What format should we add next?

[amacsmith]

What graph format should we add next?

Five formats ship today: understand-anything@1, gitnexus@1, code-review-graph@1, bundle@1, and generic@1. The bar for adding a new first-class format is low — a JSON Schema, an ok and bad fixture, and a real graph from a real producer. The question is which producer is worth a schema slot next.

Candidates we've considered but haven't shipped:

• ts-morph dump — JS/TS specific; per-file symbol table with type info. Strong fit for agents working in TypeScript monorepos.
• ctags JSON — universal, ancient, exhaustive. Lowest-effort producer; almost every language has a generator. Trade-off: shallow semantics.
• Semgrep findings — pattern-match results rather than a structural graph; useful for "find all places that match X" agent queries.
• Sourcegraph SCIP — language-agnostic precise indexes (the successor to LSIF). High fidelity, high producer cost.

A few more were floated and skipped: tree-sitter S-expression dumps (too tied to a single grammar), Doxygen XML (heavy), language-server symbol exports (no portable serialization).

If you have a strong preference, drop a thumbs-up on the comment for that option below — or comment with a real producer + sample graph if you want to lobby for one not on the list. We'll prioritize whichever has both demand and a producer that already emits the format in the wild.

Schema-contribution flow: CONTRIBUTING.md.