> ## Documentation Index > Fetch the complete documentation index at: https://ai-kb.automationanywhere.com/llms.txt > Use this file to discover all available pages before exploring further. # Smart Tables > Learn how to use Smart Tables to store and look up data **Smart Tables** are a structured data management feature within the Knowledge Base (KB) that allow you to store, organize, and extract information in a tabular format. Unlike standard spreadsheets, Smart Tables are integrated with AI capabilities, enabling automatic data extraction from documents, relationship mapping between different data sets (Collections), and direct querying via AI Agents. Key capabilities include: * **Defining structured schemas** — columns, data types, and formats * **Extracting data automatically** from uploaded files (e.g., Resumes, Invoices) * **Visualizing data** through built-in dashboards * **Querying table data** using natural language via Agents *** ## Accessing Smart Tables Smart Tables are accessed through the **Knowledge Base (KB)** section of the application. Once created, you can view and manage your Smart Tables by navigating to the **KB Data** section, where all tables are listed and can be opened for viewing or editing. KB Data section showing list of Smart Tables *** ## Creating Smart Tables There are three primary methods to create a Smart Table within the Knowledge Base. Access these options by clicking the creation menu in the KB interface. Three options menu for creating Smart Tables ### Method A: Create Empty Smart Table This option creates a blank table with default system columns (Title, Description, Number). A sidebar configuration panel opens automatically, allowing you to define your custom schema from scratch. You can keep these default fields, modify them, or delete them to create your own structure. Sidebar panel with default fields ### Method B: Create from Template You can select from pre-defined templates to quickly set up tables for common use cases. Multiple templates are available for different scenarios. For example, the **Evaluations Template** imports a pre-defined set of columns designed for testing agent responses (Question Text, Expected Answer, Actual Response). Template selection dialog ### Method C: Import from File You can upload a **CSV file** to automatically create a Smart Table. The system will parse the CSV headers to create columns and import the rows as data entries. CSV import interface *** ## Column Types and Configuration When configuring a Smart Table, each column requires specific definitions to ensure accurate data storage and extraction. A column consists of a **Name**, **Type**, **Format**, and **Tool Source**. Providing a clear **Description** for each column is critical. This description guides the Large Language Model (LLM) on what specific information to extract or generate for that field. ### Configuring Columns When creating or editing a Smart Table, use the sidebar panel to configure each column. After setting up a field's properties (Name, Type, Format, Tool, and Description), click **Save Field** to apply the configuration. You can edit existing fields at any time by clicking on the field name in the column configuration panel, making changes, and saving again. Column configuration sidebar ### Field Types and Formats | Field Type | Description | Format Options | | ----------------- | --------------------------------------------- | ------------------------------------------------------------------------- | | **Text** | Standard alphanumeric string | Text, Phone Number, Email | | **Number** | Numeric values | Number, Decimal, Currency | | **Single Select** | Select one value from a predefined list | Dropdown with options (e.g., Skill Level: Beginner, Intermediate, Expert) | | **Multi-select** | Select multiple values from a predefined list | List of options (e.g., Languages: Python, Java, HTML, CSS) | | **Checkbox** | Boolean values (True/False) | True/False (e.g., Is Female?) | | **Date** | Calendar dates | Date formats (DD/MM/YYYY, etc.) | | **User** | Reference to a user in the system | Username or User Email | | **File** | Attachment field for documents | File upload | | **Collection** | Link to another Smart Table (relational) | Select target Smart Table | Field type dropdown options ### Tool Sources The **Tool** setting defines how data enters the cell: * **User Input** — Data is manually entered by the user or extracted directly from a file without modification * **LLM Generated** — The AI model generates the content based on the column description and context (e.g., summarizing a resume) * **Agent Generated** — An Agent processes a query and populates this field with the response * **Comparison** — Compares two fields and determines if they match or calculates similarity scores Tool source dropdown *** ## Data Extraction (Document Processing) Smart Tables can automatically populate rows by extracting data from uploaded documents. This is particularly useful for processing standardized documents like resumes or forms. Each uploaded document creates one row in the Smart Table, though multiple documents can be uploaded