Ajouter ou remplacer des entrées de dataset

curl --request POST \
  --url https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
[
  {
    "messages": [
      {}
    ],
    "row_id": "<string>",
    "group_id": "<string>",
    "tools": [
      {
        "type": "function",
        "function": {
          "name": "<string>",
          "description": "<string>",
          "parameters": {},
          "strict": true
        }
      }
    ],
    "tool_choice": "<unknown>",
    "response_format": "<unknown>",
    "metadata": {},
    "provenance": {}
  }
]
'

import requests

url = "https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries"

payload = [
    {
        "messages": [{}],
        "row_id": "<string>",
        "group_id": "<string>",
        "tools": [
            {
                "type": "function",
                "function": {
                    "name": "<string>",
                    "description": "<string>",
                    "parameters": {},
                    "strict": True
                }
            }
        ],
        "tool_choice": "<unknown>",
        "response_format": "<unknown>",
        "metadata": {},
        "provenance": {}
    }
]
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.post(url, json=payload, headers=headers)

print(response.text)

const options = {
  method: 'POST',
  headers: {Authorization: 'Bearer <token>', 'Content-Type': 'application/json'},
  body: JSON.stringify([
    {
      messages: [{}],
      row_id: '<string>',
      group_id: '<string>',
      tools: [
        {
          type: 'function',
          function: {name: '<string>', description: '<string>', parameters: {}, strict: true}
        }
      ],
      tool_choice: '<unknown>',
      response_format: '<unknown>',
      metadata: {},
      provenance: {}
    }
  ])
};

fetch('https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries', options)
  .then(res => res.json())
  .then(res => console.log(res))
  .catch(err => console.error(err));

<?php

$curl = curl_init();

curl_setopt_array($curl, [
  CURLOPT_URL => "https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries",
  CURLOPT_RETURNTRANSFER => true,
  CURLOPT_ENCODING => "",
  CURLOPT_MAXREDIRS => 10,
  CURLOPT_TIMEOUT => 30,
  CURLOPT_HTTP_VERSION => CURL_HTTP_VERSION_1_1,
  CURLOPT_CUSTOMREQUEST => "POST",
  CURLOPT_POSTFIELDS => json_encode([
    [
        'messages' => [
                [
                                
                ]
        ],
        'row_id' => '<string>',
        'group_id' => '<string>',
        'tools' => [
                [
                                'type' => 'function',
                                'function' => [
                                                                'name' => '<string>',
                                                                'description' => '<string>',
                                                                'parameters' => [
                                                                                                                                
                                                                ],
                                                                'strict' => true
                                ]
                ]
        ],
        'tool_choice' => '<unknown>',
        'response_format' => '<unknown>',
        'metadata' => [
                
        ],
        'provenance' => [
                
        ]
    ]
  ]),
  CURLOPT_HTTPHEADER => [
    "Authorization: Bearer <token>",
    "Content-Type: application/json"
  ],
]);

$response = curl_exec($curl);
$err = curl_error($curl);

curl_close($curl);

if ($err) {
  echo "cURL Error #:" . $err;
} else {
  echo $response;
}

package main

import (
	"fmt"
	"strings"
	"net/http"
	"io"
)

func main() {

	url := "https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries"

	payload := strings.NewReader("[\n  {\n    \"messages\": [\n      {}\n    ],\n    \"row_id\": \"<string>\",\n    \"group_id\": \"<string>\",\n    \"tools\": [\n      {\n        \"type\": \"function\",\n        \"function\": {\n          \"name\": \"<string>\",\n          \"description\": \"<string>\",\n          \"parameters\": {},\n          \"strict\": true\n        }\n      }\n    ],\n    \"tool_choice\": \"<unknown>\",\n    \"response_format\": \"<unknown>\",\n    \"metadata\": {},\n    \"provenance\": {}\n  }\n]")

	req, _ := http.NewRequest("POST", url, payload)

	req.Header.Add("Authorization", "Bearer <token>")
	req.Header.Add("Content-Type", "application/json")

	res, _ := http.DefaultClient.Do(req)

	defer res.Body.Close()
	body, _ := io.ReadAll(res.Body)

	fmt.Println(string(body))

