SoterAIInstall

Summary

Analyze each input text for prompt injection, jailbreaks, and other threats, then route clean results to Safe and risky results to Flagged. Choose cloud, local, or automatic detection; optionally add conversation context, audit metadata, and custom replies.

Inputs

• Input Text (required) — The message to analyze, such as a value mapped from {{$json.chatInput}}.
• Detection Engine — Select Auto (cloud first, with local fallback), Cloud Only, or Local Only. Cloud detection requires a SoterAI credential; local detection runs bundled rules without network access and has reduced coverage.
• Customer Replies — Optional reply text for allowed, blocked, off-topic, prompt injection, redacted, rephrase-needed, or sensitive-data results. Replies are returned in userMessage; the workflow must deliver them.
• Options — Optional settings include request timeout, batch concurrency, session ID, project ID, audit metadata, reuse of identical items, raw API response inclusion, and whether Auto mode may fall back to local detection. Metadata is redacted before sending; raw responses are recursively sanitized.

Output shape

Two routed outputs: Safe for clean items and Flagged for risky items; each carries the analysis result.

Connect both outputs to handle each outcome. Identical items may reuse an API call by default and are marked reusedResult. If n8n Continue On Fail is enabled, failed items leave through Flagged. The operation notice describes the result as a risk report.

Links

• n8n community nodes installation guide