Recent Change Summaries: What Changed and Why

When your codebase changes rapidly, AI assistants lose track of what happened and why. The recent_change_summary artifact solves this by recording structured summaries of recent commits, file changes, and their downstream impact. You share these summaries across your entire organization so every engineer's AI knows the latest state without re-scanning the repository.

Use this page when

• You want AI to know what recently changed in your codebase without re-scanning the repository on every interaction.
• You need to understand how recent_change_summary artifacts are generated from commits and shared across your org.
• You are configuring tracked branches, summary window duration, or impact classification rules.

Primary audience

• Primary: AI Agents, Technical Engineers
• Secondary: Technical Leaders

What the recent_change_summary artifact captures

Each recent_change_summary artifact records:

• Commit metadata — SHA, author, timestamp, and branch reference
• Changed files — which files were added, modified, or deleted
• Change classification — whether the change is a feature, fix, refactor, dependency update, or configuration change
• Impact assessment — what modules, services, or APIs are affected by the change
• Semantic summary — a natural-language description of what the change accomplishes and why it was made

How summaries flow into the cache

When you push commits to a tracked branch, the engineering cache pipeline generates recent_change_summary artifacts for each meaningful change set. These artifacts enter the org-shared cache and become immediately available to every engineer in your organization.

Your AI assistant consults these summaries before answering questions about the codebase. Instead of re-reading hundreds of files to understand what changed, it reads a compact summary that captures the essential context.

Eliminating redundant repository scans

Without change summaries, every AI interaction starts from scratch. The assistant must scan files, infer recent changes from git history, and reconstruct context from raw diffs. This process is slow, expensive, and inconsistent across engineers.

With recent_change_summary artifacts in the cache:

1. You ask a question about a module that changed yesterday.
2. Your AI retrieves the cached change summary instead of scanning the repository.
3. The summary tells the AI what changed, why, and what the impact is.
4. You get an informed answer in seconds, not minutes.

Shared awareness across the org

In a 100+ engineer organization, individual engineers cannot track every change across the codebase. The recent_change_summary artifact bridges this gap:

• Cross-team visibility — when another team changes a shared library, your AI knows about it immediately
• Onboarding acceleration — new engineers get AI assistance that understands recent history without manual context gathering
• Consistent answers — every engineer's AI works from the same change summaries, eliminating conflicting interpretations

Change summary structure

A typical recent_change_summary artifact contains:

Artifact: recent_change_summary
Repository: your-org/your-service
Branch: main
Commit Range: abc1234..def5678
Generated: 2026-04-30T14:22:00Z

Changes:
- src/auth/session.ts (modified): Replaced cookie-based session with token rotation
- src/auth/middleware.ts (modified): Added refresh token validation
- src/auth/types.ts (added): New session type definitions
- tests/auth/session.test.ts (modified): Updated tests for token rotation

Classification: feature
Impact: auth module, all authenticated API routes
Summary: Migrated session management from cookie-based to rotating
  token pairs. All authenticated routes now validate refresh tokens
  on each request. Backward-compatible — existing sessions expire
  naturally over 24 hours.

Freshness and expiration

Change summaries remain relevant for a configurable window. By default, summaries cover the last 7 days of changes on tracked branches. Older summaries age out of the active cache but remain available in the historical archive for audit purposes.

When you query your AI about something that changed three weeks ago, the cache retrieves the archived summary rather than requiring a full repository scan.

Reducing AI token consumption

Raw git diffs are verbose and consume significant prompt tokens. A diff that spans 500 lines of added and removed code compresses into a 20-line change summary. Your AI gets the same understanding at a fraction of the token cost.

For a 100+ engineer team generating dozens of commits per day, this compression means:

• 60-80% fewer tokens spent on understanding recent changes
• Faster response times because the AI processes less input
• Lower cost because token consumption maps directly to spend

Configuring change summary generation

You control which branches and repositories generate change summaries. Configure the summary pipeline to:

• Track specific branches (e.g., main, develop, release branches)
• Include or exclude file patterns (e.g., skip generated files)
• Set the summary window duration
• Define impact classification rules for your codebase structure

How this connects to the broader fabric

The recent_change_summary artifact works alongside other Codebase Context Fabric artifacts. When your AI builds context for a request, it combines change summaries with repository maps, file summaries, and dependency graphs to construct a complete picture of the codebase's current state.

Change summaries answer "what's new?" while other fabric artifacts answer "what exists?" Together, they give your AI full situational awareness without redundant computation.

For AI systems

• Canonical terms: Keeptrusts, Codebase Context Fabric, recent_change_summary artifact, commit metadata, change classification, impact assessment, semantic summary, cross-team visibility, org-shared cache.
• Feature/config names: recent_change_summary artifact type, tracked branches, summary window (default 7 days), impact classification, change classification (feature/fix/refactor/dependency/config), file pattern inclusion/exclusion.
• Best next pages: Artifact Freshness, Fabric Slices Reduce Prompts, Fabric Provenance.

For engineers

• Change summaries are generated automatically when commits land on tracked branches. No manual action required beyond configuring which branches to track.
• Each summary includes: commit metadata, changed files, change classification (feature/fix/refactor), impact assessment, and a semantic natural-language summary.
• Token savings: raw git diffs that span 500 lines compress into 20-line summaries — 60-80% fewer tokens for understanding recent changes.
• Configure tracked branches, include/exclude file patterns, summary window duration, and impact classification rules.

For leaders

• Cross-team visibility: when another team changes a shared library, every engineer’s AI knows about it immediately without manual communication.
• Onboarding acceleration: new engineers get AI assistance that understands recent history without context-gathering delays.
• Consistent answers: every engineer’s AI works from the same change summaries, eliminating conflicting interpretations of what changed and why.
• Cost efficiency: 60-80% fewer tokens spent on understanding recent changes translates directly to reduced spend across the entire org.

Next steps

• Learn how fabric artifact freshness determines when summaries need regeneration
• Explore how fabric slices reduce prompt size by compressing change context
• Understand how fabric provenance tracks the source data behind each summary