Pull the public images
Core application, converters and the Dokuta video-to-docs stack are all public on Docker Hub, AMD64 and ARM64. Nothing to request, nothing to sign in to.
Image list on GitHub →Docsie Self-Hosted · Kubernetes preview
Your knowledge.
Your infrastructure.
Deploy Docsie on Kubernetes you control. Public Helm charts, public Docker images for the whole stack, and your choice of local or hosted AI models. Data, storage and inference stay where you put them.
Free for personal use. Free 30-day business evaluation for up to 100 users. Docsie is proprietary software with public deployment tooling: every image is public on Docker Hub. No registry credentials, no access request.
Install in three steps
Helm on your cluster
Clone the repository, build chart dependencies, set your hostname, storage and admin account in a values file, then run the explicit-kubeconfig installer. PostgreSQL, Redis, MinIO and Elasticsearch are included in the platform chart.
Kubernetes install guide →Point it at your AI
Connect Ollama or any OpenAI-compatible endpoint for generation. Choose local, hosted or mixed providers independently for text, vision, embeddings, transcription and speech.
Ollama walkthrough →# v0.2.0-preview.2
git clone https://github.com/LikaloLLC/docsie-self-hosted.git
cd docsie-self-hosted
git checkout v0.2.0-preview.2
helm dependency build charts/docsie-platform --skip-refresh
bash scripts/install-kubernetes.sh /path/to/kubeconfig docsie /path/to/values.yaml
git clone https://github.com/LikaloLLC/docsie-self-hosted.git
cd docsie-self-hosted
git checkout v0.2.0-preview.2
helm dependency build charts/docsie-platform --skip-refresh
bash scripts/install-kubernetes.sh /path/to/kubeconfig docsie /path/to/values.yaml
Installing with an AI coding agent? Point it at AGENTS.md and describe the outcome you want, such as "internal SOPs plus video-to-docs on our existing cluster using our Ollama server". The agent selects the required components, prepares configuration and verifies the result.
What you can install today
| Area | Status | Notes |
|---|---|---|
| Core knowledge base | Preview | Published Kubernetes chart with PostgreSQL, Redis, MinIO, Elasticsearch, web and worker processes, document and image converters. All images public. |
| Local or hosted AI | Configurable | Ollama and OpenAI-compatible endpoints documented. Verify chat, retrieval, vision and embeddings for the models you pick. |
| Dokuta video-to-docs | Public images | UI, API and worker images public on Docker Hub. Needs text and vision models plus a transcription endpoint for narrated recordings. |
| Local transcription | Bring your own | Any Whisper-compatible endpoint works. A bundled Whisper server is not packaged yet. |
| Local TTS and live voice agents | Pending | Chatterbox and LiveKit packaging in progress. Voice APIs are disabled by default. |
| Learn, Forms, translation, comparison, policy review, automation | Application features | Present in the application. Confirm image version and dependencies for the workflow you need. |
| AWS one-click and offline bundle | Not released | AWS scaffolding in the repo is experimental. Air-gapped deployments are assisted by our team for now. |
Read the release status before choosing a deployment. It states exactly what was validated.
Simple access, clear terms
Free for your own projects
One active user, personal non-commercial use. Bring your own machine or cluster and your own AI access. Work for an employer, client or business requires a business license.
30 days. Up to 100 users.
Evaluate Docsie in your environment with your team. Install from the public images; ask us for installation help and evaluation terms. Extensions can be arranged if your evaluation needs more time.
Keep your deployment
Continue with a paid license after your evaluation. A commercial license upgrades the existing installation without moving content. Infrastructure, storage and inference costs remain yours.
Before you deploy
Is Docsie open source?
No. The Helm charts, installer scripts and documentation are public on GitHub. The Docsie application is proprietary, with free personal use, a 30-day business evaluation and commercial licensing. Anonymous readers of a published portal do not count as application seats.
Are all the images public?
Yes. The core Docsie application, the image-processor and PDF-converter images, and the four Dokuta images (UI, API, Celery worker, LiveKit voice worker) all pull from Docker Hub without credentials. The Helm charts reference them by immutable digest.
What infrastructure do we need?
A Kubernetes cluster with a working default StorageClass, Helm 3 and kubectl. The platform chart brings PostgreSQL, Redis, MinIO and Elasticsearch with it. Production images target Linux x86_64. For local models you also need a machine with enough RAM, VRAM and disk for the model you choose.
Can we keep every AI call inside our network?
Yes, if you configure it that way. Point generation at Ollama or another OpenAI-compatible server on your private network and supply your own transcription endpoint. Self-hosting the application does not automatically make every AI call local, so verify each route (text, vision, embeddings, transcription, speech) before calling the setup offline.
Can I run this on a Mac or ARM server?
The Dokuta images ship for linux/arm64 as well as amd64. The documented Mac rehearsal for the full platform uses emulation and locally prepared images. Production architecture is Linux x86_64. Ask us about an assisted local deployment.
Does the preview enforce licensing?
The preview image does not include offline license enforcement. License terms still apply. Request evaluation terms from Docsie rather than assuming an activation command exists.
What happens after the business trial?
Continued business use requires a paid license. You renew the existing installation without moving its content. Evaluation terms are provided with your trial.
Plan your self-hosted evaluation
Tell us where you want Docsie to run, which AI providers you use and what your team wants to evaluate. We help validate your configuration and set evaluation terms.
Related: Self-hosted documentation platform · On-premise knowledge base · Air-gapped documentation · Self-hosted ChatGPT for docs