#601

Global Rank · of 601 Skills

byted-tos-doc-process AI Agent Skill

View Source: bytedance/agentkit-samples

Medium

Installation

npx skills add bytedance/agentkit-samples --skill byted-tos-doc-process

12

Installs

Bytedance TOS Document Process Skill

This skill provides document processing functions for files in Bytedance's TOS via the doc-preview feature, implemented by generating pre-signed URLs with the Volcengine TOS SDK.

Note: This approach is necessary because the SDK's get_object method does not directly support doc_* keyword arguments. All document processing parameters must be passed as query parameters in a pre-signed URL.

Quick Start

1. Client Initialization

import os
import tos
from tos.enum import HttpMethodType
from urllib.request import urlopen

def create_client() -> tos.TosClientV2:
    """Initializes a TosClientV2 from environment variables."""
    try:
        # ... (full implementation in scripts)
        return tos.TosClientV2(
            ak=os.getenv('TOS_ACCESS_KEY'),
            sk=os.getenv('TOS_SECRET_KEY'),
            endpoint=os.getenv('TOS_ENDPOINT'),
            region=os.getenv('TOS_REGION'),
            security_token=os.getenv('TOS_SECURITY_TOKEN'),
        )
    except Exception as e:
        print(f"Error initializing client: {e}")
        return None

client = create_client()

2. Basic Workflow (Pre-signed URL)

# (Assumes 'client' is initialized and 'bucket_name', 'object_key' are set)

# 1. Preview document as a PDF and save locally
try:
    # Build query params for doc-preview
    pdf_params = {
        "x-tos-process": "doc-preview",
        "x-tos-doc-dst-type": "pdf"
    }
    presigned_pdf = client.pre_signed_url(
        HttpMethodType.Http_Method_Get,
        bucket_name,
        object_key,
        query=pdf_params
    )
    
    # Download the content from the pre-signed URL
    with urlopen(presigned_pdf.signed_url) as response, open("local_preview.pdf", "wb") as f_out:
        f_out.write(response.read())
    print("PDF preview saved to local_preview.pdf")

except Exception as e:
    print(f"Error converting to PDF: {e}")

# 2. Preview page 3 as a PNG image
try:
    png_params = {
        "x-tos-process": "doc-preview",
        "x-tos-doc-dst-type": "png",
        "x-tos-doc-page": "3",
        "x-tos-doc-image-dpi": "150"
    }
    presigned_png = client.pre_signed_url(
        HttpMethodType.Http_Method_Get,
        bucket_name,
        object_key,
        query=png_params
    )
    with urlopen(presigned_png.signed_url) as response, open("page_3.png", "wb") as f_out:
        f_out.write(response.read())
    print("Page 3 saved as page_3.png")

except Exception as e:
    print(f"Error converting to PNG: {e}")

# 3. Get total page count from response headers
try:
    presigned_head = client.pre_signed_url(
        HttpMethodType.Http_Method_Get,
        bucket_name,
        object_key,
        query={"x-tos-process": "doc-preview", "x-tos-doc-dst-type": "pdf"}
    )
    with urlopen(presigned_head.signed_url) as response:
        total_pages = response.headers.get("x-tos-total-page")
        print(f"Document has {total_pages} pages.")
except Exception as e:
    print(f"Error getting page count: {e}")

Core Operations

All document processing is achieved by generating a pre-signed URL with process=\"doc-preview\" and other x-tos-doc-* parameters in the query string.

1. Convert to PDF (x-tos-doc-dst-type='pdf')

Converts an entire document into a single PDF file.

# See Quick Start example

2. Convert to Image (x-tos-doc-dst-type='png' or 'jpg')

Converts a specific page of a document into an image.

# See Quick Start example
# Use query params like "x-tos-doc-page", "x-tos-doc-image-dpi", etc.

3. Convert to HTML (x-tos-doc-dst-type='html')

Fetches a temporary HTML page containing a token for the final preview URL. This requires a second step to parse the HTML and decode the token.

