Guía completa de comandos curl para servicios de modelos de IA

Comandos para veirficar la conectividad con el servicio de modelos:

Bash

export BASE_URL="http://0.0.0.0:8000/v1"
export API_KEY="EMPTY"
export MODEL="gpt-4.1-mini"

curl -sS --fail --connect-timeout 10 --max-time 60 \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  "${BASE_URL}/models" | python -m json.tool

PowerShell

$env:BASE_URL="http://0.0.0.0:8000/v1"
$env:API_KEY="EMPTY"
$env:MODEL="gpt-4.1-mini"

curl.exe -sS --fail --connect-timeout 10 --max-time 60 `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  "$env:BASE_URL/models" | python -m json.tool

LLM: /v1/chat/completions (no streaming)

Ejemplo básico

Bash

curl -sS --fail --connect-timeout 10 --max-time 120 \
  "${BASE_URL}/chat/completions" \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"${MODEL}\",
    \"messages\": [
      {\"role\":\"system\",\"content\":\"Eres un asistente útil.\"},
      {\"role\":\"user\",\"content\":\"Di hola en una sola oración.\"}
    ]
  }" | python -m json.tool

PowerShell

$body = @{
  model = $env:MODEL
  messages = @(
    @{ role="system"; content="Eres un asistente útil." },
    @{ role="user"; content="Di hola en una sola oración." }
  )
} | ConvertTo-Json -Depth 20

curl.exe -sS --fail --connect-timeout 10 --max-time 120 `
  "$env:BASE_URL/chat/completions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

Ejemplo con parámetros avanzados

Bash

curl -sS --fail --connect-timeout 10 --max-time 120 \
  "${BASE_URL}/chat/completions" \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"${MODEL}\",
    \"temperature\": 0.2,
    \"top_p\": 0.9,
    \"max_tokens\": 256,
    \"seed\": 42,
    \"stop\": [\"\\n\\n###\"],
    \"messages\": [
      {\"role\":\"system\",\"content\":\"Responde brevemente.\"},
      {\"role\":\"user\",\"content\":\"Resume qué es un mutex.\"}
    ]
  }" | python -m json.tool

PowerShell

$body = @{
  model = $env:MODEL
  temperature = 0.2
  top_p = 0.9
  max_tokens = 256
  seed = 42
  stop = @("`n`n###")
  messages = @(
    @{ role="system"; content="Responde brevemente." },
    @{ role="user"; content="Resume qué es un mutex." }
  )
} | ConvertTo-Json -Depth 30

curl.exe -sS --fail --connect-timeout 10 --max-time 120 `
  "$env:BASE_URL/chat/completions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

LLM: /v1/responses (no streaming)

Ejemplo básico

Bash

curl -sS --fail --connect-timeout 10 --max-time 120 \
  "${BASE_URL}/responses" \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"${MODEL}\",
    \"input\": \"Di hola en una sola oración.\"
  }" | python -m json.tool

PowerShell

$body = @{
  model = $env:MODEL
  input = "Di hola en una sola oración."
} | ConvertTo-Json -Depth 20

curl.exe -sS --fail --connect-timeout 10 --max-time 120 `
  "$env:BASE_URL/responses" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

Ejemplo con parámetros avanzados

Bash

curl -sS --fail --connect-timeout 10 --max-time 120 \
  "${BASE_URL}/responses" \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"${MODEL}\",
    \"temperature\": 0.2,
    \"top_p\": 0.9,
    \"max_output_tokens\": 256,
    \"seed\": 42,
    \"input\": \"Explica qué es un mutex en 3 puntos.\"
  }" | python -m json.tool

PowerShell

$body = @{
  model = $env:MODEL
  temperature = 0.2
  top_p = 0.9
  max_output_tokens = 256
  seed = 42
  input = "Explica qué es un mutex en 3 puntos."
} | ConvertTo-Json -Depth 30

curl.exe -sS --fail --connect-timeout 10 --max-time 120 `
  "$env:BASE_URL/responses" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

Salida en streaming (SSE)

/v1/chat/completions streaming

Bash

curl -N --no-buffer -sS --fail --connect-timeout 10 --max-time 0 \
  "${BASE_URL}/chat/completions" \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"${MODEL}\",
    \"stream\": true,
    \"messages\": [
      {\"role\":\"user\",\"content\":\"Cuenta del 1 al 20.\"}
    ]
  }"