}

HttpResponse<String> response = Unirest.post("https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries")
  .header("Authorization", "Bearer <token>")
  .header("Content-Type", "application/json")
  .body("[\n  {\n    \"messages\": [\n      {}\n    ],\n    \"row_id\": \"<string>\",\n    \"group_id\": \"<string>\",\n    \"tools\": [\n      {\n        \"type\": \"function\",\n        \"function\": {\n          \"name\": \"<string>\",\n          \"description\": \"<string>\",\n          \"parameters\": {},\n          \"strict\": true\n        }\n      }\n    ],\n    \"tool_choice\": \"<unknown>\",\n    \"response_format\": \"<unknown>\",\n    \"metadata\": {},\n    \"provenance\": {}\n  }\n]")
  .asString();

require 'uri'
require 'net/http'

url = URI("https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries")

http = Net::HTTP.new(url.host, url.port)
http.use_ssl = true

request = Net::HTTP::Post.new(url)
request["Authorization"] = 'Bearer <token>'
request["Content-Type"] = 'application/json'
request.body = "[\n  {\n    \"messages\": [\n      {}\n    ],\n    \"row_id\": \"<string>\",\n    \"group_id\": \"<string>\",\n    \"tools\": [\n      {\n        \"type\": \"function\",\n        \"function\": {\n          \"name\": \"<string>\",\n          \"description\": \"<string>\",\n          \"parameters\": {},\n          \"strict\": true\n        }\n      }\n    ],\n    \"tool_choice\": \"<unknown>\",\n    \"response_format\": \"<unknown>\",\n    \"metadata\": {},\n    \"provenance\": {}\n  }\n]"

response = http.request(request)
puts response.read_body

{
  "dry_run": true,
  "created": 123,
  "updated": 123,
  "revision": 123,
  "entry_counts": {
    "total": 123,
    "train": 123,
    "val": 123
  },
  "group_conflicts": [
    "<string>"
  ]
}

{
  "error": {
    "message": "Project 'missing' not found in entity 'your-team'",
    "type": "not_found"
  }
}

Ajouter ou remplacer des entrées de dataset

Écrit jusqu’à 5,000 lignes de chat OpenAI en une seule transaction. Envoyez un tableau JSON ou les mêmes lignes au format NDJSON (une ligne de données par ligne de texte) avec Content-Type: application/x-ndjson. Toute ligne invalide entraîne le rejet de la requête et est signalée par son index ; la ligne N du fichier NDJSON correspond à l’index N - 1. mode=append rejette un row_id qui existe déjà (409 entry_exists) ; mode=upsert remplace la ligne en conservant les valeurs stockées de group_id, split, metadata et provenance pour les champs qu’elle omet. Les nouvelles entrées sont affectées par défaut à train et à leur propre groupe. Une écriture qui placerait un groupe dans plusieurs partitions renvoie 409 group_split_conflict. L’écriture crée une nouvelle révision du dataset et marque les réétiquetages et évaluations terminés comme obsolètes.

POST

projects

{alias}

datasets

{datasetId}

entries

Ajouter ou remplacer des entrées de dataset

curl --request POST \
  --url https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
[
  {
    "messages": [
      {}
    ],
    "row_id": "<string>",
    "group_id": "<string>",
    "tools": [
      {
        "type": "function",
        "function": {
          "name": "<string>",
          "description": "<string>",
          "parameters": {},
          "strict": true
        }
      }
    ],
    "tool_choice": "<unknown>",
    "response_format": "<unknown>",
    "metadata": {},
    "provenance": {}
  }
]
'

import requests

url = "https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries"