and processed to create multiple rows in the same table. ### Workflow: Resume Extraction Example Create a Smart Table named "Resumes" with the following columns: * Name (Text) * Contact Number (Text — Phone Number format) * Years of Experience (Number — Decimal format) * Languages (Multi-select: Python, Java, HTML, CSS) * Is Female (Checkbox) * Resume File (File) Navigate to the Knowledge Base and upload a file (e.g., a PDF resume). During upload, select the "Resumes" Smart Table. The system processes the document and maps content to the defined columns — for example, finding "3 years" in the text and populating Years of Experience, or extracting a phone number for Contact Number. Navigate to the Smart Table to view the newly created row with extracted data. File upload modal with Smart Table assignment If a file is uploaded without an assignment, it can be assigned later via the document settings. Go to the document, click **Assign Smart Table**, select the target Smart Table, and click **Apply** to trigger the extraction process retroactively. Document settings with Smart Table assignment Enable **Quick Upload** when assigning a Smart Table during upload to automatically extract and populate data as soon as the file upload completes. Resume Smart Table with extracted data *** ## Collections and Relationships **Collections** allow for many-to-many mappings between Smart Tables. This is useful when a single row in a main table contains complex list data that requires its own structure. ### Example: Skills Mapping Instead of storing skills as a simple comma-separated string (e.g., "Python, Java"), you can create a relational structure with detailed skill information: 1. **Create "Skills" Table** — Define columns: * Skill Name (Text) — Name of the skill the user possesses * Skill Level (Single Select) — Options: Beginner, Intermediate, Expert 2. **Configure Main Table** — In the "Resumes" table, create a column named "Skills". 3. **Set Type to Collection** — Set the column type to **Collection** and link it to the "Skills" Smart Table. 4. **Result** — When a resume is processed, the system extracts skills and populates the "Skills" table with detailed rows (e.g., "Python - Expert", "Java - Beginner", "AWS - Intermediate") linked back to the specific candidate via the many-to-many relationship. Skills Smart Table with linked data ### Viewing Collection Data When viewing the main table (e.g., Resumes), the Collection field displays a link to the related entries. Clicking on it navigates to the linked table (e.g., Skills) where you can see all the detailed entries for that candidate. The relationship is bidirectional — you can view which resume each skill belongs to by checking the Content Key in the Skills table. Collection field in Resume table *** ## Data Management Once data is populated, you have several tools to manage and organize it. ### Views and Tabs Smart Tables provide two views: * **Table View** — The default spreadsheet-like view for viewing and editing rows and columns * **Dashboard View** — A visualization view where you can create charts and widgets to analyze your data Table View and Dashboard View tabs ### Importing Data into an Existing Table You can bulk-import data into an existing Smart Table using the **Import Data** button in the table toolbar. This is useful when you need to refresh a table from an external file or add new rows without recreating the table from scratch. **Import Data** updates an existing table. To create a new Smart Table from a file, use **Import from File** when creating a table (see [Method C: Import from File](#method-c-import-from-file) above). The import workflow has three steps: Click **Import Data** and upload a file. Supported formats are **CSV**, **Excel** (XLSX, XLS), **TXT**, and **JSON**. * For **CSV**, **Excel**, and **TXT** files, the first row is used as column headers. * **JSON** files should contain exported data type configurations and will recreate the table structure with fields and views. Choose how the imported data should be applied to the existing table: * **Replace All Data** — Deletes all existing rows before importing the new data. Use this when you want to fully refresh the table contents. * **Append New Rows** — Adds the imported rows to the existing data without removing current entries. Map each **Source Column** from your file to a **Target Column** in the Smart Table and set the **Data Type** for each field. A **Data Preview** at the bottom of the dialog shows a sample of the rows to be imported. Review the **Import Summary**, which shows the file name, table name, number of columns, and number of rows. Click **Import Data** to complete the import. ### Adding and Managing Rows * **Insert Row** — Click **Add Row** or **Insert Row** to manually add a new blank row * **Automatic Row Creation** — When a document is uploaded and assigned to a Smart Table, a new row is automatically created and populated with extracted data Add Row button and action ### Editing and Re-computing * **Manual Edit** — Click any cell to manually update the value * **Execute Cell/Row** — For LLM-generated or Agent-generated fields, click **Execute Cell** or **Execute Row** to compute or refresh that row's data * **Re-compute All** — Refreshes all rows simultaneously. Useful after updating column descriptions or agent configurations to re-process the entire dataset Execute Cell/Row and Re-compute All buttons Each row is automatically assigned a unique **Content Key** identifier that links it back to the source document. This enables traceability between extracted data and original files. Content Key column ### Grouping and Organizing You can group data by specific columns to organize the view. For example, grouping a candidate list by the "Language" multi-select field lets you see all Python developers together. Columns can also be rearranged by dragging and repositioning them in the table view. Group By functionality Column rearrangement ### Renaming Smart Tables Click on the table name at the top and enter a new name. This helps organize multiple tables with descriptive identifiers (e.g., "Resumes", "Skills", "Evaluations"). Rename Smart Table ### Deleting Fields and Rows * **Delete Column** — When editing a column configuration, select **Delete Field** to remove the column entirely * **Delete Row** — Click on a row and select **Delete** to remove individual entries *** ## Dashboards and Visualization Smart Tables include a built-in **Dashboard** view to visualize stored data. ### Creating Widgets 1. Navigate to the **Dashboard** tab within the Smart Table 2. Click **Add Widget** 3. Configure the chart: * **Widget Title** — Provide a descriptive title (e.g., "Total Candidates by Gender") * **Chart Type** — Select from Pie Chart, Bar Chart, Line Chart, and others * **X-Axis (Category)** — Select the grouping field (e.g., Is Female) * **Y-Axis (Value)** — Select the metric to measure, with aggregation functions like Count, Sum, or Average * **Y-Series Column Name** — Optionally customize the label shown in the chart legend 4. Save the widget to add it to the dashboard Add Widget dialog Dashboard view with Pie Chart *** ## Querying Smart Tables via Agents One of the most powerful features of Smart Tables is the ability to query structured data using natural language through an **Agent**. ### Setup Process Go to the Agent builder and create or open an existing agent. Add the **Database Manager** toolkit to the agent. In the toolkit settings, select one or more Smart Tables the agent should have access to. This allows the agent to query across multiple related tables simultaneously. Deselect unnecessary options like "Database Query" or "Database Schema" if you only want the agent to access Smart Table data specifically. Open the agent chat and ask questions in natural language, for example: * "How many candidates have more than 5 years of experience?" * "List all Python experts" * "How many rows are there in the Resumes table?" The agent executes a query against the Smart Table and returns the answer in natural language. Agent builder with Database Manager toolkit Database Manager settings showing Smart Table selection Agent query example *** ## Templates Templates accelerate setup by providing pre-configured column structures for common use cases. You can select from existing templates but cannot create custom templates through the UI. ### Evaluation Template Designed for testing Agent performance and validating AI responses. Includes: * **Question Text** — The input prompt for the agent (User Input) * **Expected Answer** — The ground truth or correct answer (User Input) * **Actual Response** — Populated by the agent during the evaluation run (Agent Generated) * **Evaluation Status** — Compares Actual Response to Expected Answer and returns a pass/fail percentage (Comparison) **Using the Evaluation Template:** 1. Create a Smart Table from the Evaluations Template 2. Add rows with your test questions and expected answers 3. Configure the Agent field to connect to your AI agent 4. Execute individual rows or run **Re-compute All** to test all questions 5. Review the Actual Response and Evaluation Status to assess agent performance Evaluations template in action *** ## Best Practices * **Column Descriptions** — Write clear, descriptive text for each column to guide the LLM during extraction. For example, name a column "Years of Experience" and add: *"Total years of professional work experience found in the resume."* * **Data Types** — Use specific types (Number vs. Text) to enable accurate sorting, filtering, and chart creation. Use Number for "Years of Experience" rather than Text. * **Format Selection** — Choose appropriate formats per field type. Use Phone Number format for contact fields; use Currency format for monetary values. * **Collections** — Use Collections for complex lists (Skills, Certifications, Education History) to maintain data integrity and enable deeper querying. * **Testing** — Test with one or two documents first to verify column descriptions are guiding the AI correctly before bulk uploading. * **Quick Upload** — Enable Quick Upload when assigning a Smart Table during upload to extract and populate data immediately. * **Re-apply Extraction** — If you modify column definitions, re-assign