About This Project¶
Project Vision¶
docpack_confluence is a bridge between Confluence knowledge bases and AI systems. It enables seamless integration of Confluence content with any AI platform, whether through API-based systems like RAG pipelines or simple drag-and-drop file uploads to AI knowledge bases.
The core idea is simple: make Confluence content AI-ready with minimal configuration.
The Three Pain Points We Solve¶
1. Precise Batch Selection¶
Confluence is not a traditional file system. Manually specifying which pages to include is tedious and error-prone. We solve this with a familiar, low-code approach:
gitignore-style syntax: Use
includeandexcludepatterns just like.gitignoreHierarchical wildcards: Use
/*suffix to include a page and all its descendantsURL or ID support: Specify pages using their Confluence URLs or page IDs
Example configuration:
{
"include": [
"https://example.atlassian.net/wiki/spaces/KB/pages/123456/Getting+Started/*",
"https://example.atlassian.net/wiki/spaces/KB/pages/789012/API+Reference"
],
"exclude": [
"https://example.atlassian.net/wiki/spaces/KB/pages/111111/Draft+Pages/*"
]
}
This approach eliminates the need to maintain a list of individual files. Add or remove entire sections with a single pattern.
Note
Currently, filtering is based on page hierarchy (parent-child relationships) using the /* syntax. Title-based filtering with DSL expressions may be added in future versions.
2. Rich Metadata Output¶
AI systems work best when they have context. Each exported page includes:
Source information: The original Confluence URL for citation and reference
Page title: For navigation and identification
Markdown content: Human-readable, AI-parseable content converted from Confluence’s Atlas Document Format
Hierarchical metadata: Breadcrumb paths showing the page’s position in the knowledge structure
The output format uses XML for metadata encapsulation with Markdown for content, providing both machine-parseability and human readability:
<document>
<source_type>Confluence Page</source_type>
<confluence_url>https://example.atlassian.net/wiki/spaces/KB/pages/123/Title</confluence_url>
<title>Page Title</title>
<markdown_content>
# Page Title
Your page content in Markdown format...
</markdown_content>
</document>
3. Single-File Packaging¶
Managing multiple files is painful for AI knowledge bases:
Version tracking nightmare: Which files changed? Which need updating?
Sync complexity: Uploading dozens of files is error-prone
Update friction: Deleting and re-uploading many files is tedious
Our solution: pack everything into a single file.
Delete the old file, upload the new one - done
Perfect for platforms like Claude Projects, ChatGPT Knowledge, or any drag-and-drop AI system
Works equally well with API-based RAG systems
Whether your AI platform has an API or just accepts file uploads, a single consolidated file makes synchronization trivial.
How It Works¶
Configure: Write a simple JSON configuration specifying your Confluence space and filtering rules
Fetch: The tool connects to Confluence API and retrieves matching pages
Transform: Pages are converted from Atlas Doc Format to Markdown with metadata
Export: Each page becomes a structured XML file, optionally merged into one bundle
Confluence Space
|
v
[Filter with include/exclude patterns]
|
v
[Convert to Markdown + Metadata]
|
v
Individual XML files --> Single packed file
(optional)
Use Cases¶
RAG Pipelines: Feed Confluence documentation into vector databases for retrieval-augmented generation
AI Knowledge Bases: Upload company wikis to Claude Projects, ChatGPT, or similar platforms
Documentation AI: Create AI assistants that understand your internal documentation
Content Migration: Export Confluence content in a portable, structured format
Current Limitations¶
Page filtering is hierarchy-based only (using
/*for descendants)Title-based filtering with DSL expressions is planned for future releases
Requires Confluence Cloud with API access
Summary¶
docpack_confluence transforms Confluence from a static wiki into an AI-ready knowledge source. With simple JSON configuration and familiar gitignore-style patterns, you can:
Select exactly the pages you need
Export with rich metadata for AI consumption
Package into a single file for easy synchronization
It’s the missing link between your Confluence knowledge base and the AI tools that can unlock its value.