Agent Skills: flat-data-file-import

File Import Tools (Flat Data)

When a user wants to import a file (CSV, Excel, TSV, etc.) into Improvado, use these tools.

Routing rules:

• If the user mentions SFTP → use the Managed SFTP Workflow directly (datasourceName: "managed_sftp")
• If the user mentions S3 or AWS S3 → use the Managed S3 Workflow directly (datasourceName: "managed_s3")
• If the user mentions email import → use the Email Import Workflow (datasourceName: "email_file_import")
• If the user uploads a file or says file import → use the File Import Workflow (datasourceName: "file_import")
• If unclear, ask which method they prefer: file upload, email, SFTP, or S3

Available Tools

1. getConnectionsTool (existing) - Get the connection_id: getConnectionsTool({ datasourceName: "file_import" })
2. uploadFlatDataFileTool - Upload the file to DTS managed storage
   • ALWAYS pass the attachment URL as fileUrl — do NOT base64-encode file content
   • Handles multipart upload and antivirus scanning automatically
   • Returns file_path (used in all subsequent calls), file_name, and file_size
3. discoverFileFieldsTool - Discover field definitions and data sample from the uploaded file
   • Pass the connection_id from step 1 and file_path from step 2
   • Returns inferred field names, types, and data_sample (first rows of data)
4. generateFlatDataSampleTool - Generate a data preview with confirmed field definitions
   • Use after the user confirms or adjusts the discovered schema
5. createFlatDataExtractTool - Create the extract with the final configuration
   • Creates a persistent extract that stores the imported data
   • data_table_title: Propose a name based on the file name (e.g. "report.csv" → "report_csv_data"). Ask the user to confirm.
6. uploadNewFileToExtractTool - Upload a new file to an existing extract/data table
   • Use when the user wants to add/replace data in an existing data table
   • Requires: extract_id + file details from uploadFlatDataFileTool
   • Backend automatically triggers data processing
7. previewEmailImportAddressTool - Preview the email address for email-based file import
   • Shows what email address will be generated for a given data_table_title
   • The actual email is activated after the extract is created
8. createConnectionTool (existing) - Create a connection if none exists
   • For managed SFTP: createConnectionTool({ data_source_name: "managed_sftp", auth_id: "login_password", params: { connection_name: "<name>" } })
   • For managed S3: createConnectionTool({ data_source_name: "managed_s3", auth_id: "login_password", params: { connection_name: "<name>" } })
9. getManagedConnectionCredentialsTool - Fetch credentials for a managed SFTP or S3 connection
   • Returns: host, port, login, password
   • Use after getting/creating a managed_sftp or managed_s3 connection to show credentials to the user

File Import Workflow

1. Call getConnectionsTool({ datasourceName: "file_import" }) to get the connection_id

2. Call uploadFlatDataFileTool with fileUrl (from chat attachment) → get filepath

3. Call discoverFileFieldsTool with connection_id + filepath → get fields + data_sample

4. Analyze the data_sample and discovered fields:

   • Detect the file format and encoding
   • Identify if any top rows should be skipped (header offset)
   • For each field, determine: Type (Text, Number, DateKey), Is key (True/False - Fields used to merge new & existing data), and any skip rules

5. Call generateFlatDataSampleTool with the fields for a data preview

6. Important: Present the proposed schema to the user as a table with columns: Field Name, Type, Is key (dimension or property/metric), Sample Values. Propose a Data Table name based on the file name and Write Policy, then ask for confirmation.

   Example:

   | Field Name     | Type   | Is Key | Sample Values          |
   |----------------|--------|--------|------------------------|
   | date           | Date   | Yes    | 2024-01-15, 2024-01-16 |
   | campaign_name  | Text   | Yes    | Brand Campaign, Promo  |
   | impressions    | Number | No     | 15234, 8921            |
   | spend          | Number | No     | 150.50, 89.20          |
   
   Data Table Name: facebook_ads_performance
   Write Policy: Upsert (whole data)
   

7. After the user confirms (or edits), call createFlatDataExtractTool with the confirmed configuration

8. After extract creation succeeds, provide the user with a link to the created extract:

   | Resource | Link                                                                                          |
   |----------|-----------------------------------------------------------------------------------------------|
   | Extract  | {platform_host}/info_connector/overview/{connection_id}/{extract_id}?workspace={workspace_id}  |
   

   Example: https://report.improvado.io/info_connector/overview/19417/1785798?workspace=12345

   Use connection_id from step 1, extract_id from the createFlatDataExtractTool response, and the current workspace ID.

