feat: CLI rework — ungate catalog, eliminate redundancies, add stack modification (#40)#41
Merged
feat: CLI rework — ungate catalog, eliminate redundancies, add stack modification (#40)#41
Conversation
The workflow permissions fix resolved 4 CodeQL code-scanning alerts (actions/missing-workflow-permissions) and should be documented under a Security heading rather than just Bug Fixes. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Dependabot PR #36 (pygments 2.19.2 → 2.20.0) fixes catastrophic backtracking CVEs but was missed by release-please because build(deps) is not a tracked conventional commit type. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Allow `mlx-stack pull` to accept arbitrary HuggingFace repo strings (containing '/') in addition to catalog IDs. HF repos bypass catalog lookup and download directly. Catalog ID behavior is unchanged. - Add hf_repo_override param to pull_model() in core/pull.py - Route MODEL arg in cli/pull.py based on '/' detection - Update help text documenting both input types - Add 26 new tests covering HF repo acceptance, error handling, flag combinations, disk space checks, and inventory tracking Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Add a third resolution path in resolve_target() that detects HF repo strings (containing '/') and handles them: checks local models dir for already-downloaded copy, creates a minimal synthetic CatalogEntry for benchmarking, finds a free port, and starts a temp vllm-mlx instance. This enables both 'mlx-stack bench mlx-community/Model-4bit' as a standalone command and 'mlx-stack pull mlx-community/Model-4bit --bench' to resolve the target correctly. Also updates bench CLI help text to document HF repo support and fixes stale references to removed 'recommend' and 'init' commands in bench and pull CLI output. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
…h '--' HF repo IDs (e.g. mlx-community/Model-4bit) were used directly as benchmark service names, creating invalid PID/log file paths since process.py uses service_name for filesystem operations. Now replaces '/' with '--' in _start_temp_instance() to produce path-safe names like bench-temp-mlx-community--Model-4bit. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Remove the profile CLI command entirely. Add hardware info section (chip, GPU cores, memory, bandwidth) to status output reading from profile.json via core/hardware.py load_profile(). Add hardware data to status --json under 'hardware' key. Handle missing/corrupt profile gracefully. Update no-stack messages to reference 'setup' instead of 'init'. Delete test_cli_profile.py and add 16 new hardware tests. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Add is_estimate to HardwareProfile.to_dict() so it persists in profile.json. Update load_profile() to read is_estimate from saved data (defaulting to False for legacy profiles). Include is_estimate in status --json hardware output. Add 9 new tests covering round-trip serialization, table display, and JSON output for estimated bandwidth. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Remove the standalone 'recommend' CLI command and absorb its functionality into 'models --recommend'. Add --budget, --intent, --show-all flags that work with --recommend, and --available flag that queries HF API. Make --recommend, --available, and --catalog mutually exclusive. Ensure --budget/--intent/--show-all require --recommend. Update display-only notice to reference 'setup' not 'init'. Delete test_cli_recommend.py and add comprehensive new tests in test_cli_models.py. Update all user-facing strings that referenced 'recommend' or 'init' as commands. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Delete cli/init.py and test_cli_init.py. Remove init import and registration from cli/main.py. Update _COMMAND_CATEGORIES to remove init from 'Setup & Configuration'. Update test_cli.py, test_cross_area.py, test_cli_up.py, and test_cli_watch.py to reflect init removal. Update launchd.py and recommend.py docstrings. core/stack_init.py preserved for internal use by setup. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
- Remove src/mlx_stack/cli/recommend.py (deregistered from main.py but file remained) - Update cli/setup.py module docstring to remove old init flow reference - Update test_cross_area.py comments to reference setup instead of init command Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
… reference Replace the post-benchmark message in cli/pull.py that referenced 'models --recommend' with a generic message: 'Results saved. These will be used for model scoring.' Add VAL-CROSS-008 test to verify the output no longer references removed commands. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
…ack modification Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
--model MODEL creates a single-tier 'standard' stack without the wizard. --no-pull skips model download in wizard, --model, and --add flows. --no-start skips stack startup. --no-pull implies --no-start. --model is mutually exclusive with --add/--remove. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
…--model Two scrutiny fixes: 1. _resolve_model_source() now returns entry.id instead of entry.name for catalog ID resolution, so tiers[].model stores the canonical ID (e.g., 'qwen3.5-8b') rather than the display name. 2. _single_model_setup() no longer auto-starts services — it always prints 'mlx-stack up' guidance per the no-auto-restart convention. Tests updated to explicitly verify both model and source fields, and to assert start_stack is never called from --model path. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
weklund
pushed a commit
that referenced
this pull request
Apr 5, 2026
🤖 I have created a release *beep* *boop* --- ## [0.3.8](v0.3.7...v0.3.8) (2026-04-04) ### Features * CLI rework — ungate catalog, eliminate redundancies, add stack modification ([#40](#40)) ([#41](#41)) ([3bee7d9](3bee7d9)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.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.
Summary
Implements phases 1-4 of #40. Reworks the CLI to eliminate redundant commands, remove the catalog-as-whitelist pattern, and add the missing stack modification path.
Closes #40
Supersedes #27
Changes
Milestone 1: Ungate Pull
pullaccepts any HuggingFace repo string (e.g.,mlx-community/Phi-5-Mini-4bit) in addition to catalog IDsMilestone 2: Absorb Profile into Status
profilecommand deletedstatusnow shows hardware info (chip, GPU cores, memory, bandwidth) with estimate indicatorsstatus --jsonoutputMilestone 3: Absorb Recommend into Models, Remove Init
recommendcommand deletedinitcommand deletedmodelsgains--recommendflag (scored tier recommendations with--budget,--intent,--show-all)modelsgains--availableflag (live HuggingFace discovery with scoring overlay)--recommend,--available,--catalogare mutually exclusiveMilestone 4: Setup Modification Flags
setup --add MODEL— add a model to existing stack (HF repo or catalog ID, repeatable)setup --as TIER— custom tier name for--addsetup --remove TIER— remove a tier (repeatable, prevents empty stack)setup --model MODEL— single-model quick setup (skips wizard)setup --no-pull— generate config without downloadingsetup --no-start— generate config without starting servicesImpact
Validation