AI Assistant

Overview and Access

The AI Assistant is a built-in chatbot powered by a large language model. It understands your database structure and can answer natural language questions about your actual business data - active projects, outstanding invoices, employee details, and much more.

Two ways to open it

• Floating button: Click the chat bubble icon in the bottom-right corner of any page. The button pulses gently to indicate it is available.
• Keyboard shortcut: Press Ctrl + K on Windows/Linux or Cmd + K on Mac to toggle the chat panel open and closed.

The chat panel

When the assistant opens, a side panel slides in from the right. It includes a message input area at the bottom, a scrollable conversation history in the middle, and a header with the assistant title. A clear button appears in the header once you start a conversation, allowing you to reset the chat at any time.

Before you send your first message, the panel displays a set of suggested questions drawn from common queries about your data. Click any suggestion to send it instantly, or type your own question in the input field and press Enter.

Asking Questions

Type your question in plain, natural language. The assistant translates your question into a database query behind the scenes, runs it against your live data, and returns a formatted answer. You do not need to know SQL or any technical query language.

Example questions

Here are some of the kinds of questions you can ask:

• "How many active projects do we have?"
• "Which invoices are overdue?"
• "What is the total payments received this month?"
• "Show me clients without any projects"
• "Which employees have React skills?"
• "Top 5 projects by contract value"
• "How many proposals were converted to projects?"
• "List maintenance contracts expiring soon"
• "Total purchase amount by supplier"
• "How many job seekers are in the system?"

The assistant understands the full database schema - 20 core business tables covering every entity in the platform. Responses are generated in real time, and you will see a typing indicator with an elapsed-time counter while the assistant processes your request.

Data Queries and Insights

The AI Assistant can query across all major entities in your workspace. This includes:

• Clients - customer records, project counts, total invoiced amounts
• Projects - active, completed, on hold, contract values, deadlines
• Invoices - pending, overdue, paid, amounts by status or currency
• Payments - collection totals, payment history, amounts by period
• Proposals - pending, approved, rejected, conversion rates
• Employees - headcount, compensation types, skills, project assignments
• Seekers - recruitment pipeline, skills, hiring status
• Maintenance contracts - active contracts, expiring soon, payment coverage
• Maintenance payments - payment history for service contracts
• Suppliers - vendor directory and purchase history
• Purchases - expense totals by period, supplier, or currency
• Referrals - referral sources and commission tracking
• Employee payments - salary payments, pending payroll, expenses by period
• Skills and tags - skill distribution, project categorization

How it works

When you ask a question, the system provides the AI with a compact representation of your database structure - table names, column names, and relationships between entities. The AI uses this context to generate a read-only database query, which is then executed against your live data. A second AI pass formats the raw results into a clear, human-readable answer. The entire process typically completes in a few seconds.

Cross-entity queries

The assistant understands how entities relate to each other and can answer questions that span multiple tables. For example:

• "Which projects for client X have unpaid invoices?"
• "How much has each client paid in total?"
• "Which employees are assigned to the most projects?"
• "Show me the total collected for each project"
• "Which clients were referred by whom?"

Aggregations and summaries

You can ask for totals, counts, averages, and rankings across any entity. The assistant handles grouping by period (this month, this year, this quarter), by currency, by status, or by any other dimension in your data.

Tips and Limits

Rate limit

Each user can send up to 10 requests per minute. If you exceed this limit, you will see a "Too many requests" message. Wait a moment and try again - the limit resets on a rolling 60-second window.

Tips for better results

• Be specific: "How many active projects do we have?" works better than "Tell me about projects."
• Use entity names: Mention the specific entity you are asking about - clients, invoices, employees, projects, proposals, maintenances, purchases, seekers, and so on.
• Ask one question at a time: The assistant handles focused questions more reliably than compound requests with multiple parts.
• Include time frames: "This month", "this year", "in the last 30 days" help the assistant scope results to the period you care about.
• Keep questions under 500 characters: The assistant enforces a 500-character limit on input to ensure reliable processing.

What the assistant cannot do

• It cannot modify data. The assistant is strictly read-only. It cannot create, update, or delete any records in your system.
• It cannot access files or documents. Uploaded files, PDFs, images, and attachments are not accessible to the chatbot.
• It cannot send emails or notifications. The assistant only retrieves and displays information.
• It cannot answer non-business questions. The assistant is designed to answer questions about your data only. General knowledge, coding help, or off-topic requests will be declined.

Handling unexpected answers

If the assistant gives an unexpected or incorrect answer, try rephrasing your question. You can also click "View SQL" on the response to see exactly what query was generated - this can help you understand how your question was interpreted and adjust accordingly.