Building a Knowledge Graph from JIRA and Confluence

The Problem

Enterprise teams drown in unstructured information spread across JIRA tickets, Confluence pages, Slack threads, and Google Docs. Finding the right context for a decision made six months ago is almost impossible.

The Solution: A Knowledge Graph

A knowledge graph connects entities (people, projects, decisions, tickets) with typed relationships, making it possible to traverse connections and surface relevant context automatically.

Architecture Overview

Implementation

Step 1: Extract Data

Pull tickets and pages using the REST APIs. Normalize the data into a common schema:

from jira import JIRA
from atlassian import Confluence

def extract_jira_tickets(project_key: str) -> list[dict]:
    jira = JIRA(server=JIRA_URL, token_auth=JIRA_TOKEN)
    issues = jira.search_issues(f"project={project_key}", maxResults=1000)
    return [normalize_issue(issue) for issue in issues]

Step 2: Build the Graph

Map entities to nodes and relationships to edges. Key relationship types:

• ASSIGNED_TO — ticket to person
• BLOCKS / BLOCKED_BY — ticket dependencies
• REFERENCES — confluence page to ticket
• DECIDED_IN — decision to meeting/page

Step 3: Query

With the graph in place, you can answer questions like:

• "What decisions were made about the auth migration?"
• "Who has context on the payment service refactor?"
• "What tickets are blocking the Q3 launch?"

Results

After indexing ~5,000 JIRA tickets and ~800 Confluence pages, search relevance improved dramatically compared to native JIRA/Confluence search. The graph structure allows multi-hop queries that traditional search simply can't handle.