AI Assistant
The Flowstate AI Assistant helps you capture and structure ideas and requests. It’s an intelligent companion that listens to your thoughts and transforms them into actionable Sparks.
What Is the AI Assistant?
Section titled “What Is the AI Assistant?”The AI Assistant is an AI-powered interface that:
- Listens to your natural language input
- Classifies your input as ideas or requests
- Generates structured Sparks from your words
- Provides a conversational way to capture thoughts
The AI Assistant appears as a floating orb on your canvas, ready to help whenever inspiration strikes.
The AI Orb
Section titled “The AI Orb”Visual Design
Section titled “Visual Design”The AI Orb is a floating, animated sphere that lives on your canvas:
- Position: Hovers in a convenient location, movable if needed
- Animation: Subtle pulsing and movement to indicate it’s active
- Color: Changes based on state (see below)
- Size: Large enough to be noticeable, small enough to stay out of the way
The orb’s design is inspired by science fiction AI companions, creating a sense of presence and personality.
Dynamic States
Section titled “Dynamic States”The AI Orb changes appearance based on what it’s doing:
Idle State
Section titled “Idle State”When the AI is waiting for input:
- Appearance: Gentle pulsing, soft glow
- Color: Neutral blue or white
- Animation: Slow, rhythmic breathing effect
- Meaning: Ready to listen
The idle state is calm and unobtrusive, indicating availability without demanding attention.
Listening State
Section titled “Listening State”When the AI is receiving input:
- Appearance: Active pulsing, brighter glow
- Color: Vibrant blue or green
- Animation: Responsive to voice input (visual feedback)
- Meaning: Actively listening
The listening state provides feedback that your input is being received.
Thinking State
Section titled “Thinking State”When the AI is processing your input:
- Appearance: Rapid pulsing, shifting patterns
- Color: Purple or orange
- Animation: Complex, dynamic movement
- Meaning: Working on your request
The thinking state shows that the AI is analyzing your input and generating a response.
Interaction
Section titled “Interaction”You can interact with the AI Orb in several ways:
Voice Input
Section titled “Voice Input”Click the orb and speak:
- Click the AI Orb
- Speak your idea or request
- The orb enters listening state
- When you’re done, click again or pause
- The orb enters thinking state
- A Spark is generated and appears in the current Room
Voice input is the most natural way to capture thoughts quickly.
Text Input
Section titled “Text Input”Type your input:
- Click the AI Orb
- A text input field appears
- Type your idea or request
- Press Enter or click Submit
- The orb enters thinking state
- A Spark is generated
Text input is useful when voice isn’t practical or when you want to be precise.
Context Awareness
Section titled “Context Awareness”The AI Orb is aware of your current context:
- Which Room you’re in
- What Sparks already exist
- Your recent activity
This context helps the AI generate more relevant and useful Sparks.
AI Classification
Section titled “AI Classification”The AI Assistant’s core capability is classifying input as ideas or requests:
How Classification Works
Section titled “How Classification Works”When you provide input, the AI analyzes:
- Intent signals: Are you proposing something new (idea) or asking for something specific (request)?
- Language patterns: Exploratory language suggests ideas; directive language suggests requests
- Context: What Room are you in? What Sparks already exist?
- Structure: Does the input describe a vision or a problem to solve?
Based on this analysis, the AI classifies the input and generates the appropriate Spark type:
- Ideas become Blueprints
- Requests become Scopes of Work
Classification Accuracy
Section titled “Classification Accuracy”The AI is trained to classify accurately, but it’s not perfect:
- Typical accuracy: 85-95% for clear input
- Ambiguous input: May require clarification
- Correction: You can change the Spark type after creation if the AI misclassified
The AI learns from corrections, improving over time.
Handling Ambiguity
Section titled “Handling Ambiguity”If the AI isn’t sure how to classify your input, it may:
- Ask clarifying questions
- Suggest a classification with low confidence
- Generate a generic Spark and let you refine it
You’re always in control and can adjust the result.
Generating Structured Artifacts
Section titled “Generating Structured Artifacts”Once the AI classifies your input, it generates a structured Spark:
Blueprint Generation (Ideas)
Section titled “Blueprint Generation (Ideas)”For ideas, the AI creates a Blueprint with:
- Vision: The core concept and what it could become
- Key features: Main components or capabilities
- Benefits: Why this idea matters
- Open questions: Areas that need exploration
- Next steps: How to move forward
The AI extracts this structure from your natural language input, organizing your thoughts into a clear format.
Scope of Work Generation (Requests)
Section titled “Scope of Work Generation (Requests)”For requests, the AI creates a Scope of Work with:
- Problem statement: What needs to be addressed
- Desired outcomes: What success looks like
- Requirements: Must-haves and constraints
- Timeline: When it’s needed
- Next steps: How to proceed
Again, the AI structures your input, turning a conversational request into a clear specification.
Iterative Refinement
Section titled “Iterative Refinement”After generating a Spark, you can:
- Review the structured content
- Ask the AI to refine specific sections
- Add details manually
- Regenerate if the result isn’t right
The AI is a starting point, not the final word. You refine the Spark to match your intent.
Powered by OpenAI
Section titled “Powered by OpenAI”The AI Assistant uses OpenAI’s language models:
Model Selection
Section titled “Model Selection”Flowstate uses GPT-4 or similar models for:
- High-quality classification
- Nuanced understanding of intent
- Coherent, well-structured output
The specific model may vary based on performance and cost considerations.
