Requirement Agent
The Requirement Agent generates application schemas from your requirements using industry context, tool templates, and a prompt or uploaded PRD.
- Better accuracy — Industry workflows and compliance requirements are automatically included.
- Consistent output — Structured format for predictable schema generation.
- Transparent results — View the raw AI response for verification and debugging.
- Fast prototyping — Demo-ready blueprints for pre-sales and onboarding.
AI features require LLM API keys to be set up. If keys are missing:
- A warning appears: "API key for [model] is not configured"
- Click Configure API Key to open LLM Configuration
- Enter your API credentials and save
See Profile → LLM Configuration for setup instructions.
Go to Projects → New Project and select Build an app with AI, or use the Build App option from an existing project's sidebar.
How it works
- Select Industry — Choose domain context
- Enter Prompt / Upload PRD — Describe your requirements
- Select AI Model — Choose generation model
- Generate — Create the blueprint
- Review Blueprint — Check the generated schema
- Modify or Create — Adjust if needed, then create the project
Setup
Industry domain
Choose the industry that best matches your application.
| Industry | Example Workflows |
|---|---|
| Aviation | Crew scheduling, duty time regulations (FAA/EASA), aircraft maintenance, disruptions |
| Healthcare | Sample tracking, equipment scheduling, audit logging, HIPAA-like compliance |
| Retail | Inventory management, POS integration, customer loyalty |
| Logistics | Shipment tracking, warehouse management, route optimization |
| HRM / ERP | Employee onboarding, payroll, leave management |
Selecting an industry activates domain-specific workflows, so your generated schema includes expected entities and compliance requirements.
Tool Templates
Tool templates act as context anchors for the agent. They help generate more specific, relevant schemas.
Example templates:
- Crew Scheduling System
- Sample Tracking System
- Procurement Management
- Employee Onboarding Portal
Select at least one relevant tool template to avoid generic outputs and improve blueprint quality.
Requirements
You can provide requirements in one or both ways:
Option A: Enter a prompt
Describe your application in the text area. Be specific about modules, workflows, and business rules.
Good prompt example:
Build a procurement management system for multiple departments with:
- Multi-level approval workflows
- Supplier onboarding and rating
- Purchase order tracking
- Budget allocation per department
- Comprehensive audit logs
Option B: Upload PRD / documents
Upload a formal PRD or requirement document.
| Format | Status |
|---|---|
| Supported | |
| DOCX | Supported |
| Max file size | Configurable (default: 10MB) |
- At least one input is required: prompt entered OR PRD uploaded
- If prompt is empty, PRD upload becomes mandatory
Model
Choose the AI model for requirement generation.
| Setting | Default |
|---|---|
| Model | Gemini (org-configured) |
Generate
When you click Next or Generate, the Requirement Agent:
- Extracts PRD text (if uploaded)
- Normalizes inputs into a structured prompt format
- Injects industry + tool context
- Calls the selected AI model
- Generates raw response and parsed blueprint
Output sections
1. AI Response (Raw)
Hyper displays the raw AI output exactly as returned for transparency.
Why this matters:
- Builds trust in the generation process
- Helps with debugging and iteration
- Strengthens pre-sales demo confidence
- Allows verification of AI assumptions
Includes:
- Copy button
- Timestamp
- Model name
2. Generated Blueprint (Parsed)
The raw response is parsed into a structured blueprint with:
| Artifact | Description |
|---|---|
| Collections/Entities | Tables with their purposes |
| Fields | Field types and validation hints |
| Relationships | Foreign keys and associations |
| Roles/ACL Suggestions | Access control recommendations |
| Module Breakdown | Logical grouping of functionality |
3. Review Tabs
Switch between views:
| Tab | Shows |
|---|---|
| Relations | Interactive ERD diagram showing table relationships |
| Collections | Table list with field details and properties |
| Access Control | Permission matrix for each role |
Blueprint review
"How does this look?"
After generation, you see:
- Summary text explaining the generated structure
- Collection chips for quick navigation to each entity
- Schema visualization in ERD or table format
- + Modify Schema button for manual adjustments
Relations tab
Displays an interactive ERD diagram showing all collections and their relationships. Use the zoom controls in the bottom-right to navigate complex schemas.
Collections tab
Shows all generated tables in the sidebar. Click any collection to see:
- Field names and types
- Property badges: Primary key, Unique, Indexed, Required
- Field-level validation hints
Access control tab
Displays the Access Control Matrix showing permissions per role across all collections.
| Column | Description |
|---|---|
| Table | Collection name |
| Role columns | One column per generated role (for example, admin, doctor, nurse) |
| Actions | Create, Read, Update, Delete, List, Detail, Select, Upload |
Permissions are represented by tri-state checkboxes:
- indicates granted permissions
- indicates denied access
- indicates no explicit rule (inherited)
Iteration and regeneration
You can iterate without restarting:
| Action | What Happens |
|---|---|
| Edit prompt | Modify your description |
| Regenerate | Generate a new blueprint |
| Keep context | Industry, tools, model selection remain intact |
Each iteration keeps your previous selections, so you can experiment with different prompt variations without starting over.
Error handling
PRD extraction failure
If text extraction fails:
- Error message: "Unable to extract text from PRD"
- You can retry the upload
- Your prompt input remains safe
Model failure / timeout
If the model fails:
- Error state is shown with retry option
- All selections remain intact:
- Industry
- Tools
- Prompt
- PRD
- Model selection
Best practices
| Recommendation | Why |
|---|---|
| Select accurate industry domain | Activates domain-specific workflows |
| Choose relevant tool template | Provides contextual anchors |
| Write clear prompts with modules + workflows | Improves generation accuracy |
| Upload PRD when available | Maximizes detail for generation |
| Review raw AI output | Verify assumptions before creating |
FAQ
Do I need to upload a PRD?
No. PRD upload is optional if you enter a prompt. If your prompt is empty, PRD upload becomes mandatory.
Why show the raw AI response?
To improve transparency and trust. It helps teams debug, iterate, and validate AI-generated assumptions.
Can I regenerate without losing my input?
Yes. You can edit your prompt and regenerate while keeping your selected domain, tools, PRD, and model.
What if the schema isn't quite right?
Click + Modify Schema to manually adjust collections, fields, and relationships before creating the project.
Related
- Creating a Project — Full project creation flow
- Collections — Working with collections after creation
- Fields — Available field types
- Hyper AI — Other AI-powered features