PowerShell

$body = @{
  model = $env:MODEL
  stream = $true
  messages = @(
    @{ role="user"; content="Cuenta del 1 al 20." }
  )
} | ConvertTo-Json -Depth 20

curl.exe -N --no-buffer -sS --fail --connect-timeout 10 --max-time 0 `
  "$env:BASE_URL/chat/completions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body

/v1/responses streaming

Bash

curl -N --no-buffer -sS --fail --connect-timeout 10 --max-time 0 \
  "${BASE_URL}/responses" \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"${MODEL}\",
    \"stream\": true,
    \"input\": \"Escribe un poema corto sobre Berlín.\"
  }"

PowerShell

$body = @{
  model = $env:MODEL
  stream = $true
  input = "Escribe un poema corto sobre Berlín."
} | ConvertTo-Json -Depth 20

curl.exe -N --no-buffer -sS --fail --connect-timeout 10 --max-time 0 `
  "$env:BASE_URL/responses" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body

Modelos de visión (VLM) - Imágenes y texto

Uso de image_url + texto

Bash

curl -sS --fail --connect-timeout 10 --max-time 180 \
  "${BASE_URL}/chat/completions" \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "'"${MODEL}"'",
    "messages": [
      {
        "role": "user",
        "content": [
          {"type":"image_url","image_url":{"url":"https://ejemplo.com/recibo.png"}},
          {"type":"text","text":"Haz OCR de esta imagen. Devuelve texto plano."}
        ]
      }
    ]
  }' | python -m json.tool

PowerShell

$body = @{
  model = $env:MODEL
  messages = @(
    @{
      role="user"
      content=@(
        @{ type="image_url"; image_url=@{ url="https://ejemplo.com/recibo.png" } },
        @{ type="text"; text="Haz OCR de esta imagen. Devuelve texto plano." }
      )
    }
  )
} | ConvertTo-Json -Depth 30

curl.exe -sS --fail --connect-timeout 10 --max-time 180 `
  "$env:BASE_URL/chat/completions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

Subida de imágenes locales

Forma A: Usar endpoint de carga

Bash
# 1) Subir (endpoint de ejemplo, cámbialo según tu servicio)
UPLOAD_URL="${BASE_URL}/uploads"
IMG_URL=$(curl -sS --fail -H "Authorization: Bearer ${API_KEY}" \
  -F "file=@./recibo.png" "${UPLOAD_URL}" | python -c "import sys,json; print(json.load(sys.stdin)['url'])")

# 2) Inferencia
curl -sS --fail --connect-timeout 10 --max-time 180 \
  "${BASE_URL}/chat/completions" \
  -H "Authorization: Bearer ${API_KEY}" -H "Content-Type: application/json" \
  -d '{
    "model":"'"${MODEL}"'",
    "messages":[{"role":"user","content":[
      {"type":"image_url","image_url":{"url":"'"${IMG_URL}"'"}},
      {"type":"text","text":"Extrae todo el texto."}
    ]}]
  }'
PowerShell
# Si el servicio tiene endpoint de subida: $env:BASE_URL/uploads
# La respuesta de subida incluye un campo 'url' (ejemplo)
$uploadResp = curl.exe -sS --fail `
  -H "Authorization: Bearer $env:API_KEY" `
  -F "file=@recibo.png" `
  "$env:BASE_URL/uploads" | ConvertFrom-Json
$imgUrl = $uploadResp.url

$body = @{
  model = $env:MODEL
  messages = @(
    @{
      role="user"
      content=@(
        @{ type="image_url"; image_url=@{ url=$imgUrl } },
        @{ type="text"; text="Extrae todo el texto." }
      )
    }
  )
} | ConvertTo-Json -Depth 30

curl.exe -sS --fail --connect-timeout 10 --max-time 180 `
  "$env:BASE_URL/chat/completions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body

