Dify capabilities that change how a self-hosted deployment is designed and governed: human review steps, knowledge pipelines, moderation hooks, triggers, version control and telemetry.
Putting a person in the loop of a Dify workflow: pausing at a Human Input node, delivering forms through the web app or email, decision buttons that route the workflow, timeouts, and where this fits in review processes that need recorded human oversight.
Building custom ingestion for Dify knowledge bases as a visual pipeline from data source to extraction, processing and storage, choosing between built-in templates and blank pipelines, and when a pipeline beats the ready-to-use knowledge base.
Screening what users send to Dify apps and what models return by wiring the moderation extension points to your own moderation or data loss prevention service, including how streamed output is chunked and what the endpoint must return.
Starting Dify workflows from events in other systems: webhook triggers with method, content type and parameter extraction, integration triggers from plugins, schedule triggers, the TRIGGER_URL setting a self-hosted instance needs, and testing without polluting production data.
Managing change in Dify chatflow and workflow apps: drafts versus the live version, named versions and release notes, restoring an old version for rollback, and combining this with DSL exports for promotion between environments.
Instrumenting a self-hosted Dify for operations teams: enabling OTLP export to your collector, sampling rates, per-service Sentry DSNs for the API, web and plugin daemon, DEPLOY_ENV tagging, and workflow log retention settings.
Using Dify's annotation reply to return curated answers for known questions instead of generating them: similarity threshold, embedding model choice, bulk import of Q and A pairs, hit tracking, and where approved answers help in regulated support use.
What Dify records for every web app, API and trigger run, how operators review and rate answers, and how to turn on scheduled cleanup of workflow logs and messages so retention matches your privacy policy.
The four built-in Community roles (Owner, Admin, Editor, Normal), invitation links and their expiry, ownership transfer, knowledge base visibility, and when custom roles push a team toward Dify Enterprise.
Moving Dify apps between workspaces and instances as YAML DSL: what an export carries and leaves out (API keys, knowledge content, logs), recreating MCP server identifiers, and keeping DSL files under version control.
What the Dify Agent app adds to a self-hosted deployment: the Linux sandbox that runs code and shell commands, the agent backend and agent SSRF proxy services, required secrets, and the isolation decisions operators must make before enabling it.
Tell us what you're working on, whether it's a deployment, an audit, a security test or a cyber range. You'll speak with an engineer who can help you scope it.