Skip to main content
Product Knowledge gives you an AI-generated map of your codebase’s feature areas, helping you understand what your product does and track changes at a higher level than individual commits.

​
What is Product Knowledge?

Product Knowledge automatically analyzes your linked repositories to discover and categorize the different features, components, and systems in your codebase. This understanding powers several Kasava features:
  • Smarter commit classification - Commits are automatically tagged with the feature areas they affect
  • Better digests - Weekly updates organized by feature instead of random commits
  • Targeted insights - AI analysis understands which features are most active or at risk
  • Architectural overview - See how your codebase is structured across layers

​
Getting Started

When you first visit the Knowledge page with no feature areas defined, you’ll see an empty state with two options:
  1. Create Manually - Define feature areas yourself with custom names and directory patterns
  2. AI Discover - Let AI analyze your commit history to automatically identify features
Product Knowledge empty state with Discover and Create buttons Once you have feature areas, the page displays a summary card, optional architecture overview, and features organized by category.

​
Accessing Product Knowledge

  1. Navigate to a product in the sidebar
  2. Click Knowledge in the Monitor section
The URL structure is: /products/[productId]/knowledge Sidebar navigation showing Knowledge under the Monitor section

​
Understanding Feature Areas

A feature area represents a logical grouping of code that serves a common purpose. Examples include:
  • User Authentication
  • Dashboard Analytics
  • Notification System
  • Payment Processing
  • API Endpoints

​
Feature Area Sources

Feature areas can come from different sources, indicated by an icon:
SourceIconDescription
AI DiscoveredSparklesAutomatically identified by analyzing your commit history
User DefinedPersonManually created by you or your team
MergedGit MergeCreated by combining two existing feature areas

​
Hierarchical Organization

Feature areas are organized in a three-level hierarchy:
  1. Categories (Level 0) - High-level groupings like “Core Features”, “Infrastructure”, “User & Auth”
  2. Feature Groups (Level 1) - Related features grouped together
  3. Features (Level 2) - Individual feature areas
The Knowledge page displays features grouped by category, sorted by activity (commit count).

​
Confidence Scores

AI-discovered features include a confidence percentage indicating how certain the AI is about the classification. Higher confidence means:
  • Clear naming patterns in the codebase
  • Consistent directory structures
  • Well-defined boundaries between features

​
AI Discovery

​
How It Works

AI Discovery analyzes your commit history to understand your codebase:
  1. Commit analysis - Each commit message and changed files are examined
  2. Pattern detection - The AI identifies recurring themes, directories, and naming conventions
  3. Feature extraction - Logical feature areas are identified and named
  4. Categorization - Features are organized into a hierarchical structure

​
Running AI Discovery

1

Navigate to Knowledge

Open your product and click Knowledge in the Monitor section of the sidebar
2

Click AI Discover

Click the AI Discover button in the summary card (top right)
3

Confirm Rebuild (if applicable)

If you have existing AI-discovered features, a confirmation dialog appears:
  • Shows how many AI-discovered features will be replaced
  • User-defined features are always preserved
  • Click Rebuild Features to proceed
4

Wait for Analysis

The AI analyzes your commit history. This typically takes a few seconds to a minute depending on repository size. A success notification shows how many areas were discovered and created.
AI Discovery requires at least one repository linked to your product. If no repositories are connected, you’ll see a notification prompting you to link repositories first.

​
Upgrade Analysis

If your feature areas don’t have hierarchical structure (older discovery format), you’ll see an “Upgrade Analysis” button. This re-analyzes all commits with improved AI categorization to generate the three-level hierarchy.
Upgrade Analysis queues commits for re-processing and may take several minutes for large repositories. You’ll be notified when complete.

​
Managing Feature Areas

​
Creating a Feature Area Manually

You can define feature areas yourself for complete control:
1

Click Add Feature

Click the + Add Feature button above the feature categories (visible when you have existing features), or click Create Manually from the empty state
2

Enter Details

Fill in the Create Feature Area dialog:
  • Name (required) - A clear name like “User Authentication”
  • Description (optional) - What this feature area covers
  • Directory Patterns (optional) - Glob patterns to match files, one per line (e.g., src/auth/**)
3

Create

Click Create to save your new feature area. It will appear in the appropriate category based on its name.
Create Feature Area dialog with name, description, and directory patterns fields User-defined features are marked with a person icon to distinguish them from AI-discovered features (sparkles icon) or merged features (git merge icon).