Forma B: Convertir imagen a base64

Bash
B64=$(python - <<'PY'
import base64
print(base64.b64encode(open("recibo.png","rb").read()).decode())
PY
)

curl -sS --fail --connect-timeout 10 --max-time 180 \
  "${BASE_URL}/chat/completions" \
  -H "Authorization: Bearer ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\":\"${MODEL}\",
    \"messages\":[
      {\"role\":\"user\",\"content\":[
        {\"type\":\"image_url\",\"image_url\":{\"url\":\"data:image/png;base64,${B64}\"}},
        {\"type\":\"text\",\"text\":\"Haz OCR y devuelve solo texto.\"}
      ]}
    ]
  }"
PowerShell
$b64 = [Convert]::ToBase64String([IO.File]::ReadAllBytes("recibo.png"))
$dataUrl = "data:image/png;base64,$b64"

$body = @{
  model = $env:MODEL
  messages = @(
    @{
      role="user"
      content=@(
        @{ type="image_url"; image_url=@{ url=$dataUrl } },
        @{ type="text"; text="Haz OCR y devuelve solo texto." }
      )
    }
  )
} | ConvertTo-Json -Depth 30

curl.exe -sS --fail --connect-timeout 10 --max-time 180 `
  "$env:BASE_URL/chat/completions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body

Modelos pequeños (SLM) - Bajo retardo y recursos

chat/completions - Mínimo retardo

Bash

curl -sS --fail --connect-timeout 3 --max-time 15 \
  "${BASE_URL}/chat/completions" \
  -H "Authorization: Bearer ${API_KEY}" -H "Content-Type: application/json" \
  -d "{
    \"model\":\"${MODEL}\",
    \"temperature\":0,
    \"max_tokens\":64,
    \"messages\":[{\"role\":\"user\",\"content\":\"Devuelve solo: OK\"}]
  }"

PowerShell

$body = @{
  model = $env:MODEL
  temperature = 0
  max_tokens = 64
  messages = @(@{role="user"; content="Devuelve solo: OK"})
} | ConvertTo-Json -Depth 10

curl.exe -sS --fail --connect-timeout 3 --max-time 15 `
  "$env:BASE_URL/chat/completions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body

Embeddings (Vectorización)

/v1/embeddings

Bash

curl -sS --fail --connect-timeout 10 --max-time 60 \
  "${BASE_URL}/embeddings" \
  -H "Authorization: Bearer ${API_KEY}" -H "Content-Type: application/json" \
  -d "{
    \"model\":\"text-embedding-3-small\",
    \"input\":[\"hola mundo\",\"prueba de embedding\"]
  }" | python -m json.tool

PowerShell

$body = @{
  model = "text-embedding-3-small"
  input = @("hola mundo","prueba de embedding")
} | ConvertTo-Json -Depth 10

curl.exe -sS --fail --connect-timeout 10 --max-time 60 `
  "$env:BASE_URL/embeddings" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

Rerank (Reordenación)

/v1/rerank (Plantilla)

Bash

curl -sS --fail --connect-timeout 10 --max-time 60 \
  "${BASE_URL}/rerank" \
  -H "Authorization: Bearer ${API_KEY}" -H "Content-Type: application/json" \
  -d '{
    "model":"modelo-rerank",
    "query":"ascend npu memory leak",
    "documents":[
      "Cómo hacer perfil de memoria en Ascend NPU.",
      "Guía para montar/desmontar en Linux.",
      "Conceptos básicos de atención en Transformers."
    ],
    "top_n":2
  }' | python -m json.tool

PowerShell

$body = @{
  model="modelo-rerank"
  query="ascend npu memory leak"
  documents=@(
    "Cómo hacer perfil de memoria en Ascend NPU.",
    "Guía para montar/desmontar en Linux.",
    "Conceptos básicos de atención en Transformers."
  )
  top_n=2
} | ConvertTo-Json -Depth 20

curl.exe -sS --fail --connect-timeout 10 --max-time 60 `
  "$env:BASE_URL/rerank" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

Llamadas a funciones (Salida estructurada)

chat/completions: tools + tool_choice

Bash

curl -sS --fail --connect-timeout 10 --max-time 120 \
  "${BASE_URL}/chat/completions" \
  -H "Authorization: Bearer ${API_KEY}" -H "Content-Type: application/json" \
  -d '{
    "model":"'"${MODEL}"'",
    "messages":[
      {"role":"user","content":"Extrae nombre y precio total de: \"Alice pagó 19.99 EUR\""}
    ],
    "tools":[
      {
        "type":"function",
        "function":{
          "name":"extraer_campos",
          "description":"Extrae campos estructurados",
          "parameters":{
            "type":"object",
            "properties":{
              "nombre":{"type":"string"},
              "total":{"type":"number"},
              "moneda":{"type":"string"}
            },
            "required":["nombre","total","moneda"],
            "additionalProperties":false
          }
        }
      }
    ],
    "tool_choice":{"type":"function","function":{"name":"extraer_campos"}}
  }' | python -m json.tool

