Self-Hosted AI Add-On for Jitbit Helpdesk

Every AI feature from the hosted Jitbit Helpdesk is now available on the self-hosted edition. Semantic knowledge base search, similar-article suggestions inside tickets, AI-drafted replies grounded in your own content, and external documentation indexing — all running on your own hardware, with your data under your control.

These features used to be hosted-only because they depended on a Python microservice running a vector database and embedding models. That microservice now ships as a container stack you can stand up yourself — no new platform, no new vendor, no cloud dependency. You run it next to your existing Helpdesk install.

Requires Jitbit Helpdesk v11.22 or newer. Older versions don't expose the "Self-hosted AI service" settings panel in the admin. If you're on an earlier version, grab the latest release from the version history page first.

What you get

• Semantic KB search for end-users. Knowledge base search results ranked by meaning, not keyword match. A user searching for "my password won't work" finds the "Reset login credentials" article even if none of those words overlap.
• Similar-article suggestions in tickets. When agents open a ticket, Helpdesk surfaces the most relevant KB articles based on semantic similarity to the ticket content.
• AI-generated reply drafts. Agents click a button and get a draft response grounded in your own KB, your writing-style rules, and the ticket context. Uses OpenAI, Azure OpenAI, or AWS Bedrock with your API key.
• External documentation indexing. Crawl your own public docs, wiki, or any internal site so AI replies can pull from your product documentation, not just your support KB.

Every feature above matches our hosted AI helpdesk. The only difference is where the data lives.

How it's packaged

A Docker Compose stack: one container for the AI service, one for Qdrant, the vector database. Runs on Windows or Linux, Intel/AMD or ARM. Point Helpdesk at it from Administration → AI features and it kicks off a background reindex of your KB automatically.

Two options for embeddings, with a real tradeoff:

• Bundled local model — free, no API key, no external calls. Strictest data isolation: nothing about your tickets or KB ever leaves your network. Needs a reasonably capable CPU on the AI host.
• OpenAI embeddings — higher search quality and no CPU load on your side (this is what we usually recommend), but text to be embedded is sent to OpenAI's API. The vector database and all stored content still live on your hardware.

You can switch between them any time — the add-on will reindex automatically.

Full setup instructions live in the setup guide.

Licensing and pricing

Why it matters

For regulated organizations — healthcare, government, finance, defense — sending customer tickets and knowledge base content to a third-party AI API has been a non-starter. The self-hosted AI add-on removes that objection:

• With the local embedding model, nothing about your customers, KB, or indexed documentation ever leaves your network.
• If you use a generative provider for reply drafts, you bring your own key. Azure OpenAI and AWS Bedrock keep your data inside your existing cloud tenancy, with a BAA if you need one.
• It runs next to your existing SQL Server and IIS/Kestrel setup. No new platform, no new vendor.

If you already run Jitbit Helpdesk self-hosted, the AI add-on is the fastest way to bring RAG-based support tooling to your team without giving up the on-prem model.