Auto Memory

Version 1.0.0 is out now! Featuring completely rewritten functionality for better performance and much better memory management

✨ What It Does

Auto Memory listens in on your conversations and detects facts, preferences, key moments, or anything useful for the assistant to remember about you. It stores these as separate memories, so future AI interactions stay personal and context-aware—without you micromanaging.

You get:

Seamless journaling of your important info
Smarter, context-rich AI assistance
No “please remember X” (unless you want to!)

Note: Make sure to enable the Memory feature in your user profile settings (Profile → Settings → Personalization) to allow models to access your memories!

💾 How It Works

Auto-extracts new or changed "facts" from recent user messages
Stores each fact separately in your Memory database
Auto-maintains Memories: merges duplicates, resolves conflicts, and prunes old/irrelevant ones
Uses advanced LLMs to understand context and nuance

🚀 Installation

Make sure your Open WebUI is version 0.5.0 or newer
Click on GET to add the extension to your Open WebUI deployment
Configure API keys and model (see below)

⚙️ Configuration

Configure via the Open WebUI extension settings or directly in code:

Setting	Description	Default
`openai_api_url`	OpenAI-compatible API endpoint	`https://api.openai.com`
`model`	LLM model for memory identification	`gpt-5-mini`
`api_key`	API key for the chosen endpoint	(empty)
`related_memories_n`	Number of related memories to check for consolidation	`5`
`related_memories_dist`	Similarity distance threshold for related memories	`0.75`
`messages_to_consider`	How many recent messages to consider (user+assistant)	`4`
`allow_unsafe_user_overrides`	SECURITY WARNING: Allow users to override API URL/model without providing their own API key. This could allow users to steal your API key or use expensive models at your expense. Only enable if you trust all users.	`false`
`debug_mode`	Enable detailed logging for troubleshooting	`false`

User-Level Settings

Users can override settings individually:

Setting	Description
`enabled`	Whether to enable Auto Memory for this user
`show_status`	Display memory save status on UI
`openai_api_url`	User-specific API endpoint (overrides global)
`model`	User-specific model (overrides global)
`api_key`	User-specific API key (overrides global)
`messages_to_consider`	User-specific message count (overrides global)

Security Note: User overrides for openai_api_url and model require the user to provide their own api_key. Without it, the admin's global settings are used for security. Admins can enable allow_unsafe_user_overrides to bypass this check, but this is not recommended.

🧠 Memory Extraction Logic

New or changed facts from User's latest message are saved.
Explicit "please remember..." requests always create a Memory.
Avoids duplicates & merges conflicts.
Automatically deletes and maintains Memories over time.

Example

Conversation:

text

-4. user: I love oranges 😍
-3. assistant: That's great!
-2. user: Actually, I hate oranges 😂
-1. assistant: omg you LIAR 😡

Memory stored:

python

["User hates oranges"]

See full logic and more cases in code.

🙌 Credits

Created by nokodo

💖 Support & Feedback

📜 License

You are free to use, run, and modify this extension for personal or internal purposes. You may NOT redistribute, publish, sublicense or sell this extension or any modified version without prior explicit written consent from the author. All copies must retain this notice. Provided “AS IS” without warranty. Earlier pre-release versions may have been available under different terms.

Keep your AI tuned in to who you really are—automatically!