Models

List the models you can finetune

The finetuning subset of GET /tools: the tools this account can train a model with, i.e. exactly the /tools entries with finetune: true, and every one of them. It is /tools run with a filter, so auth, query parameters, gating (feature flags, tier, organization restrictions) and each item are /tools' own: an item here is byte-for-byte the /tools item for the same tool and caller, so finetune is always true. Finetuning tools are NOT removed from /tools. custom=true behaves as on /tools (custom tools are not listed by either), so it returns an empty array.

Train one with POST /finetune (or POST /finetune-batch), naming the item's name in model. This is NOT GET /finetuned-models, which lists the models you have already TRAINED (and runs them by their inferenceType); /models lists which tools you can train them WITH.

get/api/models

Request

  • Base URL: https://app.tamarind.bio
  • URL: https://app.tamarind.bio/api/models
  • Auth: API key in header x-api-key

Response

The finetuning tools available to this account

namestring

The value to send as type when submitting.

finetuneboolean

True for a finetuning tool, which is submitted with POST /finetune (name it in model) rather than /submit-job.

displayNamestring
descriptionstring
githubstring
paperstring
outputTypesstring[]

What the tool declares it produces — molecular (pdb, sequence, sdf, smiles) alongside file and score types (csv, score, cif, …). A few tools use compound values such as pdb-list, so match by containment rather than equality.

This describes the ARTIFACTS a run leaves behind. To chain stages, match on taskType and the molecule types: a scoring stage passes its input type through even though its outputTypes says score. Absent means the tool declares nothing, which is unknown rather than "produces nothing"; and a type here is not proof the tool generated it, since a scoring tool can echo its input.

taskTypestring

The tool's pipeline task category (structure-prediction, inverse-folding, score, …) — the fact pipelines chain on. Absent for a tool that declares none.

filterMetricsstring[]

Metric names accepted in a pipeline stage's filter settings. Deliberately narrower than outputs.columns — /submit-pipeline rejects a filter on a column that is not filterable. Absent for tools whose filters are not validated.

Example response

[
  {
    "name": "alphafold",
    "settings": [
      {
        "extension": [
          "pdb",
          "cif"
        ]
      }
    ],
    "outputTypes": [
      "pdb"
    ],
    "taskType": "structure-prediction"
  }
]

Changes

No recorded changes to this operation across all 1 revision of this API.