# Step 1: Get the HTML content via a pre-signed URL
html_params = {"x-tos-process": "doc-preview", "x-tos-doc-dst-type": "html"}
presigned_html = client.pre_signed_url(HttpMethodType.Http_Method_Get, bucket_name, object_key, query=html_params)

with urlopen(presigned_html.signed_url) as response:
    html_content = response.read().decode('utf-8')

# Step 2: Parse and decode (see scripts/doc_preview_html_url.py for full logic)
# ... logic to extract and base64-decode the token ...
# final_url = decode_preview_url(token)

4. Batch Export Pages (image-mode=1)

Exports a range of pages as images directly to a TOS bucket.

# Use query params: "image-mode", "start-page", "end-page", "x-tos-save-bucket", "x-tos-save-object"
batch_params = {
    "x-tos-process": "doc-preview",
    "x-tos-doc-dst-type": "jpg",
    "image-mode": "1",
    "start-page": "2",
    "end-page": "5",
    "x-tos-save-bucket": "output-bucket",
    "x-tos-save-object": "exported/page_{Page}.jpg" # {Page} is a placeholder
}
presigned_batch = client.pre_signed_url(HttpMethodType.Http_Method_Get, bucket_name, object_key, query=batch_params)
# The response body (from urlopen) contains JSON metadata about the batch job

Authorization

Authentication is handled by tos.TosClientV2. Provide credentials via environment variables.

Required Environment Variables

  • TOS_ACCESS_KEY
  • TOS_SECRET_KEY
  • TOS_ENDPOINT
  • TOS_REGION

Optional for STS

  • TOS_SECURITY_TOKEN

Best Practices

  • Error Handling: Always wrap HTTP requests in try...except blocks for HTTPError and URLError.
  • Parameter Reference: Refer to REFERENCE.md for a mapping of doc_preview_params.py arguments to x-tos-* query keys and to the official TOS documentation for authoritative details.
  • HTML Preview: Be aware of the two-step process and the custom domain requirement for recent buckets.
  • Total Pages Header: The x-tos-total-page header is a convenient way to get the page count.

Additional Resources

  • For detailed parameters, see REFERENCE.md.
  • For end-to-end examples, see WORKFLOWS.md.
  • For executable Python examples, see the scripts/ directory.
  • For the definitive list of all processing parameters, always consult the official Volcengine TOS Document Preview documentation.

Installs

Installs 12
Global Rank #601 of 601

Security Audit

ath Safe
socket Safe
Alerts: 0 Score: 90
snyk Medium
EU EU-Hosted Inference API

Power your AI Agents with the best open-source models.

Drop-in OpenAI-compatible API. No data leaves Europe.

Explore Inference API

GLM

GLM 5

$1.00 / $3.20

per M tokens

Kimi

Kimi K2.5

$0.60 / $2.80

per M tokens

MiniMax

MiniMax M2.5

$0.30 / $1.20

per M tokens

Qwen

Qwen3.5 122B

$0.40 / $3.00

per M tokens

How to use this skill

1

Install byted-tos-doc-process by running npx skills add bytedance/agentkit-samples --skill byted-tos-doc-process in your project directory. Run the install command above in your project directory. The skill file will be downloaded from GitHub and placed in your project.

2

No configuration needed. Your AI agent (Claude Code, Cursor, Windsurf, etc.) automatically detects installed skills and uses them as context when generating code.

3

The skill enhances your agent's understanding of byted-tos-doc-process, helping it follow established patterns, avoid common mistakes, and produce production-ready output.

What you get

Skills are plain-text instruction files — not executable code. They encode expert knowledge about frameworks, languages, or tools that your AI agent reads to improve its output. This means zero runtime overhead, no dependency conflicts, and full transparency: you can read and review every instruction before installing.

Compatibility

This skill works with any AI coding agent that supports the skills.sh format, including Claude Code (Anthropic), Cursor, Windsurf, Cline, Aider, and other tools that read project-level context files. Skills are framework-agnostic at the transport level — the content inside determines which language or framework it applies to.

Data sourced from the skills.sh registry and GitHub. Install counts and security audits are updated regularly.

EU Made in Europe

Chat with 100+ AI Models in one App.

Use Claude, ChatGPT, Gemini alongside with EU-Hosted Models like Deepseek, GLM-5, Kimi K2.5 and many more.

Customer Support