The --auto-memory flag enables automatic extraction and storage of important information from conversations.
Quick Start
praisonai "My name is John and I prefer Python" --auto-memory
Usage
Basic Auto Memory
praisonai "I work at Acme Corp as a software engineer" --auto-memory
🧠 Auto Memory enabled
╭─ Agent Info ─────────────────────────────────────────────────────────────────╮
│ 👤 Agent: DirectAgent │
│ Role: Assistant │
│ 🧠 Auto Memory: Enabled │
╰──────────────────────────────────────────────────────────────────────────────╯
╭────────────────────────────────── Response ──────────────────────────────────╮
│ Nice to meet you! It's great to know you're a software engineer at Acme │
│ Corp. How can I help you today? │
╰──────────────────────────────────────────────────────────────────────────────╯
💾 Memories Extracted:
┌─────────────────────┬────────────────────────────┐
│ Type │ Content │
├─────────────────────┼────────────────────────────┤
│ Entity (Person) │ User works at Acme Corp │
│ Entity (Role) │ Software Engineer │
│ Long-term │ User's workplace: Acme Corp│
└─────────────────────┴────────────────────────────┘
With User Isolation
praisonai "Remember my preferences" --auto-memory --user-id user123
🧠 Auto Memory enabled (user: user123)
╭────────────────────────────────── Response ──────────────────────────────────╮
│ I'll remember your preferences. What would you like me to remember? │
╰──────────────────────────────────────────────────────────────────────────────╯
💾 Memory stored for user: user123
Combine with Other Features
# Auto memory with session
praisonai "Learn about my project" --auto-memory --session my-project
# Auto memory with planning
praisonai "Plan my learning path" --auto-memory --planning
# Auto memory with metrics
praisonai "Remember this" --auto-memory --metrics
How It Works
- Conversation Analysis: The system analyzes the conversation
- Information Extraction: Important information is identified
- Categorization: Information is categorized (entities, facts, preferences)
- Storage: Memories are stored for future retrieval
- Retrieval: Memories are automatically injected into future conversations
| Type | Description | Example |
|---|
| Entities | People, places, organizations | ”User works at Google” |
| Facts | Factual information | ”Project deadline is Dec 31” |
| Preferences | User preferences | ”Prefers Python over JavaScript” |
| Context | Contextual information | ”Working on ML project” |
Use Cases
Personal Assistant
# First conversation
praisonai "I'm learning Rust and prefer hands-on examples" --auto-memory
# Later conversation (memories recalled)
praisonai "Give me a coding exercise" --auto-memory
🧠 Auto Memory enabled
📚 Recalled Memories:
• User is learning Rust
• User prefers hands-on examples
╭────────────────────────────────── Response ──────────────────────────────────╮
│ Since you're learning Rust and prefer hands-on examples, here's a practical │
│ exercise: │
│ │
│ **Exercise: Build a Simple CLI Calculator** │
│ │
│ ```rust │
│ use std::io; │
│ │
│ fn main() { │
│ // Your code here │
│ } │
│ ``` │
╰──────────────────────────────────────────────────────────────────────────────╯
Project Context
# Share project details
praisonai "I'm building an e-commerce platform using Django and React" --auto-memory
# Ask for help (context remembered)
praisonai "How should I structure my API?" --auto-memory
Team Preferences
# Different team members
praisonai "I prefer detailed explanations" --auto-memory --user-id alice
praisonai "I prefer concise answers" --auto-memory --user-id bob
Viewing Stored Memories
Use the memory command to view what’s been stored:
# Show memory statistics
praisonai memory show
# Search memories
praisonai memory search "Python"
Memory Statistics
┏━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ Property ┃ Value ┃
┡━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━┩
│ user_id │ default │
│ short_term_count │ 5 │
│ long_term_count │ 12 │
│ entity_count │ 8 │
│ storage_path │ .praison/memory/default │
└──────────────────┴─────────────────────────┘
Recent Memories:
• User prefers Python for backend
• User works at Acme Corp
• Project: E-commerce platform
• Tech stack: Django + React
Memory Persistence
Memories are stored locally and persist across sessions:
.praison/
└── memory/
├── default/
│ ├── short_term.json
│ ├── long_term.json
│ └── entities.json
└── user123/
├── short_term.json
├── long_term.json
└── entities.json
Best Practices
Use --user-id to keep memories separate for different users or projects.
Auto memory increases token usage as memories are injected into prompts. Monitor with --metrics.
User Isolation
Use --user-id for multi-user scenarios
Regular Cleanup
Periodically clear old memories with praisonai memory clear
Combine with Sessions
Use with --session for project-specific memory
Monitor Usage
Use --metrics to track memory-related token costs
Privacy Considerations
Memories are stored locally on your machine. No data is sent to external servers for memory storage.
- Memories are stored in
.praison/memory/
- Use
praisonai memory clear all to delete all memories
- Each user ID has isolated memory storage