PowerShell

$body = @{
  model = $env:MODEL
  messages = @(@{ role="user"; content="Extrae nombre y precio total de: ""Alice pagó 19.99 EUR""" })
  tools = @(
    @{
      type="function"
      function=@{
        name="extraer_campos"
        description="Extrae campos estructurados"
        parameters=@{
          type="object"
          properties=@{
            nombre=@{type="string"}
            total=@{type="number"}
            moneda=@{type="string"}
          }
          required=@("nombre","total","moneda")
          additionalProperties=$false
        }
      }
    }
  )
  tool_choice = @{ type="function"; function=@{ name="extraer_campos" } }
} | ConvertTo-Json -Depth 50

curl.exe -sS --fail --connect-timeout 10 --max-time 120 `
  "$env:BASE_URL/chat/completions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

responses: JSON Schema

Bash

curl -sS --fail --connect-timeout 10 --max-time 120 \
  "${BASE_URL}/responses" \
  -H "Authorization: Bearer ${API_KEY}" -H "Content-Type: application/json" \
  -d '{
    "model":"'"${MODEL}"'",
    "input":"Extrae nombre y precio total de: Alice pagó 19.99 EUR",
    "response_format":{
      "type":"json_schema",
      "json_schema":{
        "name":"campos_recibo",
        "schema":{
          "type":"object",
          "properties":{
            "nombre":{"type":"string"},
            "total":{"type":"number"},
            "moneda":{"type":"string"}
          },
          "required":["nombre","total","moneda"],
          "additionalProperties":false
        }
      }
    }
  }' | python -m json.tool

PowerShell

$body = @{
  model = $env:MODEL
  input = "Extrae nombre y precio total de: Alice pagó 19.99 EUR"
  response_format = @{
    type="json_schema"
    json_schema=@{
      name="campos_recibo"
      schema=@{
        type="object"
        properties=@{
          nombre=@{type="string"}
          total=@{type="number"}
          moneda=@{type="string"}
        }
        required=@("nombre","total","moneda")
        additionalProperties=$false
      }
    }
  }
} | ConvertTo-Json -Depth 60

curl.exe -sS --fail --connect-timeout 10 --max-time 120 `
  "$env:BASE_URL/responses" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

Procesamiento por lotes y concurrente

Bash: xargs concurrente

printf "%s\n" "q1: hola" "q2: mutex" "q3: semáforo" | \
  xargs -P 8 -I {} bash -lc '
    curl -sS --fail --connect-timeout 5 --max-time 60 "'"${BASE_URL}"'/chat/completions" \
      -H "Authorization: Bearer '"${API_KEY}"'" -H "Content-Type: application/json" \
      -d "{\"model\":\"'"${MODEL}"'\",\"messages\":[{\"role\":\"user\",\"content\":\"{}\"}]}" \
    | python -c "import sys,json; print(json.load(sys.stdin)[\"choices\"][0][\"message\"][\"content\"])"
  '

PowerShell: ForEach-Object -Parallel

$qs = @("q1: hola","q2: mutex","q3: semáforo")

$qs | ForEach-Object -Parallel {
  $body = @{
    model = $env:MODEL
    messages = @(@{role="user"; content=$_})
  } | ConvertTo-Json -Depth 10

  curl.exe -sS --fail --connect-timeout 5 --max-time 60 `
    "$env:BASE_URL/chat/completions" `
    -H "Authorization: Bearer $env:API_KEY" `
    -H "Content-Type: application/json" `
    -d $body
} -ThrottleLimit 8 | Out-String | Write-Host

Carga de archivos (OCR, documentos, tablas)

Cargar directamente PDF/imagen

Bash

export PARSE_URL="http://177.24.2.3:9005/file_parse"
export SERVER_URL="http://177.24.2.3:8000"

curl -sS --fail --connect-timeout 10 --max-time 3600 \
  "${PARSE_URL}" \
  -F "files=@./informe.pdf" \
  -F "backend=vlm-http-client" \
  -F "server_url=${SERVER_URL}" \
  -F "parse_method=auto" \
  -F "formula_enable=true" \
  -F "table_enable=true" \
  -F "start_page_id=0" \
  | python -m json.tool

