Helix StudioGenome IDE
← Docs
Helix CLI docs
Browse Helix CLI docs

Helix CLI (headless) quick reference

Helix CLI lets you simulate CRISPR/Prime edits and generate HTML reports without the GUI. Ideal for GPU servers and batch pipelines; Studio can open the same .helix sessions for interactive drill-down.

New in 1.0.x:

  • helix partner run produces seeded partner demos + support bundle for intake.
  • helix run --out … --backends all emits repro bundles; verify with helix verify --kind repro.
  • Lightcone audit packs can be dumped/verified headlessly (helix experiment dump --audit-pack …).

Commands

  • Simulate (create/update a session):

    • helix-cli simulate guides.json batch.helix
    • append to existing: helix-cli simulate more_guides.json batch.helix --append
    • backend override (optional): --backend gpu|native-cpu|cpu-reference

  • Reports (HTML per run):

    • all runs: helix-cli report --session batch.helix --outdir reports/
    • single run: helix-cli report --session batch.helix --run-id demo_crispr_g2 --outdir reports/

Config schema

See docs/cli_config_schema.md for the full JSON schema with CRISPR + PRIME examples.

Common per-run fields:

  • guide_id (required), edit_type (crispr | prime), sequence (required)
  • target_locus (optional), pam_index, cut_index (optional)
  • metadata (optional dict, propagated to reports/UI)

Prime-specific:

  • prime.pbs_sequence (required), prime.rt_sequence (required)
  • prime.nick_guide_id / prime.nick_cut_index (optional)
  • prime.spacer (optional)
  • prime.pam_profile (optional, default SpRY_NNN)
  • prime.draws (optional), prime.seed (optional)

Typical workflow

  1. Prepare a config JSON with your guides/runs.
  2. Run helix-cli simulate ... to produce a .helix session.
  3. Run helix-cli report ... to generate HTML reports (one per run).
  4. Open the .helix file in Helix Studio for interactive inspection (Guide Inspector, Outcome Explorer, Compare, PCR CTA).

Notes

  • Reports include guide info, physics + backend, outcomes (JSON + CSV), and run metadata.
  • CLI uses the same RunModel/EngineInfo schema as Studio; artifacts are interchangeable.
  • On headless servers, you don't need Qt; CLI runs fully without a display.

Batch parallelism

Batch scans in Studio can parallelize candidate evaluation without changing outputs.

Config key: batch_parallelism.workers (in the session config)

Default: 1

Set to N to enable deterministic parallel candidate evaluation during batch scans. Results are identical; only runtime changes. Each worker keeps its own reference cache, so memory use scales with the worker count.