​
Directory Patterns

Directory patterns help match commits to feature areas based on file paths. Use glob syntax: 
src/auth/**
lib/authentication/**
components/login/*
PatternMatches
src/auth/**All files in src/auth and subdirectories
*.tsAll TypeScript files
src/**/test*.tsTest files anywhere in src

​
Viewing Feature Area Details

Click any feature area row to expand and see detailed information:
SectionDescription
DescriptionWhat this feature covers
StatsCommits (30 days), files changed, active contributors
Top ContributorsPeople who’ve worked on this feature most, with avatars
Recent WorkAI-generated descriptions of recent changes (intent/purpose)
Recent CommitsLatest commit messages with short SHA and timestamps
Directory PatternsGlob patterns associated with this feature
Source BadgeShows if AI Discovered, User Defined, or Merged
The expanded view loads additional details on demand to keep the interface fast. Expanded feature area showing stats, contributors, and recent commits

​
Modifying a Feature Area

To make changes to an existing feature area, use the three-dot menu on the feature row:
  1. Hover over the feature area row
  2. Click the three-dot menu on the right
  3. Choose from available actions:
    • Edit - Update name, description, or directory patterns
    • Combine with… - Merge this feature into another
    • Delete - Remove the feature area
For significant changes, consider creating a new user-defined feature and combining the existing one into it. This preserves commit history while giving you full control over the definition.

​
Combining Feature Areas

Sometimes AI creates similar feature areas that should be one. Combine them:
1

Open Combine Dialog

Click the three-dot menu on the feature you want to merge away, then select “Combine with…”
2

Select Target

Search and select the feature area you want to combine into
3

Review Preview

The dialog shows what the combined result will look like:
  • Combined commit count
  • Merged directory patterns
4

Confirm

Click “Combine” to merge the features. The source feature will be absorbed into the target.
Combining is not reversible. The source feature area is removed and its history is merged into the target.

​
Deleting a Feature Area

  1. Hover over the feature area row
  2. Click the three-dot menu
  3. Select Delete
  4. Confirm deletion
Deleted feature areas are soft-deleted and commits previously associated with them may be reclassified.

​
Correcting AI-Discovered Features

The AI does its best to understand your codebase, but it may not always get things right. Here’s how to correct it:

​
Wrong Name or Description

  1. Edit the feature area to update the display name or description
  2. Your changes override the AI-generated values

​
Incorrect Categorization

If a feature is in the wrong category:
  1. Create a new user-defined feature with the correct name
  2. Combine the misclassified AI feature into your new one
  3. The user-defined feature takes precedence

​
Too Many Similar Features

Use the Combine feature to merge duplicate or overly-granular features into one.

​
Missing Feature

If the AI missed an important feature:
  1. Create it manually with “Add Feature”
  2. Add directory patterns to help future commits match correctly

​
Rebuilding from Scratch

If the AI-discovered features are too far off:
  1. Delete individual incorrect features, or
  2. Run “AI Discover” again to completely regenerate all AI features
User-defined features are always preserved when running AI Discovery.

​
Architecture Overview

The Knowledge page shows Architectural Layers when detected, representing the high-level structure of your codebase. Common layers include:
  • Frontend - UI components, pages, styles
  • Backend - API routes, services, handlers
  • Database - Schemas, migrations, models
  • Infrastructure - CI/CD, configuration, deployment
Layers are displayed as a compact grid showing which repositories map to each layer and how many repos contribute to each layer. Architecture overview grid showing layers and repository counts

​
Summary Card

The summary card at the top of the Knowledge page provides a quick overview:
  • Features - Number of identified feature areas
  • Commits - Total commits analyzed across all features
  • Categories - Number of top-level categories
  • AI discovered - Count of features found by AI analysis
  • User defined - Count of features you created manually
  • Last sync - When AI Discovery was last run
The card also displays a narrative summary describing your product’s feature distribution and primary focus areas. Product Knowledge summary card showing stats and AI Discover button

​
Best Practices

Re-run AI Discovery after major changes - When you add new features or significantly restructure code, run AI Discovery to update the feature map. Your user-defined features will be preserved.
Use directory patterns for precision - When creating user-defined features, add glob patterns like src/auth/** to ensure accurate commit classification. Multiple patterns can be specified, one per line.
Combine aggressively - If you see duplicate or overlapping features, use the Combine with… action. Fewer, well-defined features are more useful than many fragmented ones.
Name features from the user perspective - “User Authentication” is better than “auth-module” because it’s clearer in digests and insights.
Upgrade Analysis for better categorization - If your feature areas don’t have the three-level hierarchy, click Upgrade Analysis to re-process commits with improved AI categorization.

Products

Manage your products and access other features

Repositories

Link repositories to enable AI analysis

Updates (Digests)

See how feature areas appear in weekly digests

Insights

AI-powered analysis using feature context