Updated at: 2025-07-07 09:44:26

Install Python Dependency Packages

To run Python nodes in workflows, administrators must upload and install Python dependency packages via Python Package Management interface in Works.

Click “” button in Works and select [+Upload] to upload a “.tar” file. AnyShare will automatically installs the package upon upload.

Once installed, users can utilize this package in Python or custom nodes.

Note：
1）Action binding for upload and installation (installation will start automatically after upload).
2）Cannot pause ongoing uploads or installations.
3）Deleted packages become invalid immediately.

Implement Python Nodes in Automated Flow

Python nodes enable custom logic beyond built-in workflow capabilities.

Note: It requires coding skills and AnyShare OpenAPI knowledge to use Python node.

Example: OCR Integration

from aishu_anyshare_api.api_client import ApiClient
import aishu_anyshare_api_efast as efast
from datetime import datetime
import requests
import json
import re
import base64
def main(doc_id):
client = ApiClient(verify_ssl=False)
try:
json_file = file_download(client, doc_id)
text = json_file.content
url = 'http://10.4.132.197:8507/lab/ocr/predict/general'
b64 = base64.b64encode(text).decode()
data = {'scene': 'chinese_print', 'image': b64}
res = requests.post(url, json=data).json()
texts = res['data']['json']['general_ocr_res']['texts']
return texts
except Exception as e:
return str(e)
def file_download(client: ApiClient, doc_id: str) -> requests.Response:
resp = client.efast.efast_v1_file_osdownload_post(efast.FileOsdownloadReq.from_dict({"docid":doc_id})) #
type: ignore
method, url, headers = resp.authrequest[0], resp.authrequest[1], resp.authrequest[2:]
header = {}
for val in headers:
arr = val.split(": ")
header[arr[0]] = arr[1].strip()
resp = requests.request(method, url, headers=header, verify=False) # type: ignore
return resp

Code Functionality:

The provided example demonstrates core implementation logic. And this script executes the following core operations:

Retrieves file content via AnyShare SDK and extracts business-required data fields from the OCR's JSON response.

Note: The specific extraction rules and the JSON data structure may vary depending on the OCR model used. You can adjust as needed.

Code Mechanics:

The workflow framework internally encapsulates AnyShare's document operation APIs as a Python SDK, enabling direct calls while automatically handling user authentication tokens. The input/output values of the main() function are recorded in the workflow context, remaining accessible throughout the entire workflow lifecycle.

Note:

1) Workflow activates upon file upload events.

2) The Python node accepts an input variable (e.g., doc_id), which carries the document ID of the triggering file for node operations.

3) The Python node supports multiple output variables that directly map to the return values of its main() method.

4) The Python node outputs processed data via return values, making it available for downstream workflow operations such as cataloging and tagging.