Find saved working memory

Searches saved preferences, instructions, lessons, and procedures so an agent adapts to how you work. Use before work starts and when tasks change phase.

Created by Chris Moen • Version 16 • 22 steps

Personalise every AI interaction

The Find Saved Working Memory template acts as a bridge between your historical data and your current tasks. It allows an AI agent to search through your specific preferences, past lessons, and established procedures so it understands how you work. Instead of starting from scratch with each new prompt, the agent retrieves the relevant context needed to follow your unique brand guidelines or technical guardrails.

How the memory retrieval works

This workflow follows a structured path to find and load the most relevant context for your current project. First, it identifies the type of memory required and searches your internal Key-Value (KV) storage and tables to find the correct data bank. It then looks for two types of information: general working memories that apply to everything you do, and niche memories specific to the current workflow.

The flow uses search functions and logical matches to pull up to eight specific memory hits. By comparing your current task against these saved entries, the system returns a condensed context packet that the agent can read and follow immediately.

Key benefits for your team

Using this template helps you maintain consistency across different projects and team members. It ensures that your AI agents don't forget important rules or preferences once a conversation ends.

  • Consistency: Keep your brand voice and technical standards uniform across all automated tasks.
  • Contextual awareness: The agent understands when a task changes phase and adjusts its behaviour based on saved procedures.
  • Reduced manual input: Stop repeating the same instructions in every prompt by storing them in a centralised memory bank.
  • Scalability: As you learn new lessons or update your workflows, the memory bank grows with you, making your agents smarter over time.

Smart search and retrieval

This sequence is designed to be efficient. It doesn't just dump all your data at once. Instead, it carefully selects and matches search hits to find the most relevant "working memory" for the moment. This targeted approach keeps your AI interactions fast and accurate, providing only the information that is actually needed to get the job done right. If you want your agents to behave like they've worked with you for years, this memory retrieval flow is the right choice.

Steps

  1. Check what memory to look for (function)
  2. Find saved memory bank settings (kv)
  3. Read saved memory bank settings (function)
  4. Find the saved memory bank (search)
  5. Select the saved memory bank (function)
  6. Remember where the memory bank is (kv)
  7. Load memories for this workflow (table)
  8. Load memories for this workflow (table)
  9. Load general working memories (table)
  10. Load general working memories (table)
  11. Search memories for this workflow (search)
  12. Search general working memories (search)
  13. Match search hits to saved memories (function)
  14. Load matching memory 1 (table)
  15. Load matching memory 2 (table)
  16. Load matching memory 3 (table)
  17. Load matching memory 4 (table)
  18. Load matching memory 5 (table)
  19. Load matching memory 6 (table)
  20. Load matching memory 7 (table)
  21. Load matching memory 8 (table)
  22. Return memory context (function)

FAQ

What does the Find Saved Working Memory flow do?

This flow searches your saved preferences, instructions, and procedures to help an agent adapt to your specific working style. You should use it before starting a new task or when a project moves into a different phase.

How does the flow find and load my saved preferences?

The flow works by checking your memory bank settings, searching for relevant records, and matching search hits to your saved memories. It then loads specific memory context from tables to ensure the agent follows your established guardrails and lessons.

Which services or integrations do I need to connect?

No external integrations are needed as the flow manages data through internal tables, search functions, and key-value storage. It retrieves everything it needs from your existing memory bank and internal workflow tables.

Can I customise the types of memories the agent retrieves?

You can customise the flow by updating the memory bank settings and the specific tables where your procedures are stored. Since the flow loads matching memories across multiple steps, you can organise your instructions into different categories like general working habits or workflow-specific rules.