Dify
The LLM app platform - visual builder, RAG, agents, observability - the catalogue’s heaviest stack, shipped honestly.
One-click deploy, from $49/mo on a Miget plan.
Dify is the most complete LLM app platform in open source: a visual builder for chatbots, agents, and multi-step workflows, with built-in RAG over your documents, model management across providers, annotations, and usage observability. 120k+ GitHub stars say the demand is real.
It is also, honestly, the heaviest template in this catalogue: eleven services mirroring upstream’s compose at the 1.14 line - api/worker/beat, web, the plugin daemon, a code-execution sandbox with its squid egress guard, same-origin nginx routing (which eliminates post-deploy URL variables), plus its own Postgres, noeviction Valkey, and bundled Qdrant. The experimental badge means exactly that: validate on a dev project, bump the version-locked images together.
Two honest notes carried from the license and pricing pages: self-hosting is single-tenant by license (one shared Dify for multiple customer tenants needs upstream permission), and Dify Cloud meters message credits ($59/month for 5,000) - which self-hosting does not.
Upstream project: Dify
#what you get
- Visual builder: chatbots, agents, and multi-step workflows
- Built-in RAG: ingest, chunk, embed, retrieve - Qdrant bundled
- Model management across providers (litellm plugs in as OpenAI-compatible)
- Plugin ecosystem via the bundled daemon
- Sandboxed code execution with SSRF-guarded egress
- Same-origin routing: no post-deploy URL variables
#topology
| Service | Role | Public |
|---|---|---|
| entry | nginx router (:5000) - api + web same-origin | yes |
| api / worker / worker_beat / web / plugin_daemon | the Dify platform (version-locked images) | no |
| sandbox / ssrf_proxy | code execution + egress guard | no |
| db / redis / vectordb | Postgres (2 DBs) / noeviction Valkey / Qdrant | no |
#miget sizing
// this stack needs
6.6 GiB RAM · 27 GB disk · 11 services
Upstream’s floor is 4 GiB; this allocation gives the real shape room. The four langgenius images are version-locked - upgrade them together with release notes open.
Hobby - recommended fit
$49/mo
4 vCPU · 8 GiB · 160 GiB disk
Headroom for your own apps: 16 GiB at $97/mo
Professional - production
$85/mo
4 vCPU · 8 GiB · 50 GiB disk
Dedicated resources, production SLOs - plan details
One Miget plan is a fixed pool of compute - the whole stack (managed databases included) deploys inside it, and anything left over runs your other apps. No per-service or per-seat math.
#vs. the managed service
What the hosted equivalents charge, against the flat Miget plan this stack fits on. Prices as of June 2026, sources linked.
| Service | Plan | Monthly | What you get |
|---|---|---|---|
| Dify on Miget ★ | 8 GiB plan | $49 | this whole stack, flat - no usage meters, and room left for your own apps |
| Dify Cloud | Professional | ~$59 | per workspace: 5,000 message credits/mo, 3 members, 50 apps; Team $159/mo |
Self-hosted Dify has no message-credit meter; LLM calls bill at your providers either way.
#vs. other PaaS
Estimated monthly cost of running this exact stack (6.6 GiB RAM, 27 GB disk, 11 containers) elsewhere, from published June 2026 rates.
| Platform | Est. monthly | Notes |
|---|---|---|
| Miget ★ | $49 flat | compose stacks first-class: one deploy, dedicated vCPU, managed Postgres/Valkey, volumes and TLS all included in the plan |
| Heroku | ~$331 | no volumes; nothing between 1 GB ($50) and 2.5 GB ($250) dynos - 2 GB containers cost far more than shown |
| Render | ~$90 | per-service instances (0.5 GB $7, 2 GB $25) - every container is its own paid service |
| DO App Platform | ~$85 | no persistent volumes - stateful containers need managed DBs/Spaces (base $5 Spaces included here) |
| Railway | ~$70 | usage-based ($10/GB RAM-mo); vCPU billed separately at $20/vCPU-mo on top |
| Fly.io | ~$43 | cheapest sticker price - but burstable shared CPUs (1/16 core; dedicated vCPUs cost ~2-3×), no compose deploys (one app per container, manual wiring), managed DBs billed extra |
Estimates assume RAM fully allocated at published on-demand rates - and sticker price isn't the whole comparison: the cheaper rows buy burstable shared CPUs, per-service wiring instead of a compose deploy, and managed databases billed separately. Heroku and DO App Platform have no persistent volumes at all - stateful stacks like this one need workarounds there.
#deploy it
On Miget
- Create a Compose Stack in app.miget.com pointing at the templates repository
- Set the stack path to
dify -
Set the required variable:
SECRET_KEY + six service secrets, openssl one-liners in .env.example - nothing else to configure
- Deploy. Miget layers
compose.miget.yaml(RAM, privacy, volumes, managed services) automatically
Locally first?
Every template is portable, vanilla Docker Compose - the Miget overrides are ignored locally:
git clone https://github.com/deployable-sh/stacks
cd miget-compose-templates/dify
docker compose up -dSame files, same behavior. The template README covers connection strings and scaling notes.
#faq
Dify, Flowise, or Open WebUI?
Open WebUI (1 container) for team chat over models. Flowise (2 services) for visual LLM flows and embeddable bots. Dify (11 services) when you want the full platform - workflows, RAG pipelines, plugins, multi-app workspaces - and accept the operational weight. The catalogue ships all three so the choice is honest.
What does the license actually restrict?
Dify’s open-source license is Apache-2.0 plus two conditions: keep the branding, and do not operate one Dify as a multi-tenant service for separate customers without permission. Your org self-hosting its own workspace - the thing this template does - is explicitly fine.
How does the cost compare to Dify Cloud?
Dify Cloud Professional is $59/month per workspace with 5,000 message credits and caps on apps and documents. This stack is $49/month flat with no credit metering - the LLM calls themselves bill at your providers either way (cap them via the litellm template).
Why so many services?
Because Dify genuinely is a platform: async ingestion needs workers, plugins need their daemon, untrusted code needs a sandbox with guarded egress, and RAG needs a vector store. The template wires all of it so one deploy works - but the page count is the honest price of the feature count.
Ship Dify today
One compose stack, 6.6 GiB of RAM, from $49/month flat, and it runs on your laptop with the same files.