Star on GitHub
Star on GitHub

Engineering Decisions Memory (EDM)

PRSense goes beyond simple code review by automatically identifying, categorizing, and archiving the Engineering Decisions made during your pull requests and code reviews. This creates a searchable, AI-powered knowledge base of why things were built the way they were.

What is EDM?

Often, the most critical context about a codebase lives in the minds of developers or gets buried in Slack threads and PR comments. When a new developer joins the team or an old feature needs to be refactored, that context is typically lost.

Engineering Decisions Memory (EDM) solves this by:

  1. Scanning PR Descriptions & Comments: Using NLP to detect when a technical decision is being made or debated.
  2. Extracting the Rationale: Identifying the underlying problem, the chosen solution, and the reasoning behind it.
  3. Structuring the Data: Categorizing the decision (e.g., Architecture, Security, Dependency, Performance) and tagging it with relevant metadata.
  4. Building a Knowledge Graph: Connecting decisions to specific code files and architectural domains.

How It Works

1. Automatic Detection

When a developer submits a PR or when reviewers leave comments, the PRSense AI Engine actively monitors the conversation. It looks for keywords and patterns that indicate a decision, such as:

  • “We decided to use X instead of Y because…”
  • “This approach handles the edge case where…”
  • “We are deprecating the old API in favor of…“

2. Decision Extraction

Once a decision is detected, PRSense extracts the core elements into a structured format:

  • Summary: A concise description of the decision.
  • Reasoning: The underlying rationale and trade-offs considered.
  • Author: The team member who made or finalized the decision.
  • Source: A direct link to the PR or comment where the decision occurred.
  • Tags: Automatically generated labels (e.g., database, auth, performance).

3. The EDM Repository

All extracted decisions are stored in the org’s private EDM Repository, accessible via the PRSense Dashboard.

[!TIP] Codebase Chat Integration The true power of EDM shines when used with the PRSense Codebase Chat. You can ask questions like “Why did we choose Redis over Memcached for session storage?” and the AI will reference the exact EDM record and PR discussion.

Viewing Decisions

You can view your organization’s captured decisions by navigating to the Decisions tab in the PRSense Dashboard.

The view includes:

  • A searchable list of all decisions.
  • Filtering by tags, authors, and date ranges.
  • Quick links back to the original source material on GitHub/GitLab.

Best Practices for EDM

To get the most out of Engineering Decisions Memory, encourage your team to:

  • Write descriptive PR summaries: The better the initial context, the more accurate the extracted decision.
  • Document trade-offs in comments: When reviewing code, explicitly state why one approach is preferred over another.
  • Use standard terminology: Consistent use of terms like “Architecture,” “Refactor,” or “Security” helps the AI categorize decisions more effectively.