HelixSpec MVP v0.1

Goal

A typed, deterministic DSL for genome editing that compiles into a canonical IR (helixspec_ir_v1) and a verifiable manifest (helixspec_manifest_v1).

Scope (v0.1)

• One file -> one IR document (no sweeps/templates yet).
• Supported edits: cut, prime, base.
• Explicit, typed fields only. No general expressions.

Grammar (high level)

• version 1
• genome <build>
• sample "name" { ploidy: <int> }
• locus NAME = chrN:start-end(+|-)
• cut edit NAME { nuclease ... guide ... target ... repair ... }
• prime edit NAME { nuclease ... pegRNA ... target ... }
• base edit NAME { editor ... guide ... edit ... target ... }

Type rules

• DNA sequences: A/C/G/T only (uppercased).
• PAM patterns: IUPAC set A/C/G/T/N/R/Y/S/W/K/M/B/D/H/V.
• Locus start/end are 1-based, start <= end.
• ref(locus, offset) resolves to pos_1based = locus.start + offset and must fall inside the locus.
• pegRNA constraints: PBS length 1-20, RTT length 1-80.

Canonicalization

• All strings normalized (Unicode NFC + LF newlines).
• Lists are ordered deterministically (loci by name, edits by edit_id).
• Canonical JSON uses sorted keys and stable formatting.

CLI

• helix compile spec.hxspec outdir
  • Writes ir.json, manifest.json, SHA256SUMS.txt.
  • --deterministic pins created_at_utc for reproducible outputs.
• helix verify <outdir|manifest.json>
  • Recomputes IR hash + manifest identity hash and checks SHA256SUMS.
• helix diff a_manifest b_manifest
  • Semantic IR diff (edit-level changes).

Non-goals (v0.1)

• No execution (helix run) yet.
• No sweeps/templates.
• No reference sequence materialization.
• No GUI integration beyond DSL compilation.