Guides¶

Practical guides for using Sekha effectively.

Use Case Guides¶

AI Coding Assistant¶

Build an AI coding assistant with persistent memory:

AI Coding Assistant →

• Remember project context across sessions
• Track architectural decisions
• Build knowledge base of code patterns
• Context-aware code suggestions

Research Assistant¶

Create a research assistant that never forgets:

Research Assistant →

• Organize research notes semantically
• Track sources and citations
• Find related research across conversations
• Build comprehensive literature reviews

Core Concepts¶

Organizing Memory¶

Best practices for organizing conversations:

• Folders - Organize by project, topic, or context
• Labels - Tag conversations with themes
• Importance - Score 1-10 for prioritization
• Summaries - Auto-generate for long conversations

Semantic Search¶

Find memories by meaning, not just keywords:

• Query in natural language
• Relevance scoring with embeddings
• Cross-conversation search
• Time-based filtering

Context Assembly¶

Buil optimal context for your LLM:

• Budget-aware context selection
• Importance-weighted prioritization
• Summary integration for compression
• Source tracking for citations

Memory Pruning¶

Keep your memory system efficient:

• Identify low-value conversations
• Prune by age, importance, or relevance
• Selective deletion with summaries
• Export before deletion

Conversation Summarization¶

Compress long conversations:

• Automatic summarization
• Configurable detail levels
• Summary-aware context assembly
• Token budget optimization

Importing Data¶

Bring existing data into Sekha:

• ChatGPT export compatibility
• Claude conversation imports
• Custom JSON format
• Bulk import tools

Backup & Restore¶

Protect your memory:

• Export conversations as JSON
• SQLite database backups
• ChromaDB vector backups
• Automated backup scripts

Integration Guides¶

Integrate Sekha with your workflow:

• Claude Desktop - MCP integration
• VS Code (Coming Soon) - Editor integration
• Obsidian (Coming Soon) - Knowledge base sync
• CLI (Coming Soon) - Terminal interface

Best Practices¶

Conversation Labels¶

Use descriptive, searchable labels:

# Good
"Architecture Discussion - Auth System"
"Bug Fix - Payment Processing"
"Research - Vector Databases"

# Not ideal
"Chat 1"
"Conversation"
"Notes"

Folder Structure¶

Organize by project or domain:

work/
  ├── project-a/
  ├── project-b/
  └── research/

personal/
  ├── learning/
  └── creative/

Importance Scoring¶

Use consistent scoring:

• 9-10 - Critical decisions, breakthroughs
• 7-8 - Important discussions, solutions
• 5-6 - Regular conversations, notes
• 3-4 - Minor discussions
• 1-2 - Trivial or test conversations

Next Steps¶

• AI Coding Assistant - Build a coding assistant
• Research Assistant - Research workflow
• Claude Desktop - MCP integration
• API Reference - Programmatic access