Prompt Splicify
with a design request.
Describe what you want to build in plain language, attach any GenBank files you have, and Splicify picks the best workflow. The chat is the front door for designing plasmids; the plasmid viewer is where you inspect and edit them. This page walks through both.
How a chat request flows
Every prompt you send hits a single endpoint, /api/chat, that runs five deterministic steps before any cloning workflow executes. You don’t need to know the internals to use it — but a quick look at the path explains why certain phrasings get certain results.
The five steps behind every prompt
Phrasing that triggers each workflow
The classifier reads your prompt and the files you attach, then picks an intent. You don’t have to memorise the rules — but if you want a specific workflow, phrase the request to match.
| Workflow | Triggers | What you upload |
|---|---|---|
| Gibson assembly | “Gibson assembly”, “assemble these fragments”, two or more sequences in the prompt | None — fragments are pasted inline |
| Gateway cloning | “Gateway”, “BP reaction”, “LR reaction”, “attB/attP/attL/attR”, “pDONR” | Vector and/or insert .gb files |
| Golden Gate primers | “Golden Gate primers”, “BsaI/BsmBI primers”, multi-fragment language with “+” separators | Optional — sequences inline |
| sgRNA Golden Gate | “sgRNA” or “gRNA” + “Golden Gate” / “BsmBI” / “lentiCRISPR” / “pX330” | CRISPR vector .gb file |
| Restriction cloning | “clone X into Y”, “restriction enzyme/digest cloning”, named Type II enzymes (EcoRI, HindIII, …) | Backbone .gb file (optional) |
| Site-directed mutagenesis | “mutate”, “delete the X”, “Y66H”, “point mutation”, “remove the NLS” | Plasmid .gb file (target or inventory) |
| Annotate plasmid | “annotate”, “identify features” | Plasmid .gb file |
| Describe a plasmid | “make / build / design a vector for X”, “I need a plasmid that …” | None |
Supported workflows
Example 1 — Gibson from fragments
Paste fragment sequences directly in the prompt. Splicify designs primers with homology overhangs at every junction and returns a full assembly + protocol.
What you'll get back
- Circular construct visualisation showing fragments, overlaps, and primers
- Primer CSV — sequences, Tm, GC%, secondary structures, quality scores
- Overlap CSV — sequence, length, Tm, quality scores
- Junction table with overlap homology and Tm per junction
- Annotated GenBank of the assembled construct
- Expert explanation — potential issues and optimisation tips
Example 2 — Golden Gate primers
Multi-fragment Golden Gate primer design: Splicify picks orthogonal 4 bp overhangs, tails BsaI / BsmBI / BbsI sites onto every primer, and verifies no Type IIs site appears inside any fragment.
Example 3 — Restriction cloning
Upload your backbone and ask Splicify to clone an insert into it. The classifier recognises “clone X into Y” phrasing, and naming the enzymes pins the digest pair. Without explicit enzymes, Splicify scans the backbone’s MCS for a unique-cutter pair that is also absent from the insert.
[attached: pUC19_empty.gb]
What you'll get back
- Selected enzyme pair, scored on MCS unique-cutter check + insert internal-cut check
- Forward and reverse primers with the chosen sites tailed on
- Synthesis-first manifest when no template exists for the insert (gBlock + ligation, no PCR)
- Diagnostic restriction map of the assembled product
- Annotated GenBank of the assembled plasmid (LLM-annotated, not the upload’s original features)
Example 4 — Site-directed mutagenesis
Upload a plasmid and describe the change. SDM understands amino-acid notation (Y66H, S65T, D10A), feature deletions (“delete the His-tag”, “remove the NLS”), terminus insertions (“insert FLAG at the N-terminus of eGFP”), and position-based edits (“delete bp 100–150”). Before resolving the target, Splicify re-annotates the input plasmid with the full pipeline so motifs (His-tag, FLAG, HA, V5, kozak, NLS, P2A/T2A/E2A/F2A) and CDS submodules are searchable even when they’re not in the uploaded file’s features.
[attached: pCMV_6xHis_GFP_demo.gb]
Example 5 — sgRNA Golden Gate
Annealed-oligo cloning of a 20-bp guide into a CRISPR vector. The handler picks the right Type IIs enzyme automatically (BsmBI for lentiCRISPR v2, BbsI for pX330) and shows the assembled plasmid annotated with the full pipeline — every feature on the backbone is visible, not just the guide cassette.
[attached: lentiCRISPR_v2_unannotated.gb]
Example 6 — Gateway cloning
Splicify scans uploaded plasmids for att sites to decide which is a donor (attP) and which carries the insert. The BP/LR product is always emitted as a downloadable GenBank, and the response viz shows the recombined plasmid with every feature annotated.
[attached: pDONR221.gb]
Example 7 — Describe a plasmid
When you describe a plasmid in plain language without naming specific fragments, Splicify routes the request to the describe-a-plasmid handler. It builds an abstract spec from your description (host, modules, topology), runs a semantic search over a 7,256-plasmid corpus to find the closest existing foundation, and proposes deterministic edits to make it match your spec — primer-tail edits for inserts under 40 bp, synthesis fragments for 40 bp and above.
What comes back
<plasmid_id>_foundation.gb— the top-hit plasmid from the corpus, freshly annotateddesigned_plasmid.gb— the foundation edited to match your spec, re-annotatedworkflow_trace.json— ranked retrieval hits, edit operations with rationale, and any spec gaps deferred to the orchestrator- Plasmid map of the designed plasmid in the viewer
The plasmid viewer
Every workflow that returns a plasmid renders it in the same viewer panel below the chat. The viewer is interactive — drag to select, double-click a feature to inspect it, and use the toolbar to add, edit, or import annotations. Selections persist across the circular and linear views.
Toolbar buttons
Selecting, editing, navigating
- Drag in either viewer to select a range — the bp count and span are shown next to the toolbar.
- Press
Deletewith a selection to remove that region from the sequence. - Double-click any feature to open its gene card: type, location, length, description, KB source, sequence (copy-pasteable).
- Wheel-scroll the linear viewer for fine navigation; click anywhere on the circular to jump.
- Cloning-feature glyphs (Type II / IIs brackets, Gateway att crossovers, PCR warnings) can be filtered by family + cutter count from the Cloning Features menu.
Tips that change the outcome
- Naming the polymerase (“Q5”, “SuperFi II”, “KOD One”) tunes primer Tm targets and gives you the matching protocol.
- Saying “linear” vs “circular” pins Gibson topology; otherwise circular is assumed.
- For SDM, prefer notation that’s unambiguous:
Y66H in eGFPbeatschange residue 66 of GFP. - For restriction cloning, naming both enzymes (“EcoRI and HindIII”) skips the auto-pick step.
- For describe-a-plasmid, mention the host (“mammalian”, “lentiviral”, “bacterial”) — it biases retrieval toward the right corpus subset.
- Files attached as inventory (multiple .gb) are scored against a target; files attached as target are the thing you’re modifying.