Memory System

One-Liner

The Memory System provides a unified interface for managing persistent knowledge across sessions, including core memory, workflow context, CLI history, vector search, and V2 pipeline operations.

---

Pain Points Solved

Pain Point	Current State	Memory System Solution
Cross-session amnesia	New session requires re-explaining project	Persistent memory across sessions
Lost decisions	Architecture decisions forgotten	Decision log persists
Repeated explanations	Same context explained multiple times	Memory auto-injection
Knowledge silos	Each developer maintains own context	Shared team memory
Scattered data sources	Core/workflow/CLI history separate	Unified search across all categories

---

Page Overview

Location: ccw/frontend/src/pages/MemoryPage.tsx

Purpose: View and manage core memory with CRUD operations and unified vector search.

Access: Navigation → Memory

Layout

+--------------------------------------------------------------------------+
|  Memory Title                                              [Refresh] [Fullscreen]|
+--------------------------------------------------------------------------+
|  [Core] [Workflow] [CLI History] [Search] [V2 Pipeline]                      |
+--------------------------------------------------------------------------+
|  Tab Content (varies by active tab)                                        |
|  +--------------------------------------------------------------------+    |
|  |                                                                    |    |
|  |  - Memory list with CRUD operations                                 |    |
|  |  - Search interface with filters                                    |    |
|  |  - Source type badges and tags                                      |    |
|  |  - V2 pipeline job monitoring (V2 tab)                              |    |
|  |                                                                    |    |
|  +--------------------------------------------------------------------+    |
+--------------------------------------------------------------------------+

---

Core Features

Feature	Description
Tabbed Interface	Five tabs: Core, Workflow, CLI History, Search, V2 Pipeline
CRUD Operations	Create, edit, delete memories with confirmation dialogs
Unified Search	Vector search across all memory categories
Source Type Filtering	Filter by core_memory, workflow, cli_history
Tag Filtering	Filter memories by custom tags
Batch Operations	Select all, delete selected
Recommendations	View memory recommendations from AI
Re-index	Rebuild memory index
Export	Export search results to file
V2 Pipeline	Monitor extraction and consolidation jobs
Immersive Mode	Fullscreen toggle for focused work

---

Memory Categories

Tab	Source Type	Description
Core	core_memory	Persistent project knowledge stored in ~/.claude/memory/
Workflow	workflow	Session context from .workflow/.memory/
CLI History	cli_history	Command execution history
Search	All	Unified vector search across all categories
V2 Pipeline	-	Extraction and consolidation job monitoring

---

Usage Guide

Basic Workflow

1. Browse Memories: Switch between Core/Workflow/CLI History tabs
2. Create Memory: Click "New Memory" button, fill in the form
3. Edit Memory: Click edit button on memory card, modify content
4. Delete Memory: Click delete button, confirm in dialog
5. Search: Switch to Search tab, enter query to find memories
6. Filter Results: Use source type and tag filters to narrow results

Key Interactions

Interaction	How to Use
Create memory	Click "New Memory" button, fill form, submit
Edit memory	Click edit (pencil) button on memory card
Delete memory	Click delete (trash) button, confirm
Search memories	Switch to Search tab, type query, filter results
Export results	Select memories, click "Export Selected"
Re-index memory	Click "Re-index" to rebuild search index
Select all	Click "Select All" checkbox to select all visible memories
Delete selected	Click "Delete Selected" to batch delete

Source Type Badges

Source Type	Color	Description
core_memory	Blue	Persistent project knowledge
workflow	Green	Workflow session context
cli_history	Amber	CLI command history

Memory Metadata

Each memory entry contains:

• ID: Unique identifier (format: CMEM-YYYYMMDD-HHMMSS)
• Content: Markdown-formatted text
• Tags: Array of tag strings for categorization
• Created/Updated: Timestamps
• Source Type: Category identifier

---

V2 Pipeline

The V2 Pipeline tab monitors background jobs for:

Job Type	Description
Extraction	Extract knowledge from sessions
Consolidation	Consolidate and summarize memories
Embedding	Generate vector embeddings for search

Job Status

Status	Description
Pending	Job queued, not started
Running	Job in progress
Done	Job completed successfully
Error	Job failed with error message

---

Components Reference

Main Components

Component	Location	Purpose
MemoryPage	@/pages/MemoryPage.tsx	Main memory management page
V2PipelineTab	@/components/memory/V2PipelineTab.tsx	V2 pipeline job monitoring
MemoryCard	(internal to MemoryPage)	Individual memory display

State Management

• Local state:

  • activeTab: TabValue
  • searchQuery: string
  • selectedMemories: Set<string>
  • filters:
  • dialogStates: create, edit, delete

• Custom hooks:

  • useMemory - Core memory CRUD operations
  • useMemoryMutations - Memory mutations (create, update, delete)
  • useUnifiedSearch - Vector search across categories
  • useUnifiedStats - Memory statistics
  • useRecommendations - AI-generated recommendations
  • useReindex - Rebuild search index

---

Configuration

No configuration required. The memory system automatically fetches data from the backend.

Memory API Endpoints

Operation	Endpoint
List memories	GET /api/memory
Create memory	POST /api/memory
Update memory	PUT /api/memory/:id
Delete memory	DELETE /api/memory/:id
Search	POST /api/memory/search
Re-index	POST /api/memory/reindex
V2 Jobs	GET /api/memory/v2/jobs

---

Related Links

• Spec System - Constraint injection
• CLI Call - Command line invocation
• Explorer - Codebase navigation
• Dashboard - Project overview