SpliceCraft v1.0.33

New features

• Fetch proteins from NCBI by accession. The Operon Design tab's protein library gains a Fetch button — type an accession (e.g. AAK98552 for luxC) and SpliceCraft pulls the protein straight from NCBI into your collection, the same way it fetches plasmids. Building an operon from published parts (like the luxCDABE + frp bioluminescence genes) is now a few keystrokes.

Bug fixes

• RNA folding now matches ViennaRNA exactly on large loops too. An internal rounding step diverged by 0.01 kcal/mol on hairpin / internal / bulge loops larger than 30 nt (rare, and below the resolution of the original validation set). The folder now reproduces ViennaRNA to the cent at every loop size.

Hardening

• A full adversarial audit of the RNA / RBS / operon engine. The structure evaluator now rejects chemically-impossible structures (non-canonical pairs, sub-3-nt hairpin loops) with a clear error instead of crashing or returning a nonsense energy. The rbs-strength / assemble-operon / design-rbs agent endpoints now always return valid JSON and reject infinite / not-a-number targets, over-long sequences, and boolean indices with a clean 400 rather than a 200-with-Infinity or a 500. Editing an operon (adding a gene, changing the promoter or terminator) now correctly invalidates the previous assembly, so an export can never ship a stale sequence. And the folding engine's shared model is built under a lock.