Gemini 全模型接口 - 原生 API 格式

curl -X POST "https://direct.evolink.ai/v1beta/models/gemini-4-argon:generateContent" \
  -H "Authorization: Bearer $EVOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      { "role": "user", "parts": [{ "text": "请写一份完整的 Rust 异步运行时设计文档，包含架构、调度器、I/O 驱动、定时器与完整示例代码。" }] }
    ],
    "generationConfig": {
      "maxOutputTokens": 262144,
      "thinkingConfig": { "thinkingLevel": "high" }
    }
  }'

{
  "candidates": [
    {
      "content": { "role": "model", "parts": [{ "text": "# Rust 异步运行时设计\n\n## 1. 总体架构\n……" }] },
      "finishReason": "CONTINUATION",
      "continuationToken": "<很长的一段字符串，原样保存>"
    }
  ]
}

curl -X POST "https://direct.evolink.ai/v1beta/models/gemini-4-argon:generateContent" \
  -H "Authorization: Bearer $EVOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      { "role": "user", "parts": [{ "text": "请写一份完整的 Rust 异步运行时设计文档，包含架构、调度器、I/O 驱动、定时器与完整示例代码。" }] }
    ],
    "generationConfig": {
      "maxOutputTokens": 262144,
      "thinkingConfig": { "thinkingLevel": "high" }
    },
    "continuationToken": "<上一次响应里的 continuationToken>"
  }'

import os
from copy import deepcopy

import requests

URL = "https://direct.evolink.ai/v1beta/models/gemini-4-argon:generateContent"
HEADERS = {"Authorization": f"Bearer {os.environ['EVOLINK_API_KEY']}"}


def generate_with_continuation(body):
    request_body = deepcopy(body)
    pieces = []
    while True:
        # 客户端读取超时设为 1,200 秒，不代表服务端保证持续连接 20 分钟。
        resp = requests.post(URL, headers=HEADERS, json=request_body, timeout=1200)
        resp.raise_for_status()
        candidates = resp.json().get("candidates", [])
        if not candidates:
            raise RuntimeError("响应中没有候选结果，请检查 promptFeedback。")
        candidate = candidates[0]
        for part in candidate.get("content", {}).get("parts", []):
            if not part.get("thought"):
                pieces.append(part.get("text", ""))
        reason = candidate.get("finishReason")
        token = candidate.get("continuationToken")
        if reason == "CONTINUATION":
            if not token:
                raise RuntimeError("响应要求续写，但没有返回 continuationToken。")
            request_body["continuationToken"] = token  # 其余字段保持不变
            continue
        return {
            "text": "".join(pieces),
            "finishReason": reason,
            "continuationToken": token,
        }


body = {
    "contents": [
        {"role": "user", "parts": [{"text": "请写一份完整的 Rust 异步运行时设计文档，包含架构、调度器、I/O 驱动、定时器与完整示例代码。"}]}
    ],
    "generationConfig": {"maxOutputTokens": 262144, "thinkingConfig": {"thinkingLevel": "high"}},
}

result = generate_with_continuation(body)
print(result["text"])
print("finishReason:", result["finishReason"])
if result["finishReason"] == "MAX_TOKENS":
    print("累计输出预算已耗尽；如有续写令牌，可手动提高 maxOutputTokens 后恢复。")

body = {
    "contents": [
        {"role": "user", "parts": [{"text": "Output integers 1 through 100, one per line. Start immediately."}]}
    ],
    "generationConfig": {"maxOutputTokens": 64, "thinkingConfig": {"thinkingLevel": "low"}},
}

first = generate_with_continuation(body)
print(first["text"], end="")
if first["finishReason"] == "MAX_TOKENS" and first["continuationToken"]:
    body["continuationToken"] = first["continuationToken"]
    body["generationConfig"]["maxOutputTokens"] = 2048  # 调用方明确提高累计预算
    resumed = generate_with_continuation(body)
    print(resumed["text"], end="")
    print("\nfinishReason:", resumed["finishReason"])
else:
    print("\nfinishReason:", first["finishReason"])