Create evaluations import

curl --request POST \
  --url https://evalgate.com/api/evaluations/{id}/runs/import \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "results": [
    {
      "status": "failed",
      "testCaseId": 123,
      "output": "<string>",
      "input": "<string>",
      "costUsd": 123,
      "latencyMs": 123,
      "assertionsJson": {},
      "messages": [
        {}
      ],
      "toolCalls": [
        {}
      ],
      "behavior": "tool_use",
      "taskType": "<string>"
    }
  ],
  "environment": "dev",
  "ci": {
    "provider": "unknown",
    "actor": "<string>",
    "branch": "<string>",
    "repo": "<string>",
    "sha": "<string>",
    "pr": 123,
    "runUrl": "<string>"
  },
  "importClientVersion": "<string>",
  "trajectoryBaselines": [
    {
      "updatedAt": "<string>",
      "sampleSize": 123,
      "behavior": "tool_use",
      "stepDistribution": {
        "p50": 123,
        "p75": 123,
        "p90": 123
      },
      "toolCallDistribution": {
        "p50": 123,
        "p75": 123
      },
      "durationDistribution": {
        "p50": 123,
        "p75": 123
      },
      "commonToolSequences": [
        [
          "<string>"
        ]
      ],
      "redundancyProfile": {
        "mean": 123,
        "p90": 123
      },
      "version": "<string>",
      "model": "<string>",
      "taskType": "<string>",
      "parallelToolGroups": [
        [
          "<string>"
        ]
      ]
    }
  ],
  "checkReport": {}
}
'

import requests

url = "https://evalgate.com/api/evaluations/{id}/runs/import"

payload = {
    "results": [
        {
            "status": "failed",
            "testCaseId": 123,
            "output": "<string>",
            "input": "<string>",
            "costUsd": 123,
            "latencyMs": 123,
            "assertionsJson": {},
            "messages": [{}],
            "toolCalls": [{}],
            "behavior": "tool_use",
            "taskType": "<string>"
        }
    ],
    "environment": "dev",
    "ci": {
        "provider": "unknown",
        "actor": "<string>",
        "branch": "<string>",
        "repo": "<string>",
        "sha": "<string>",
        "pr": 123,
        "runUrl": "<string>"
    },
    "importClientVersion": "<string>",
    "trajectoryBaselines": [
        {
            "updatedAt": "<string>",
            "sampleSize": 123,
            "behavior": "tool_use",
            "stepDistribution": {
                "p50": 123,
                "p75": 123,
                "p90": 123
            },
            "toolCallDistribution": {
                "p50": 123,
                "p75": 123
            },
            "durationDistribution": {
                "p50": 123,
                "p75": 123
            },
            "commonToolSequences": [["<string>"]],
            "redundancyProfile": {
                "mean": 123,
                "p90": 123
            },
            "version": "<string>",
            "model": "<string>",
            "taskType": "<string>",
            "parallelToolGroups": [["<string>"]]
        }
    ],
    "checkReport": {}
}
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({
    results: [
      {
        status: 'failed',
        testCaseId: 123,
        output: '<string>',
        input: '<string>',
        costUsd: 123,
        latencyMs: 123,
        assertionsJson: {},
        messages: [{}],
        toolCalls: [{}],
        behavior: 'tool_use',
        taskType: '<string>'
      }
    ],
    environment: 'dev',
    ci: {
      provider: 'unknown',
      actor: '<string>',
      branch: '<string>',
      repo: '<string>',
      sha: '<string>',
      pr: 123,
      runUrl: '<string>'
    },
    importClientVersion: '<string>',
    trajectoryBaselines: [
      {
        updatedAt: '<string>',
        sampleSize: 123,
        behavior: 'tool_use',
        stepDistribution: {p50: 123, p75: 123, p90: 123},
        toolCallDistribution: {p50: 123, p75: 123},
        durationDistribution: {p50: 123, p75: 123},
        commonToolSequences: [['<string>']],
        redundancyProfile: {mean: 123, p90: 123},
        version: '<string>',
        model: '<string>',
        taskType: '<string>',
        parallelToolGroups: [['<string>']]
      }
    ],
    checkReport: {}
  })
};

fetch('https://evalgate.com/api/evaluations/{id}/runs/import', 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://evalgate.com/api/evaluations/{id}/runs/import",
  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([
    'results' => [
        [
                'status' => 'failed',
                'testCaseId' => 123,
                'output' => '<string>',
                'input' => '<string>',
                'costUsd' => 123,
                'latencyMs' => 123,
                'assertionsJson' => [
                                
