Skip to main content

Speech Insights SFTP

Speech Insights SFTP is the WinnerWare ingestion-source module for teams that receive call recordings through SFTP.

Overview​

This module is used when call recordings must be collected through a more locked-down file-transfer path. It extends Speech Insights by adding an SFTP connection option inside the Ingestion Sources area.

Key Features​

  • Adds SFTP as an available Speech Insights ingestion-source type.
  • Supports both password-based and key-file-based login.
  • Supports a host key fingerprint for safer server validation.
  • Supports server-type and proxy-bypass options.
  • Stores sensitive credentials and keys securely.

How It Works​

Add an SFTP source to Speech Insights​

After the module is enabled, admins can create a new ingestion source and choose SFTP as the source type. That source then becomes part of the standard Speech Insights ingestion workflow.

Choose the login approach that fits the environment​

The editor supports either a traditional username-and-password login or a key-file-based login. This makes the module flexible enough for environments with stricter security requirements.

Keep source mapping aligned with the incoming data​

Like other Speech Insights source types, the SFTP source depends on the shared source field map so WinnerWare can correctly interpret the recording file name, call date, and other incoming call metadata.

Configuration​

Enable the source module​

Turn on this module after the main Speech Insights feature is enabled.

Common source fields​

Every SFTP source also uses the shared Speech Insights source fields:

  • Name
  • Client
  • Locale
  • Candidate locales (optional, for calls in more than one language)
  • Active?
  • the Metadata File Map for recording and call metadata

See Speech Insights for what each field does.

SFTP connection settings​

The SFTP editor adds these connection fields:

SettingPurpose
SFTP hostServer name or IP address.
SFTP portConnection port. The default is 22.
Login typeChoose Normal or Key file authentication.
UsernameAccount name for the SFTP connection.
PasswordUsed when Normal login is selected. After you save, the password is stored securely. Enter a new value only to replace it.
Private key (PEM/OpenSSH)Used when Key file login is selected.
Private key passphraseOptional passphrase for the private key.
Root pathLimits ingestion to a specific server folder.
Server typeLets the connection adapt to the target server: Auto, Unix, or Windows.
Bypass proxyControls whether the connection should skip a proxy path.
Ignore certificate errorsAllows the connection to continue without enforcing host validation.
Host key fingerprintCaptures the expected server fingerprint when validation is in use.

Usage​

Use SFTP when security policy requires it​

This module is a strong fit when recordings are delivered through an SFTP-controlled location and the environment requires stronger server validation or key-based authentication.

Check the source before you activate it​

On Speech Insights -> Ingestion Sources, open the Actions menu of the source:

  1. Click Test connection. WinnerWare connects with the saved settings and lists the root folder. The result shows on the source row. A failure usually means a wrong host, port, credential, key, host key fingerprint, or Root path.
  2. Click Get statistics to see how many audio and metadata files are waiting and how many are Ready to ingest.
  3. Turn on Active? and click Ingest recordings, or wait for the hourly run.

To set up a similar server, use Clone. The copy keeps the connection settings and credentials (including the private key) and starts as inactive.

What happens when something fails​

  • If the server cannot be reached during a run, WinnerWare logs the error, but the source list can show a normal Processed time with 0 recordings ingested. If a source ingests nothing when you expect recordings, click Test connection.
  • An unexpected error that stops a run shows as Last error on the source until a later run completes.
  • A single recording that fails is left on the server and tried again on the next run.
  • A recording that cannot be decoded is moved to a failed folder next to it on the server and is not tried again.
  • Files changed in the last 2 minutes are left for the next run, so files that are still uploading are not read.
  • Recordings are downloaded to temporary disk storage instead of memory, so large recordings do not exhaust server memory.

Operational Notes​

  • The module registers SFTP as a Speech Insights recording provider.
  • Passwords, private keys, and key passphrases are protected before they are stored.
  • The editor supports both normal login and key-file login, and it can hide or show the relevant fields based on the selected authentication mode.