Batches
CSV reference
This page lists what a batch CSV can contain and the rules the importer checks before it prepares jobs. The selected tool and primary operation decide which columns are available; Column reference and examples in the batch form shows the current names, and Template downloads them as headers.
Column values
Each column has one purpose:
| Column | Value |
|---|---|
name | Optional job name, up to 200 characters. Blank names receive a generated name. |
Numbered input, such as protein_1 | Accepted text or file:filename. The number identifies a member within one job. |
.format | An accepted format for that input. Declaring a format does not convert the data. |
.kind | An accepted molecule type. Use the correct protein, DNA, RNA or ligand input. |
.label | Optional label for the input member. |
setting.<id> | A row-specific setting. Blank cells inherit shared settings. |
Use decimal points for numbers, true/false or 1/0 for booleans, and exact option values for dropdown settings. Multiple-choice settings use a JSON array, and settings that expect JSON text keep that representation.
In row mode, optional empty inputs are omitted and required inputs cannot be blank. For a supported operation that needs no inputs, named rows create separate jobs.
Molecular attributes
These attributes apply only to supported Boltz-2 inputs:
| Suffix | Supported inputs | Value |
|---|---|---|
.chainId | Protein, DNA, RNA, SMILES and CCD ligands | Chain identifier consistent with the experiment |
.copies | Same inputs | Integer from 1 to 10 |
.cyclic | Protein, DNA and RNA | true or false |
.modifications | Protein, DNA and RNA | JSON array, for example [{"position":2,"ccd":"MSE"}] |
.msa | Protein | Exact attached A3M reference, such as file:target.a3m |
An attribute needs a value in its matching input column. The importer checks the JSON shape, not the chemistry, so positions and residue identities still have to match the input.
File formatting
Save the spreadsheet as UTF-8 CSV or TSV, using comma, semicolon or tab separators. Export Excel workbooks first; native .xlsx files are not imported.
Quote any cell that contains a separator or line break, and double the quote characters inside it. A modification array becomes this cell:
"[{""position"":2,""ccd"":""MSE""}]"Headers must be nonempty and unique. Every row needs the same number of cells, so keep the separators for empty cells at the end of a row. Empty lines are ignored, and a UTF-8 BOM, Windows line endings and quoted multiline FASTA content are all accepted.
Import issues
The preview lists every problem before you can submit, and Download all … issues saves the full list. Invalid rows are never skipped silently.
| Issue | What to do |
|---|---|
| Unknown column | Map it to an accepted column or choose Ignore. Each target can be mapped only once. |
| Missing or ambiguous file | Add the file to the matching input, with a unique and exact filename. MSA references need unambiguous names too. |
| Unexpected job count | Check the pairing mode, especially how it treats empty cells. |
| Invalid row or setting | Fix the reported input, format, value or tool constraint. |
| Repeated names or inputs | Duplicates stay as separate paid jobs, with a warning. |
Size limits
An import supports up to 250 data rows, 250 generated jobs, 2,048 columns and a 5 MB CSV. The prepared run also has a 10 MB input-data budget, which repeated large files use up quickly.
Complex layouts and many different setting combinations can hit the run's configuration limits before the row limit. When that happens, split the batch or reduce the number of distinct settings. Each tool's own limits and your plan's limits still apply, and the importer does not accept ZIP files or remote URLs.