                ],
                'messages' => [
                                [
                                                                
                                ]
                ],
                'toolCalls' => [
                                [
                                                                
                                ]
                ],
                'behavior' => 'tool_use',
                'taskType' => '<string>'
        ]
    ],
    'environment' => 'dev',
    'ci' => [
        'provider' => 'unknown',
        'actor' => '<string>',
        'branch' => '<string>',
        'repo' => '<string>',
        'sha' => '<string>',
        'pr' => 123,
        'runUrl' => '<string>'
    ],
    'importClientVersion' => '<string>',
    'trajectoryBaselines' => [
        [
                'updatedAt' => '<string>',
                'sampleSize' => 123,
                'behavior' => 'tool_use',
                'stepDistribution' => [
                                'p50' => 123,
                                'p75' => 123,
                                'p90' => 123
                ],
                'toolCallDistribution' => [
                                'p50' => 123,
                                'p75' => 123
                ],
                'durationDistribution' => [
                                'p50' => 123,
                                'p75' => 123
                ],
                'commonToolSequences' => [
                                [
                                                                '<string>'
                                ]
                ],
                'redundancyProfile' => [
                                'mean' => 123,
                                'p90' => 123
                ],
                'version' => '<string>',
                'model' => '<string>',
                'taskType' => '<string>',
                'parallelToolGroups' => [
                                [
                                                                '<string>'
                                ]
                ]
        ]
    ],
    'checkReport' => [
        
    ]
  ]),
  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://evalgate.com/api/evaluations/{id}/runs/import"

	payload := strings.NewReader("{\n  \"results\": [\n    {\n      \"status\": \"failed\",\n      \"testCaseId\": 123,\n      \"output\": \"<string>\",\n      \"input\": \"<string>\",\n      \"costUsd\": 123,\n      \"latencyMs\": 123,\n      \"assertionsJson\": {},\n      \"messages\": [\n        {}\n      ],\n      \"toolCalls\": [\n        {}\n      ],\n      \"behavior\": \"tool_use\",\n      \"taskType\": \"<string>\"\n    }\n  ],\n  \"environment\": \"dev\",\n  \"ci\": {\n    \"provider\": \"unknown\",\n    \"actor\": \"<string>\",\n    \"branch\": \"<string>\",\n    \"repo\": \"<string>\",\n    \"sha\": \"<string>\",\n    \"pr\": 123,\n    \"runUrl\": \"<string>\"\n  },\n  \"importClientVersion\": \"<string>\",\n  \"trajectoryBaselines\": [\n    {\n      \"updatedAt\": \"<string>\",\n      \"sampleSize\": 123,\n      \"behavior\": \"tool_use\",\n      \"stepDistribution\": {\n        \"p50\": 123,\n        \"p75\": 123,\n        \"p90\": 123\n      },\n      \"toolCallDistribution\": {\n        \"p50\": 123,\n        \"p75\": 123\n      },\n      \"durationDistribution\": {\n        \"p50\": 123,\n        \"p75\": 123\n      },\n      \"commonToolSequences\": [\n        [\n          \"<string>\"\n        ]\n      ],\n      \"redundancyProfile\": {\n        \"mean\": 123,\n        \"p90\": 123\n      },\n      \"version\": \"<string>\",\n      \"model\": \"<string>\",\n      \"taskType\": \"<string>\",\n      \"parallelToolGroups\": [\n        [\n          \"<string>\"\n        ]\n      ]\n    }\n  ],\n  \"checkReport\": {}\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://evalgate.com/api/evaluations/{id}/runs/import")
  .header("Authorization", "Bearer <token>")
  .header("Content-Type", "application/json")
  .body("{\n  \"results\": [\n    {\n      \"status\": \"failed\",\n      \"testCaseId\": 123,\n      \"output\": \"<string>\",\n      \"input\": \"<string>\",\n      \"costUsd\": 123,\n      \"latencyMs\": 123,\n      \"assertionsJson\": {},\n      \"messages\": [\n        {}\n      ],\n      \"toolCalls\": [\n        {}\n      ],\n      \"behavior\": \"tool_use\",\n      \"taskType\": \"<string>\"\n    }\n  ],\n  \"environment\": \"dev\",\n  \"ci\": {\n    \"provider\": \"unknown\",\n    \"actor\": \"<string>\",\n    \"branch\": \"<string>\",\n    \"repo\": \"<string>\",\n    \"sha\": \"<string>\",\n    \"pr\": 123,\n    \"runUrl\": \"<string>\"\n  },\n  \"importClientVersion\": \"<string>\",\n  \"trajectoryBaselines\": [\n    {\n      \"updatedAt\": \"<string>\",\n      \"sampleSize\": 123,\n      \"behavior\": \"tool_use\",\n      \"stepDistribution\": {\n        \"p50\": 123,\n        \"p75\": 123,\n        \"p90\": 123\n      },\n      \"toolCallDistribution\": {\n        \"p50\": 123,\n        \"p75\": 123\n      },\n      \"durationDistribution\": {\n        \"p50\": 123,\n        \"p75\": 123\n      },\n      \"commonToolSequences\": [\n        [\n          \"<string>\"\n        ]\n      ],\n      \"redundancyProfile\": {\n        \"mean\": 123,\n        \"p90\": 123\n      },\n      \"version\": \"<string>\",\n      \"model\": \"<string>\",\n      \"taskType\": \"<string>\",\n      \"parallelToolGroups\": [\n        [\n          \"<string>\"\n        ]\n      ]\n    }\n  ],\n  \"checkReport\": {}\n}")
  .asString();