payload = [
    {
        "messages": [{}],
        "row_id": "<string>",
        "group_id": "<string>",
        "tools": [
            {
                "type": "function",
                "function": {
                    "name": "<string>",
                    "description": "<string>",
                    "parameters": {},
                    "strict": True
                }
            }
        ],
        "tool_choice": "<unknown>",
        "response_format": "<unknown>",
        "metadata": {},
        "provenance": {}
    }
]
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.post(url, json=payload, headers=headers)

print(response.text)

const options = {
  method: 'POST',
  headers: {Authorization: 'Bearer <token>', 'Content-Type': 'application/json'},
  body: JSON.stringify([
    {
      messages: [{}],
      row_id: '<string>',
      group_id: '<string>',
      tools: [
        {
          type: 'function',
          function: {name: '<string>', description: '<string>', parameters: {}, strict: true}
        }
      ],
      tool_choice: '<unknown>',
      response_format: '<unknown>',
      metadata: {},
      provenance: {}
    }
  ])
};

fetch('https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries', options)
  .then(res => res.json())
  .then(res => console.log(res))
  .catch(err => console.error(err));

<?php

$curl = curl_init();

curl_setopt_array($curl, [
  CURLOPT_URL => "https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries",
  CURLOPT_RETURNTRANSFER => true,
  CURLOPT_ENCODING => "",
  CURLOPT_MAXREDIRS => 10,
  CURLOPT_TIMEOUT => 30,
  CURLOPT_HTTP_VERSION => CURL_HTTP_VERSION_1_1,
  CURLOPT_CUSTOMREQUEST => "POST",
  CURLOPT_POSTFIELDS => json_encode([
    [
        'messages' => [
                [
                                
                ]
        ],
        'row_id' => '<string>',
        'group_id' => '<string>',
        'tools' => [
                [
                                'type' => 'function',
                                'function' => [
                                                                'name' => '<string>',
                                                                'description' => '<string>',
                                                                'parameters' => [
                                                                                                                                
                                                                ],
                                                                'strict' => true
                                ]
                ]
        ],
        'tool_choice' => '<unknown>',
        'response_format' => '<unknown>',
        'metadata' => [
                
        ],
        'provenance' => [
                
        ]
    ]
  ]),
  CURLOPT_HTTPHEADER => [
    "Authorization: Bearer <token>",
    "Content-Type: application/json"
  ],
]);

$response = curl_exec($curl);
$err = curl_error($curl);

curl_close($curl);

if ($err) {
  echo "cURL Error #:" . $err;
} else {
  echo $response;
}

package main

import (
	"fmt"
	"strings"
	"net/http"
	"io"
)

func main() {

	url := "https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries"

	payload := strings.NewReader("[\n  {\n    \"messages\": [\n      {}\n    ],\n    \"row_id\": \"<string>\",\n    \"group_id\": \"<string>\",\n    \"tools\": [\n      {\n        \"type\": \"function\",\n        \"function\": {\n          \"name\": \"<string>\",\n          \"description\": \"<string>\",\n          \"parameters\": {},\n          \"strict\": true\n        }\n      }\n    ],\n    \"tool_choice\": \"<unknown>\",\n    \"response_format\": \"<unknown>\",\n    \"metadata\": {},\n    \"provenance\": {}\n  }\n]")

	req, _ := http.NewRequest("POST", url, payload)

	req.Header.Add("Authorization", "Bearer <token>")
	req.Header.Add("Content-Type", "application/json")

	res, _ := http.DefaultClient.Do(req)

	defer res.Body.Close()
	body, _ := io.ReadAll(res.Body)

	fmt.Println(string(body))

