feat(setup): scaffold capability + NPU slots (pick-free), guide models interactively#1091
Merged
Merged
Conversation
…s interactively Closes #1090. Builds on the pick-free installer (#d2c2101e): we still never choose a model for the operator, but setup now provisions the capability + NPU slot *structure*. - orchestrate: SlotSelection.model_id is now Optional; apply_setup creates an empty scaffold slot (device/profile/port wired, model.default unset, no pull) when model_id is falsy. Modelless slots are already a supported runtime state. - setup_command: `--auto` scaffolds the capability set (chat/embed/rerank/stt/ tts/vision, +coder when an agent ext is on) as empty slots — no model picks. `_SETUP_SLOTS` extended with embed/stt/tts/rerank/vision ports; `--no-slots` still seeds truly zero slots. - setup_ui: interactive `hal0 setup` guides each capability — pick a fitting model, scaffold empty, or skip (never auto-selects). New "capabilities" step + pane copy; review table shows "(scaffold)". - suggest: add rerank; fix stt capability string ("asr" -> "stt", which matched no curated model); vision matches chat+mmproj models. - install.sh: first-run `--auto` drops `--no-slots` so a fresh box gets the empty capability-slot scaffold (models chosen later); "Next steps" reworded. Tests: modelless apply_setup, --auto scaffold set + --no-slots, rerank/stt/embed suggestions, interactive pick/scaffold/skip. ruff + affected pytest green (pre-existing unrelated hermes-shim env failures excluded). Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
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.
Closes #1090.
Builds on the pick-free installer (d2c2101): we still never choose a model for the operator, but setup now provisions the capability + NPU slot structure so a fresh box has the scaffolding, not just chat/coder.
What changes
Core — modelless scaffold slots (
install/orchestrate.py)SlotSelection.model_idis nowstr | None. When falsy,apply_setupcreates an empty scaffold slot (device/profile/port wired,model.defaultunset) and plans no pull. Modelless slots are already a supported runtime state (slot.modelless_ready_blocked).--autoscaffolds empty capability slots (cli/setup_command.py)--autonow scaffoldschat/embed/rerank/stt/tts/vision(+coderwhen an agent extension is on) as empty slots — pick-free: slots yes, models no._SETUP_SLOTSextended with embed/stt/tts (mirroring_SLOT_METAports) + rerank/vision.--no-slotsstill seeds truly zero slots.Interactive
hal0 setupguides, doesn't decide (cli/setup_ui.py,setup_copy.py)capabilitiesstep + pane copy; review table shows(scaffold — choose later).Suggestions (
install/suggest.py)rerank; fixsttcapability string ("asr"→"stt"— no curated model usesasr, so speech suggestions returned nothing);visionmatches chat+mmproj models.Installer (
installer/install.sh)--autodrops--no-slotsso a fresh box gets the empty capability-slot scaffold; "Next steps" reworded.img/ NPU notesimgstays on the ComfyUIcomfyui_defaultssidecar (not a llama-server slot) — unchanged.npu_opt_inrouting (stt/agent/utility→npudevice); dormant passengers (stt-npu/embed-npu) stillderive_device → Noneby design.sttis only offered interactively when the NPU trio is on.Tests
test_orchestrate: modellessapply_setupcreates empty slot, no pull.test_setup_command:--autoscaffold set + allmodel_id is None;--no-slotsseeds nothing.test_suggest: rerank/stt/embed suggestions return.test_setup_ui: interactive pick / scaffold / skip; capabilities step planned + pane copy.ruff check+ruff format --checkclean; affected pytest green (316 passed intests/install+tests/cli; the 3 failingtest_agent_*hermes-shim tests are pre-existing/environmental and import none of these modules).Left open for review — touches the just-refactored install path.
🤖 Generated with Claude Code