PowerShell

$PARSE_URL="http://177.24.2.3:9005/file_parse"
$SERVER_URL="http://177.24.2.3:8000"

curl.exe -sS --fail --connect-timeout 10 --max-time 3600 `
  "$PARSE_URL" `
  -F "files=@informe.pdf" `
  -F "backend=vlm-http-client" `
  -F "server_url=$SERVER_URL" `
  -F "parse_method=auto" `
  -F "formula_enable=true" `
  -F "table_enable=true" `
  -F "start_page_id=0" `
  | python -m json.tool

Pipe: Descargar y reenviar

Bash

curl -L "https://www.energy.gov/sites/default/files/2023-01/informe.pdf" \
| curl -sS --fail --connect-timeout 10 --max-time 3600 \
  "http://177.24.2.3:9005/file_parse" \
  -F "files=@-;filename=informe.pdf" \
  -F "backend=vlm-http-client" \
  -F "server_url=http://177.24.2.3:8000" \
  -F "parse_method=auto" \
  -F "formula_enable=true" \
  -F "table_enable=true" \
  -F "start_page_id=0" \
  | python -m json.tool

PowerShell

$tmp = Join-Path $env:TEMP "informe.pdf"
curl.exe -L "https://www.energy.gov/sites/default/files/2023-01/informe.pdf" -o $tmp

curl.exe -sS --fail --connect-timeout 10 --max-time 3600 `
  "http://177.24.2.3:9005/file_parse" `
  -F "files=@$tmp" `
  -F "backend=vlm-http-client" `
  -F "server_url=http://177.24.2.3:8000" `
  -F "parse_method=auto" `
  -F "formula_enable=true" `
  -F "table_enable=true" `
  -F "start_page_id=0" `
  | python -m json.tool

Generación de imágenes (texto a imagen)

/v1/images/generations (Plantilla)

Bash

curl -sS --fail --connect-timeout 10 --max-time 300 \
  "${BASE_URL}/images/generations" \
  -H "Authorization: Bearer ${API_KEY}" -H "Content-Type: application/json" \
  -d '{
    "model":"modelo-imagen",
    "prompt":"Un diagrama esquemático de una canalización OCR, estilo minimalista",
    "size":"1024x1024"
  }' | python -m json.tool

PowerShell

$body = @{
  model="modelo-imagen"
  prompt="Un diagrama esquemático de una canalización OCR, estilo minimalista"
  size="1024x1024"
} | ConvertTo-Json -Depth 10

curl.exe -sS --fail --connect-timeout 10 --max-time 300 `
  "$env:BASE_URL/images/generations" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body | python -m json.tool

ASR (Reconocimiento de voz) y TTS (Síntesis de voz)

ASR: /v1/audio/transcriptions (multipart)

Bash

curl -sS --fail --connect-timeout 10 --max-time 300 \
  "${BASE_URL}/audio/transcriptions" \
  -H "Authorization: Bearer ${API_KEY}" \
  -F "model=whisper-1" \
  -F "file=@./voz.wav" \
  -F "language=en" \
  | python -m json.tool

PowerShell

curl.exe -sS --fail --connect-timeout 10 --max-time 300 `
  "$env:BASE_URL/audio/transcriptions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -F "model=whisper-1" `
  -F "file=@voz.wav" `
  -F "language=en" `
  | python -m json.tool

TTS: /v1/audio/speech (Guardar en archivo)

Bash

curl -sS --fail --connect-timeout 10 --max-time 300 \
  "${BASE_URL}/audio/speech" \
  -H "Authorization: Bearer ${API_KEY}" -H "Content-Type: application/json" \
  -d '{
    "model":"tts-1",
    "voice":"alloy",
    "input":"Hola desde TTS."
  }' --output salida.mp3

PowerShell

$body = @{
  model="tts-1"
  voice="alloy"
  input="Hola desde TTS."
} | ConvertTo-Json -Depth 10

curl.exe -sS --fail --connect-timeout 10 --max-time 300 `
  "$env:BASE_URL/audio/speech" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -d $body `
  --output salida.mp3