}

HttpResponse<String> response = Unirest.post("https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries")
  .header("Authorization", "Bearer <token>")
  .header("Content-Type", "application/json")
  .body("[\n  {\n    \"messages\": [\n      {}\n    ],\n    \"row_id\": \"<string>\",\n    \"group_id\": \"<string>\",\n    \"tools\": [\n      {\n        \"type\": \"function\",\n        \"function\": {\n          \"name\": \"<string>\",\n          \"description\": \"<string>\",\n          \"parameters\": {},\n          \"strict\": true\n        }\n      }\n    ],\n    \"tool_choice\": \"<unknown>\",\n    \"response_format\": \"<unknown>\",\n    \"metadata\": {},\n    \"provenance\": {}\n  }\n]")
  .asString();

require 'uri'
require 'net/http'

url = URI("https://forge.coreweave.com/api/distillation/v1/projects/{alias}/datasets/{datasetId}/entries")

http = Net::HTTP.new(url.host, url.port)
http.use_ssl = true

request = Net::HTTP::Post.new(url)
request["Authorization"] = 'Bearer <token>'
request["Content-Type"] = 'application/json'
request.body = "[\n  {\n    \"messages\": [\n      {}\n    ],\n    \"row_id\": \"<string>\",\n    \"group_id\": \"<string>\",\n    \"tools\": [\n      {\n        \"type\": \"function\",\n        \"function\": {\n          \"name\": \"<string>\",\n          \"description\": \"<string>\",\n          \"parameters\": {},\n          \"strict\": true\n        }\n      }\n    ],\n    \"tool_choice\": \"<unknown>\",\n    \"response_format\": \"<unknown>\",\n    \"metadata\": {},\n    \"provenance\": {}\n  }\n]"

response = http.request(request)
puts response.read_body

{
  "dry_run": true,
  "created": 123,
  "updated": 123,
  "revision": 123,
  "entry_counts": {
    "total": 123,
    "train": 123,
    "val": 123
  },
  "group_conflicts": [
    "<string>"
  ]
}

{
  "error": {
    "message": "Project 'missing' not found in entity 'your-team'",
    "type": "not_found"
  }
}

Autorisations

Authorization

string

header

requis

Une clé API wandb ou un JWT wandb. Les sessions du navigateur sont échangées par le backend de l’application web Model Distillation ; les capacités des tâches sont réservées à un usage interne.

En-têtes

Wandb-Entity

string

entity personnel ou d’équipe accessible. Omettez ce champ pour utiliser l’entity par défaut de l’utilisateur authentifié.

Paramètres de chemin

alias

string

requis

Pattern: ^[a-z0-9-]{1,64}$

datasetId

string<uuid>

requis

Paramètres de requête

mode

enum<string>

défaut:append

append rejette les lignes dont le row_id existe déjà ; upsert les remplace.

Options disponibles:

append,

upsert

dry_run

boolean

défaut:false

Validez la modification et indiquez son résultat, puis annulez-la.

Corps

Required array length: 1 - 5000 elements

messages

object[]

requis

Required array length: 2 - 1000 elements

Show child attributes

row_id

string

Identité stable au sein du dataset, utilisée par les opérations upsert pour retrouver les entrées correspondantes. Par défaut, une empreinte des messages et des champs de la requête.

Required string length: 1 - 512

group_id

string

Lignes qui doivent appartenir à la même partition, comme les tours d'une même conversation. Par défaut, row_id.

Required string length: 1 - 512

split

enum<string>

train, ou val pour exclure la ligne de l'entraînement et la réserver à l'évaluation. Les nouvelles lignes sont affectées par défaut à train.

Options disponibles:

train,

val

tools

object[]

Maximum array length: 128

Show child attributes

tool_choice

any

response_format

any

metadata

object

Attributs utilisables pour le filtrage, comme l'utilisateur ou le segment de clientèle.

Show child attributes

provenance

object

Origine de la ligne, comme le système source et l'identifiant de trace.

Show child attributes

Réponse

Les entrées ont été écrites, ou le seraient dans le cadre d'une simulation à blanc.

dry_run

boolean

requis

created

integer

requis

Lignes devenues de nouvelles entrées.

updated

integer

requis

Entrées existantes remplacées par les lignes.

revision

integer

requis

La révision du dataset après l’écriture ; inchangée lors d’une simulation à blanc.

entry_counts

object

requis

Show child attributes

group_conflicts

string[]

requis

Groupes présents dans plusieurs partitions. Une écriture réelle comportant des conflits est rejetée.

Dernière modification le 8 octobre 2026