Types are a retrieval tool. They are not just labels. When you store a memory with a clear type, it becomes easier to rank and easier to use safely. If you are unsure which type to pick, start withDocumentation Index
Fetch the complete documentation index at: https://docs.4stax.com/llms.txt
Use this file to discover all available pages before exploring further.
fact and preference. Add project and decision once you want stronger structure.
| Type | Description |
|---|---|
fact | Stable facts about you or your work |
preference | How you like to work |
project | Ongoing project context |
decision | Architectural or strategic decisions |
skill | Things you can do (and how you do them) |
episodic | Session-specific context worth keeping |
Examples
fact
Facts are stable and broadly useful.
Examples:
- “We use Next.js and TypeScript”
- “The API is deployed on Cloudflare”
preference
Preferences are stable preferences about style and workflow.
Examples:
- “Prefer TypeScript over JavaScript”
- “Prefer concise answers with code snippets”
project
Project entries are durable context that is not a single fact. It helps to attach a project label.
Examples:
- “Project 4stax is a local first memory layer expanding into a vault and API”
- “This repo is the documentation site. It uses Mintlify and MDX pages”
decision
Decisions should include a reason. The reason helps later when you reconsider.
Examples:
- “Decision: use SQLite locally because it is inspectable and simple”
- “Decision: do not store secrets in the vault to reduce risk”
skill
Skills describe repeatable workflows.
Examples:
- “I can set up a Next.js app with Tailwind and deploy it”
- “I can write API docs and keep code examples copy paste ready”
episodic
Episodic entries are useful later, but not permanent identity level facts.
Examples:
- “This week we are shipping v0.1.0 and focusing on onboarding”
- “Current bug: MCP host does not reload until full restart”