Create Chat Completion with Vision (Streaming) - Base64

Interface Description: Given a prompt, the model returns one or more predicted completions. It can also return the probabilities of alternative tokens at each position. This interface is primarily used to create chat completions that involve image recognition (Base64-based) and text prompts.

Official Reference: OpenAI API Documentation

---

Endpoint Details

• Method: POST
• URL: https://api.codingplanx.ai/v1/chat/completions
• Content-Type: application/json

---

Request Headers

---

Request Body

---

Request Example

{
  "model": "gpt-4o-mini",
  "messages": [
    {
      "role": "system",
      "content": "You are a helpful assistant."
    },
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "What is in this image? Please describe it in detail."
        },
        {
          "type": "image_url",
          "image_url": {
            "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAiEAAAIhCAYAAACYF2qHAAAACXBIWXMAAA7EAAAOxAGVKw4bAAAgAElEQVR4nOy9ebxtWVXf+x1zrrX2Pud2VRQUjUVf0glIYvOJLdJ8iCagIor... (omitted long base64 string)"
          }
        }
      ]
    }
  ]
}

---

Response Body

---

Response Example (Successful 200 OK)

{
    "id": "chatcmpl-123",
    "object": "chat.completion",
    "created": 1677652288,
    "choices": [
        {
            "index": 0,
            "message": {
                "role": "assistant",
                "content": "\r
\r
Hello there, how may I assist you today?"
            },
            "finish_reason": "stop"
        }
    ],
    "usage": {
        "prompt_tokens": 9,
        "completion_tokens": 12,
        "total_tokens": 21
    }
}

---

Frequently Asked Questions (FAQs)

Q1: How do I enable streaming output? A: Set the stream parameter to true in the request body. Once enabled, the API will no longer wait for the entire response to be generated. Instead, it will return data chunks in data: {...} format via Server-Sent Events (SSE) until the data: [DONE] flag is received.

Q2: What are the requirements for Base64 encoding in vision tasks? A: The image Base64 string must include the correct MIME type prefix, such as data:image/png;base64, or data:image/jpeg;base64,. It is recommended to compress the image dimensions appropriately before converting to Base64 to reduce network latency and prompt_tokens consumption (excessively large images consume many tokens and may exceed context limits).

Q3: Why is the model's response truncated? A: Check the value of choices[0].finish_reason in the response body. If it is length, it means the generated text reached the max_tokens limit you set, or the entire conversation exceeded the model's maximum context window. You can resolve this by increasing the max_tokens value.

Q4: How can I force the model to output data in JSON format? A: First, set "response_format": { "type": "json_object" } in the request body. Important: In addition to setting this parameter, you must explicitly instruct the model to output JSON using natural language in the messages (either in the system prompt or user prompt), e.g., "Please output the result in JSON format." Failure to do so may result in the model generating endless whitespace until tokens are exhausted.

Q5: What should I do if I encounter a 401 Unauthorized error? A: Ensure that the Authorization parameter in the request header correctly carries the API Key. The format must strictly be Bearer YOUR_API_KEY (note the single space between Bearer and the Key).