require 'uri'
require 'net/http'

url = URI("https://evalgate.com/api/evaluations/{id}/runs/import")

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  \"results\": [\n    {\n      \"status\": \"failed\",\n      \"testCaseId\": 123,\n      \"output\": \"<string>\",\n      \"input\": \"<string>\",\n      \"costUsd\": 123,\n      \"latencyMs\": 123,\n      \"assertionsJson\": {},\n      \"messages\": [\n        {}\n      ],\n      \"toolCalls\": [\n        {}\n      ],\n      \"behavior\": \"tool_use\",\n      \"taskType\": \"<string>\"\n    }\n  ],\n  \"environment\": \"dev\",\n  \"ci\": {\n    \"provider\": \"unknown\",\n    \"actor\": \"<string>\",\n    \"branch\": \"<string>\",\n    \"repo\": \"<string>\",\n    \"sha\": \"<string>\",\n    \"pr\": 123,\n    \"runUrl\": \"<string>\"\n  },\n  \"importClientVersion\": \"<string>\",\n  \"trajectoryBaselines\": [\n    {\n      \"updatedAt\": \"<string>\",\n      \"sampleSize\": 123,\n      \"behavior\": \"tool_use\",\n      \"stepDistribution\": {\n        \"p50\": 123,\n        \"p75\": 123,\n        \"p90\": 123\n      },\n      \"toolCallDistribution\": {\n        \"p50\": 123,\n        \"p75\": 123\n      },\n      \"durationDistribution\": {\n        \"p50\": 123,\n        \"p75\": 123\n      },\n      \"commonToolSequences\": [\n        [\n          \"<string>\"\n        ]\n      ],\n      \"redundancyProfile\": {\n        \"mean\": 123,\n        \"p90\": 123\n      },\n      \"version\": \"<string>\",\n      \"model\": \"<string>\",\n      \"taskType\": \"<string>\",\n      \"parallelToolGroups\": [\n        [\n          \"<string>\"\n        ]\n      ]\n    }\n  ],\n  \"checkReport\": {}\n}"

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

{
  "runId": 123,
  "score": 123,
  "flags": true,
  "dashboardUrl": "<string>"
}

{
  "runId": 123,
  "score": 123,
  "flags": [
    "<string>"
  ],
  "dashboardUrl": "<string>"
}

{
  "error": {
    "message": "<string>",
    "details": "<unknown>",
    "requestId": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
  }
}

{
  "error": {
    "message": "<string>",
    "details": "<unknown>",
    "requestId": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
  }
}

{
  "error": {
    "message": "<string>",
    "details": "<unknown>",
    "requestId": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
  }
}

{
  "error": {
    "message": "<string>",
    "details": "<unknown>",
    "requestId": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
  }
}

{
  "error": {
    "message": "<string>",
    "details": "<unknown>",
    "requestId": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
  }
}

{
  "error": {
    "message": "<string>",
    "details": "<unknown>",
    "requestId": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
  }
}

{
  "error": {
    "message": "<string>",
    "details": "<unknown>",
    "requestId": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
  }
}