add openai api

This commit is contained in:
2026-06-22 11:06:01 +08:00
parent 30ead8fa43
commit ac4078d4af
9 changed files with 397 additions and 0 deletions
+76
View File
@@ -11,6 +11,8 @@
- `GET /output/...`:访问生成的 Markdown、JSON 和原始文件
- `GET /health`:健康检查
- `GET /docs`Swagger API 文档
- `GET /v1/models`OpenAI 兼容模型列表
- `POST /v1/chat/completions`OpenAI 兼容文档转 Markdown 接口
---
@@ -104,6 +106,78 @@ curl -F "file=@/path/to/page.png" -F "force=true" http://127.0.0.1:8000/api/reco
Markdown 访问地址:`http://服务器IP:8000/output/<文档名>/<文档名>.md`
### OpenAI 兼容接口
适用于已有 OpenAI SDK / LangChain / 各类 Agent 框架的调用方式。
若配置了 `OPENOCR_ROOT_PATH=/openocr`,则路径前缀为 `/openocr/v1/...`
**列出模型**
```bash
curl https://jsuse.com/openocr/v1/models
```
**文档转 Markdownchat/completions**
```bash
curl https://jsuse.com/openocr/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "opendoc-0.1b",
"messages": [
{
"role": "user",
"content": [
{"type": "text", "text": "Convert this document to markdown"},
{
"type": "image_url",
"image_url": {"url": "https://example.com/page.png"}
}
]
}
]
}'
```
也支持 `data:image/png;base64,...` 形式的图片 URL。
响应中的 `choices[0].message.content` 即为 Markdown 文本。
若设置了 `OPENOCR_API_KEY`,需附加请求头:
```bash
-H "Authorization: Bearer your-api-key"
```
Python 示例(OpenAI SDK):
```python
from openai import OpenAI
client = OpenAI(
base_url="https://jsuse.com/openocr/v1",
api_key="your-api-key-or-empty",
)
response = client.chat.completions.create(
model="opendoc-0.1b",
messages=[
{
"role": "user",
"content": [
{"type": "text", "text": "Convert to markdown"},
{
"type": "image_url",
"image_url": {"url": "https://example.com/page.png"},
},
],
}
],
)
print(response.choices[0].message.content)
```
---
## 日常开发与更新
@@ -203,6 +277,8 @@ docker compose up -d --build # 重新构建并启动
| `OPENOCR_LAYOUT_THRESHOLD` | `0.4` | 版面分析阈值 |
| `OPENOCR_MAX_LENGTH` | `2048` | 识别最大长度 |
| `OPENOCR_CORS_ORIGINS` | `*` | 跨域来源,多个用逗号分隔 |
| `OPENOCR_API_KEY` | *(空)* | OpenAI 兼容接口 Bearer Token,留空则不校验 |
| `OPENOCR_OPENAI_MODEL` | `opendoc-0.1b` | OpenAI 兼容接口使用的模型 ID |
| `OPENOCR_LOG_LEVEL` | `info` | 日志级别 |
---