docs: sync API specs from openprotein-api

source/_static/js/foldSpec.js

@@ -240,7 +240,7 @@ const foldSpec = {
         tags: ["fold requests", "esmfold2"],
         summary: "ESMFold2",
         description:
-          "Create structure prediction using ESMFold2, an all-atom structure prediction\nmodel. Folds protein/DNA/RNA/ligand complexes, optionally conditioned on an MSA.\n\nArgs:\n  - `sequences`: List of chain/molecule entities in the input. Each entry describes a protein, nucleic acid, or ligand, including its sequence, identifier(s), and optional attributes such as msa_id, SMILES string, CCD code.\n    - `msa_id` should refer to the id of an msa job which included this protein as a query, or `null` for single sequence mode.\n    - `smiles` and `ccd` are mutually exclusive for ligands.\n  - `diffusion_samples`: Number of diffusion samples to use. Controls how many independent structure samples are generated per input. Default is 1.\n  - `num_steps`: Number of sampling steps to use. Sets the number of steps in the diffusion process for each sample. Default is 200.\n  - `num_recycles`: Number of recycling steps to use. Determines how many times the model refines its prediction iteratively. Default is 3.\n  - `seed`: Random seed for reproducible sampling. `null` lets the system decide.",
+          "Create structure prediction using ESMFold2, an all-atom structure prediction\nmodel. Folds protein/DNA/RNA/ligand complexes, optionally conditioned on an MSA.\n\nArgs:\n  - `sequences`: List of chain/molecule entities in the input. Each entry describes a protein, nucleic acid, or ligand, including its sequence, identifier(s), and optional attributes such as msa_id, SMILES string, CCD code.\n    - `msa_id` should refer to the id of an msa job which included this protein as a query, or `null` for single sequence mode.\n    - `smiles` and `ccd` are mutually exclusive for ligands.\n  - `diffusion_samples`: Number of diffusion samples to use. Controls how many independent structure samples are generated per input. Default is 1.\n  - `num_steps`: Number of sampling steps to use. Sets the number of steps in the diffusion process for each sample. Default is 100.\n  - `num_recycles`: Number of recycling steps to use. Determines how many times the model refines its prediction iteratively. Default is 3.\n  - `seed`: Random seed for reproducible sampling. `null` lets the system decide.",
         requestBody: {
           description: "Request for structure prediction.",
           content: {
@@ -328,7 +328,7 @@ const foldSpec = {
         tags: ["fold requests", "esmfold2"],
         summary: "ESMFold2-Fast",
         description:
-          "Create structure prediction using ESMFold2-Fast, an inference-optimized\nsingle-sequence variant of ESMFold2 whose folding trunk has half the depth\n(24 vs 48 layers). Folds protein/DNA/RNA/ligand complexes.\n\nUnlike ESMFold2, ESMFold2-Fast is single-sequence only and does not accept\nan MSA (`msa_id`).\n\nArgs:\n  - `sequences`: List of chain/molecule entities in the input. Each entry describes a protein, nucleic acid, or ligand, including its sequence, identifier(s), and optional attributes such as SMILES string, CCD code.\n    - `smiles` and `ccd` are mutually exclusive for ligands.\n  - `diffusion_samples`: Number of diffusion samples to use. Controls how many independent structure samples are generated per input. Default is 1.\n  - `num_steps`: Number of sampling steps to use. Sets the number of steps in the diffusion process for each sample. Default is 200.\n  - `num_recycles`: Number of recycling steps to use. Determines how many times the model refines its prediction iteratively. Default is 3.\n  - `seed`: Random seed for reproducible sampling. `null` lets the system decide.",
+          "Create structure prediction using ESMFold2-Fast, an inference-optimized\nsingle-sequence variant of ESMFold2 whose folding trunk has half the depth\n(24 vs 48 layers). Folds protein/DNA/RNA/ligand complexes.\n\nUnlike ESMFold2, ESMFold2-Fast is single-sequence only and does not accept\nan MSA (`msa_id`).\n\nArgs:\n  - `sequences`: List of chain/molecule entities in the input. Each entry describes a protein, nucleic acid, or ligand, including its sequence, identifier(s), and optional attributes such as SMILES string, CCD code.\n    - `smiles` and `ccd` are mutually exclusive for ligands.\n  - `diffusion_samples`: Number of diffusion samples to use. Controls how many independent structure samples are generated per input. Default is 1.\n  - `num_steps`: Number of sampling steps to use. Sets the number of steps in the diffusion process for each sample. Default is 100.\n  - `num_recycles`: Number of recycling steps to use. Determines how many times the model refines its prediction iteratively. Default is 3.\n  - `seed`: Random seed for reproducible sampling. `null` lets the system decide.",
         requestBody: {
           description: "Request for structure prediction.",
           content: {
@@ -2006,7 +2006,7 @@ const foldSpec = {
             type: "integer",
             description: "Number of sampling steps to use.",
             minimum: 1,
-            default: 200,
+            default: 100,
           },
           num_recycles: {
             type: "integer",
@@ -2047,7 +2047,7 @@ const foldSpec = {
             ],
           ],
           diffusion_samples: 1,
-          num_steps: 200,
+          num_steps: 100,
           num_recycles: 3,
         },
       },
@@ -2091,7 +2091,7 @@ const foldSpec = {
             type: "integer",
             description: "Number of sampling steps to use.",
             minimum: 1,
-            default: 200,
+            default: 100,
           },
           num_recycles: {
             type: "integer",
@@ -2131,7 +2131,7 @@ const foldSpec = {
             ],
           ],
           diffusion_samples: 1,
-          num_steps: 200,
+          num_steps: 100,
           num_recycles: 3,
         },
       },

source/_static/js/predictorSpec.js

@@ -1450,7 +1450,15 @@ const predictorSpec = {
           type: {
             type: "string",
             description: "Type of kernel to use with GP.",
-            enum: ["linear", "rbf", "matern21", "matern32"],
+            enum: [
+              "linear",
+              "rbf",
+              "matern12",
+              "matern32",
+              "matern52",
+              "periodic",
+              "rational_quadratic",
+            ],
             example: "rbf",
           },
           multitask: {
@@ -1459,6 +1467,20 @@ const predictorSpec = {
             example: true,
             default: false,
           },
+          period: {
+            type: "number",
+            format: "double",
+            description:
+              "Period length for the periodic kernel. Only valid when type is periodic.",
+            example: 1,
+          },
+          alpha: {
+            type: "number",
+            format: "double",
+            description:
+              "Scale-mixture parameter for the rational_quadratic kernel; must be > 0. Only valid when type is rational_quadratic.",
+            example: 1,
+          },
         },
       },
       TrainRequestGP: {