FALLBACK rather than always using ${PY}. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… slots The installer already never downloaded weights (--no-pull); it only recorded a hardware-recommended model name per slot (chat always, coder if an agent extension is on, plus ComfyUI capability defaults) into the slot configs. We no longer want to ship those picks — a fresh box should boot with services up and let the operator choose every model afterward. - setup_command.build_auto_selections gains with_slots (default True); when False it seeds no slot picks and no ComfyUI defaults, leaving extensions untouched. - new `hal0 setup --no-slots` flag threads it through; interactive/manual `hal0 setup` still picks recommended models. - install.sh first-run setup now runs `--auto --no-pull --no-slots --no-extensions`, seeding the first-run sentinel + wiring but zero model picks. Messaging updated to "no model picks, no downloads". Fresh-boot consequence: OpenWebUI/Hermes come up with no model configured until the operator runs `hal0 setup` / `hal0 model pull`. Docs follow-up tracked in #1082, #1083, #1084. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
thinmintdev
added a commit
that referenced
this pull request
Jul 6, 2026
The pick-free installer (#1085) runs `hal0 setup --auto --no-pull --no-slots --no-extensions` — it writes the first-run sentinel + wiring only and seeds NO model slot. #1088 corrected first-model.mdx/cli.mdx but setup.mdx and the README still described the old "seeds the recommended Main slot" behavior. - setup.mdx first-run Aside: real installer command + "no model slots, no download; a fresh box boots with no model configured". - README (install one-liner, feature list, setup-TUI note): drop "recommended Main slot", say the installer seeds no model picks and nothing downloads. Closes #1082. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
thinmintdev
added a commit
that referenced
this pull request
Jul 6, 2026
The pick-free installer (#1085) runs `hal0 setup --auto --no-pull --no-slots --no-extensions` — it writes the first-run sentinel + wiring only and seeds NO model slot. #1088 corrected first-model.mdx/cli.mdx but setup.mdx and the README still described the old "seeds the recommended Main slot" behavior. - setup.mdx first-run Aside: real installer command + "no model slots, no download; a fresh box boots with no model configured". - README (install one-liner, feature list, setup-TUI note): drop "recommended Main slot", say the installer seeds no model picks and nothing downloads. Closes #1082. Co-Authored-By: Claude Opus 4.8 (1M context) <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
1 participant
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.
Two related installer-robustness changes, split into one commit each.
1. Auto-resolve a Hindsight-compatible Python (
60444fa)The Hindsight memory engine builds its own venv and pulls
litellm, which publishesrequires-python >=3.10,<3.14. On a Python 3.14 host (Ubuntu 25.10+/26.04) the main hal0 venv is fine, but litellm's metadata gate madepip install hindsight-apifail with a wall of "Could not find a version" resolver output before the install could fall back.Now resolved up front in preflight (
resolve_hindsight_python):python3.13→3.11from PATH--ignore-requires-python(litellm runs fine on 3.14 — classifier dropped over a since-resolved fastuuid wheel gap, [Feature]: Support Python 3.14 BerriAI/litellm#26343), skipping the doomed first attempt and its noiseAdds
HAL0_HINDSIGHT_PYTHONoverride and a read-onlypreflight_hindsight_pythonwired intopreflight_allsohal0 doctorreports the decision.2. Stop shipping model picks (
fc50839)The installer already never downloaded weights (
--no-pull); it only recorded a hardware-recommended model name per slot (chat always, coder if an agent extension is on, plus ComfyUI capability defaults) into the slot configs. We no longer ship those picks — a fresh box boots with services up and the operator chooses every model afterward.build_auto_selectionsgainswith_slots(defaultTrue);False→ zero slot picks, zero ComfyUI defaults, extensions untouchedhal0 setup --no-slotsflag threads it through; interactive/manualhal0 setupstill picks recommended models--auto --no-pull --no-slots --no-extensionsFresh-boot consequence: OpenWebUI/Hermes come up with no model configured until the operator runs
hal0 setup/hal0 model pull. Docs follow-up: #1082, #1083, #1084.Testing
bash -nclean oninstall.sh+preflight.shresolve_hindsight_pythonexercised across clean / PATH-scan / forced-fallback / override / skip casestests/cli/test_setup_command.py→ 6/6 pass;with_slots=Falseverified to seed 0 slots / 0 comfyui picks with extensions still on--no-slotsconfirmed registered on the CLI🤖 Generated with Claude Code