the Smart Table to existing documents to re-extract data with the updated configuration. * **Comparison Fields** — Use Comparison for validation scenarios, such as comparing expected vs. actual results in QA workflows. Smart Tables can also trigger and interact with **Workflows (Actions)**, enabling advanced logic such as "If Resume Qualifies, send email." Refer to the Workflow Automation documentation for details. ## Before You Start Use the same credentials as other project APIs — API key / secret or your app's auth headers. The account needs permission to use the project's Knowledge Base features. Without it, these calls return `403`. Use your environment's API host (e.g. `https://ai-api.automationanywhere.com`). Paths below are appended to that host unless your deployment uses a global prefix. *** ### 1. List Existing Tables ```bash Request theme={null} GET /project/{project_id}/data-types ``` Returns a list of all data types for the project. Each item includes an `id` you use as `data_type_id` in subsequent requests. *** ### 2. Create a New Smart Table ```bash Request theme={null} POST /project/{project_id}/data-types ``` #### Body Parameters | Field | Required | Description | | ------------- | -------- | --------------------------------------------------------------------------------------------------------- | | `title` | Yes | Name shown in the UI. | | `description` | Yes | Short description. | | `metadata` | No | Defines initial columns: each key is a column name; each value is a sample used to infer the column type. | ```json Example Request theme={null} { "title": "Support tickets", "description": "Inbound requests", "metadata": { "subject": "Sample subject", "priority": 1, "open": true, "tags": ["billing", "login"] } } ``` ```json Example Response theme={null} { "message": "Data type created successfully", "data_type_id": "abc123...", "view_id": "view-uuid-here" } ``` Save the returned `data_type_id` — you'll need it for all row operations. *** ### 3. Add or Change Columns ```bash Request theme={null} POST /project/{project_id}/data-type/{data_type_id}/column ``` ```json Example Request theme={null} { "column_name": "status", "column_type": "text", "description": "Workflow status", "options": {}, "default_value": "open" } ``` Confirm exact `column_type` values with your OpenAPI or integration docs for your environment. *** ### 4. Add a Single Row ```bash Request theme={null} POST /project/{project_id}/data-type/{data_type_id}/row ``` Column values go inside the `metadata` object. ```json Example Request theme={null} { "metadata": { "subject": "Cannot reset password", "priority": 2, "open": true, "tags": ["auth"] } } ``` ```json Example Response theme={null} { "message": "Row added successfully", "row_id": 5, "row": { "id": 5, "subject": "Cannot reset password", "priority": 2, "open": true, "tags": ["auth"] } } ``` Keys in `metadata` must match existing column names. Do not include `id` — the server assigns it automatically. *** ### 5. Add Many Rows at Once ```bash Request theme={null} POST /project/{project_id}/data-type/{data_type_id}/rows ``` Each row is a plain object of column values — **not** nested under `metadata`. ```json Example Request theme={null} { "rows": [ { "subject": "Issue A", "priority": 1, "open": true }, { "subject": "Issue B", "priority": 3, "open": false } ] } ``` ```json Example Response theme={null} { "message": "Successfully added 2 rows", "rows_added": 2, "row_ids": [6, 7] } ``` *** ### 6. Create a Table from a File ```bash Request theme={null} POST /project/{project_id}/import-table ``` **Format:** `multipart/form-data` with a file upload (CSV or Excel). | Field | Description | | ----------------- | ----------------------------------------------------------------------- | | `title` | Table name shown in the UI. | | `description` | Short description. | | `column_mappings` | JSON string describing how file columns map to table columns and types. | | `file` | The CSV or Excel file to import. | *** ## Quick Reference | Action | Method | Path | | ---------------- | ------ | ------------------------------------------------------- | | List tables | `GET` | `/project/{project_id}/data-types` | | Create table | `POST` | `/project/{project_id}/data-types` | | Add column | `POST` | `/project/{project_id}/data-type/{data_type_id}/column` | | Add one row | `POST` | `/project/{project_id}/data-type/{data_type_id}/row` | | Add many rows | `POST` | `/project/{project_id}/data-type/{data_type_id}/rows` | | Import from file | `POST` | `/project/{project_id}/import-table` | *** ## Troubleshooting The user or API key does not have access to the project or its Knowledge Base section. Verify that the account has the correct permissions enabled. Usually caused by a column name mismatch or a value whose type doesn't match the column definition. Check that all keys correspond to existing column names and that value types match (e.g. don't send a string for a `number` column). The request body structure differs between the two endpoints: * **Single row:** `{ "metadata": { ... } }` * **Bulk rows:** `{ "rows": [ { ... }, { ... } ] }` — no `metadata` wrapper per row.