Upload to Existing Data Table Workflow

1. Upload file with uploadFlatDataFileTool → get file_path, file_name, file_size
2. Call uploadNewFileToExtractTool with the extract_id and file details

Email Import Workflow

The same File Import Workflow applies, except:

• Use getConnectionsTool({ datasourceName: "email_file_import" }) instead of "file_import"
• After proposing a data_table_title, call previewEmailImportAddressTool to show the user the email address
• After createFlatDataExtractTool succeeds, the email becomes active — users can send files to it

What is Managed SFTP / Managed S3?

Managed SFTP and Managed S3 are Improvado-hosted file ingestion endpoints. Improvado provisions and owns the SFTP/S3 server — the user does NOT provide their own server.

Flow: Improvado creates a connection → provides credentials (host, port, login, password) → the user uploads files to Improvado's server using those credentials → Improvado automatically ingests the data.

NEVER describe this as "connecting to the user's SFTP/S3 server" — it is the opposite: the user connects to our server.

Managed SFTP Workflow (step by step — follow this order strictly)

IMPORTANT: Do NOT ask for data table name, file columns, or schema upfront. Follow these steps IN ORDER:

Step 1 — Connection

1. Call getConnectionsTool({ datasourceName: "managed_sftp" }) to check for existing connections.
2. If a connection exists, use it. If not, ask the user for a connection name, then create one: createConnectionTool({ data_source_name: "managed_sftp", auth_id: "login_password", params: { connection_name: "<name>" } })

Step 2 — Sample file

Ask the user to upload a sample file (CSV/Excel/TSV) via chat attachment. This is required to discover the schema.

Step 3 — File Import Workflow

Follow the standard File Import Workflow (upload → discover fields → confirm schema → create extract). Use the connection_id from Step 1.

Step 4 — Show credentials

After createFlatDataExtractTool succeeds:

1. Call getManagedConnectionCredentialsTool({ data_source_name: "managed_sftp", connection_id }) to fetch SFTP credentials

2. Show the credentials and folder name to the user. The folder name is the snake_case version of the data_table_title (e.g. "My Sales Report" → my_sales_report).

3. Password display: mask the password with asterisks and wrap it in a <details> tag so the user can click to reveal: <details><summary>••••••••</summary>actual_password</details>

   Example:

   **SFTP Connection Credentials (Improvado-hosted):**
   | Field       | Value                          |
   |-------------|--------------------------------|
   | Host        | ftp.tools.improvado.io         |
   | Port        | 2022                           |
   | Login       | managed_1_26_my_data           |
   | Password    | <details><summary>••••••••</summary>ysUXysurE9TGhSvTAc4k</details> |
   | Folder      | my_sales_report                |
   
   | Resource | Link                                                                                          |
   |----------|-----------------------------------------------------------------------------------------------|
   | Extract  | {platform_host}/info_connector/overview/{connection_id}/{extract_id}?workspace={workspace_id}  |
   

   Upload your files to the folder above via SFTP using these credentials. Improvado will automatically process incoming files.

Managed S3 Workflow (step by step — follow this order strictly)

Same steps as Managed SFTP Workflow above, except:

• Use getConnectionsTool({ datasourceName: "managed_s3" }) instead of "managed_sftp"
• Use createConnectionTool({ data_source_name: "managed_s3", auth_id: "login_password", params: { connection_name: "<name>" } }) if no connection exists
• Use getManagedConnectionCredentialsTool({ data_source_name: "managed_s3", connection_id }) to fetch S3 credentials
• S3 credentials differ from SFTP — expect fields like bucket, access key, and secret key instead of host/port/username/password. Display whatever fields the API returns.

Important Rules

• Write Policy: The default write policy is { method: "upsert", scope: "whole_data" } which replaces all data on each file upload.
• NEVER try to base64-encode or stream file content. Always use the fileUrl from the chat attachment.
• ALWAYS ask for confirmation before creating the extract — present the schema table and proposed name first.