Privacy and Security
Section titled “Privacy and Security”Your input is processed by OpenAI’s API:
- Input is sent to OpenAI servers
- OpenAI’s privacy policy applies
- Flowstate does not store raw input longer than necessary
- Generated Sparks are stored in Flowstate’s database
Review OpenAI’s privacy policy to understand how your data is handled.
API Limits
Section titled “API Limits”The AI Assistant is subject to API rate limits:
- Typical limit: 50 requests per hour per user
- Limits may vary based on your plan
- If you hit the limit, you’ll be notified and can try again later
Limits ensure fair usage and control costs.
AI Presence Block Type
Section titled “AI Presence Block Type”The AI Orb is represented on the canvas as an “ai_presence” block type:
Block Properties
Section titled “Block Properties”- Type:
ai_presence - Position: X and Y coordinates on the canvas
- State: Current state (idle, listening, thinking)
- Visibility: Always visible or hideable
Canvas Integration
Section titled “Canvas Integration”The AI presence block integrates with the canvas:
- Appears alongside other tiles (Rooms, Offices, Shelves, etc.)
- Can be positioned anywhere
- Persists across sessions
- Syncs state across devices
This makes the AI Assistant a first-class citizen of your workspace.
Use Cases
Section titled “Use Cases”Rapid Idea Capture
Section titled “Rapid Idea Capture”When inspiration strikes:
- Click the AI Orb
- Speak or type your idea
- The AI generates a Blueprint
- Continue working without losing the thought
The AI Assistant removes friction from idea capture, ensuring good ideas aren’t lost.
Structured Request Creation
Section titled “Structured Request Creation”When you need something:
- Click the AI Orb
- Describe what you need
- The AI generates a Scope of Work
- Share it with your team
The AI Assistant turns vague requests into clear specifications.
Brainstorming Sessions
Section titled “Brainstorming Sessions”During brainstorming:
- Open a Room for the session
- Use the AI Orb to capture each idea as it’s discussed
- The AI generates Blueprints for each idea
- Review and refine after the session
The AI Assistant acts as a scribe, capturing ideas so you can focus on thinking.
Meeting Notes
Section titled “Meeting Notes”During meetings:
- Use the AI Orb to capture action items and decisions
- The AI generates Sparks for each item
- Sparks are automatically organized in the Room
- Follow up is easier because everything is structured
The AI Assistant turns meeting notes into actionable artifacts.
Best Practices
Section titled “Best Practices”Be Specific
Section titled “Be Specific”The more detail you provide, the better the AI can structure your input. Instead of “we need a new feature,” try “we need a way for users to export their data as CSV files.”
Use Natural Language
Section titled “Use Natural Language”Don’t try to format your input. Just speak or type naturally. The AI is designed to understand conversational language.
Review and Refine
Section titled “Review and Refine”Don’t treat AI-generated Sparks as final. Review them, add details, and refine the structure. The AI gives you a head start, not a finished product.
Provide Context
Section titled “Provide Context”If you’re in a specific Room, the AI uses that context. If you’re working on a specific project, mention it. Context helps the AI generate more relevant Sparks.
Correct Misclassifications
Section titled “Correct Misclassifications”If the AI classifies an idea as a request (or vice versa), correct it. This feedback helps the AI learn and improves future classifications.
Use Voice When Appropriate
Section titled “Use Voice When Appropriate”Voice input is faster for capturing quick thoughts. Text input is better for precise, detailed input. Choose based on the situation.
Technical Details
Section titled “Technical Details”Data Flow
Section titled “Data Flow”- User provides input (voice or text)
- Input is sent to Flowstate backend
- Backend sends input to OpenAI API with classification prompt
- OpenAI returns classification and structured content
- Backend creates Spark in database
- Spark appears on canvas in real time
This flow typically takes 2-5 seconds from input to Spark creation.
AI Orb Component
Section titled “AI Orb Component”The AI Orb is implemented as a React component:
- State management: Tracks current state (idle, listening, thinking)
- Animation: Uses CSS animations and transitions
- Input handling: Captures voice and text input
- API integration: Communicates with backend for AI processing
The component is optimized for smooth animations and responsive interaction.
Voice Input Processing
Section titled “Voice Input Processing”Voice input uses the Web Speech API:
- User clicks the orb
- Browser requests microphone permission
- Speech recognition starts
- Audio is converted to text in real time
- Text is sent to backend when user finishes speaking
Voice input is processed locally in the browser, then sent to the backend as text.
Storage
Section titled “Storage”AI-generated Sparks are stored like any other Spark:
- Spark metadata (type, Room, position)
- Structured content (Blueprint or Scope of Work)
- Version history
- Timestamps
The fact that a Spark was AI-generated is tracked for analytics but doesn’t affect functionality.
Integration with Other Features
Section titled “Integration with Other Features”The AI Assistant integrates with other Flowstate features:
Sparks generated by the AI appear in the current Room, maintaining context and organization.
Canvas
Section titled “Canvas”The AI Orb is a canvas element, positioned and managed like other tiles.
Sparks
Section titled “Sparks”AI-generated Sparks are identical to manually created Sparks, with full versioning and collaboration features.
Future Enhancements
Section titled “Future Enhancements”Planned improvements to the AI Assistant include:
- Multi-turn conversations (ask follow-up questions)
- Proactive suggestions (AI suggests Sparks based on activity)
- Custom classification types (beyond ideas and requests)
- Voice commands for canvas navigation
- Integration with external AI services
- Personalized AI behavior based on your preferences
- Collaborative AI (AI assists entire teams, not just individuals)
The AI Assistant makes capturing and structuring thoughts effortless. By combining natural language understanding with structured output, it helps you move from inspiration to action faster.