Memory
Solmyr doesn't start from scratch every time. It has persistent memory — who you are, what you're building, what you've told it, and everything it's produced. Three layers, always available, always informing every action.
Three layers of memory
Memory in Solmyr isn't one thing. It's three distinct systems that combine to give the AI full context about your business.
Business Context
Layer 1Everything you shared during setup — your vision, focus areas, market scope, budget, team size, experience levels, constraints, and chosen strategic direction. This is the foundational "who you are" data. It's captured once during the wizard and feeds into every AI prompt as persistent context.
Files & Knowledge Base
Layer 2Research reports, competitor analyses, strategy documents, notes — everything Solmyr produces or you upload. Organized in folders, browsable and previewable. The AI creates files as it works (research outputs, analysis results) and can reference them later. This is the "what we know" layer.
Custom Instructions
Layer 3Persistent rules you set — either through Settings or by telling Solmyr in chat. "Always check in every 30 minutes." "Prefer boards over lists." These are injected into the AI's context as standing instructions, separate from the business context. Think of it as your personal operating manual for the AI.
The Files screen
When you click "Memory" in the sidebar, you land on the file browser. It's the most tangible part of memory — folders and files you can actually see and read.
Folder browser
A sidebar tree of folders organized by topic. Solmyr creates and names folders as it works — research goes in research folders, strategy in strategy folders. Navigate like any file system.
File preview
Click any file to preview its contents in-app. Markdown renders with full formatting — no downloads, no context switching. You read what the AI wrote in the same interface where you manage everything else.
Uploads
Upload your own documents to expand the knowledge base. Business plans, market research, brand guidelines — anything the AI should reference when making decisions or producing work.
How memory becomes context
Raw data sitting in a database is useless. Here's how the three layers are assembled into a prompt every time the AI acts.
Layer 1: Business identity
Your vision, focus areas, market scope, resources, and constraints become the foundation for every conversation. Because this rarely changes, it's carried forward across sessions so the AI always starts from the same shared understanding.
Layer 2: Master plan
The latest version of your strategic plan is injected as a "Current Master Plan" section. If no plan exists yet, the AI gets a note saying so and adjusts behavior accordingly.
Layer 3: Operational state
Current tasks, pending decisions, workspace contents, and recent activity are assembled fresh each cycle. This is the dynamic layer — it changes with every action.
Custom instructions
Your persistent rules are appended as a separate section. The AI treats these as standing orders that override default behavior where applicable.
The result: every time Solmyr thinks, it has full awareness of who you are, what your strategy is, what's happening right now, and how you want it to behave. No "remind me what we're doing" moments.
Where memory starts
The setup wizard is the primary input for business context. Every question you answer — your vision, what you value, where you operate, your budget and team, your constraints — gets saved as part of the shared memory. If you did the operator flow, the AI's research on your business pre-fills the same fields.
Documents you attach during setup (pitch decks, existing plans, brand guides) are passed to the AI during the initial build process. The wizard also captures detailed optional fields — business background, current challenges, goals, financials, customer profiles, and work-style preferences — all of which become part of the extended context section.
Note: There's no dedicated editor for business context after setup. If your fundamentals change (new market, different budget), the effective approach is to update the master plan and custom instructions, or re-run setup. The business context is your starting point — the plan is where evolution happens.
Custom instructions
Custom instructions are the one part of memory you can edit anytime. Set them in Settings or just tell Solmyr in chat — "remember to always use formal language" and it'll update the instructions for you.
These are persistent across sessions. Unlike chat messages that scroll out of context, instructions stay in the AI's prompt indefinitely. They're ideal for behavioral preferences, communication style, recurring constraints, or workflow rules you don't want to repeat.
Examples
- "Always check in every 30 minutes"
- "Prefer boards over lists"
- "Use formal language in all outputs"
- "Budget cap is $500/month — flag anything over"
Memory vs. the Plan
Memory
Static foundation. Who you are, what your constraints are, what files exist, how you want the AI to behave. Rarely changes after initial setup (except files and instructions).
Master Plan
Living strategy. Where you're going, what phase you're in, what decisions have been made. Updated by the AI after every significant milestone or pivot.
Memory tells the AI who it's working for. The plan tells it what it's working toward.
Next up
Feedback
Automated feedback capturing — how the system learns from your responses.