Seguridad y observabilidad

Reintentos + mostrar tiempo de ejecución

Bash

curl -sS --fail \
  --connect-timeout 10 --max-time 120 \
  --retry 5 --retry-all-errors --retry-delay 1 \
  -w "\nHTTP=%{http_code} total=%{time_total}s connect=%{time_connect}s\n" \
  "${BASE_URL}/chat/completions" \
  -H "Authorization: Bearer ${API_KEY}" -H "Content-Type: application/json" \
  -H "X-Trace-Id: trace-$(date +%s)" \
  -d "{\"model\":\"${MODEL}\",\"messages\":[{\"role\":\"user\",\"content\":\"ping\"}]}"

PowerShell

$trace = "trace-$([DateTimeOffset]::Now.ToUnixTimeSeconds())"
$body = @{ model=$env:MODEL; messages=@(@{role="user"; content="ping"}) } | ConvertTo-Json -Depth 10

curl.exe -sS --fail `
  --connect-timeout 10 --max-time 120 `
  --retry 5 --retry-all-errors --retry-delay 1 `
  -w "`nHTTP=%{http_code} total=%{time_total}s connect=%{time_connect}s`n" `
  "$env:BASE_URL/chat/completions" `
  -H "Authorization: Bearer $env:API_KEY" `
  -H "Content-Type: application/json" `
  -H "X-Trace-Id: $trace" `
  -d $body

Proxy (HTTP/HTTPS)

Bash

export HTTPS_PROXY="http://127.0.0.1:7890"
curl -sS --fail "${BASE_URL}/models" -H "Authorization: Bearer ${API_KEY}"

PowerShell

$env:HTTPS_PROXY="http://127.0.0.1:7890"
curl.exe -sS --fail "$env:BASE_URL/models" -H "Authorization: Bearer $env:API_KEY"

Servicios de infernecia personalizados (no OpenAI)

Inefrencia con cuerpo JSON

Bash

curl -sS --fail --connect-timeout 10 --max-time 120 \
  "http://HOST:PUERTO/inferir" \
  -H "Content-Type: application/json" \
  -d '{
    "modelo":"MODELO",
    "entradas":{"texto":"hola"},
    "parametros":{"max_new_tokens":128,"temperature":0.2}
  }'

PowerShell

$body = @{
  modelo="MODELO"
  entradas=@{ texto="hola" }
  parametros=@{ max_new_tokens=128; temperature=0.2 }
} | ConvertTo-Json -Depth 20

curl.exe -sS --fail --connect-timeout 10 --max-time 120 `
  "http://HOST:PUERTO/inferir" `
  -H "Content-Type: application/json" `
  -d $body

Inferencia con carga multipart

Bash

curl -sS --fail --connect-timeout 10 --max-time 600 \
  "http://HOST:PUERTO/inferir_archivo" \
  -F "archivo=@./entrada.pdf" \
  -F "modelo=MODELO" \
  -F "params={\"tabla\":true,\"formula\":true};type=application/json"

PowerShell

curl.exe -sS --fail --connect-timeout 10 --max-time 600 `
  "http://HOST:PUERTO/inferir_archivo" `
  -F "archivo=@entrada.pdf" `
  -F "modelo=MODELO" `
  -F "params={""tabla"":true,""formula"":true};type=application/json"

Errores comunes y solución de problemas

  • 401 Unauthorized: Token inválido o faltante
  • 404 Not Found: Endpoint incorrecto o no disponible
  • 413 Payload Too Large: Imagen demasiado grande
  • 415 Unsupported Media Type: Tipo de contenido incorrecto
  • 429 Too Many Requests: Límite de tasa excedido
  • 5xx: Errores del servidor

Tabla de parámetros (general)

  • temperature: Controla aleatoriedad
  • top_p: Muestreo de núcleo
  • max_tokens: Límite superior de tokens de salida
  • stop: Secuencia de parada
  • seed: Valor para resultados reproducibles
  • presence_penalty / frequency_penalty: Penalización para repetición
  • response_format / json_schema: Formato estructurado
  • stream=true: Streaming SSE
  • tools / tool_choice: Llamadas a funciones

Etiquetas: API curl IA modelos OpenAI

Publicado el 9-13 17:42

Etiquetas populares

©2026 Friki Work