AI Services Templates
This inventory describes the v2.12.8 source manifests in the ai category. Available means the manifest is not marked Coming Soon; it is not proof of a successful deployment, a current browser card, or a security review. The browser's unfiltered catalog is paginated; use category/search controls to find an entry. See Template System.
Template List
| Template ID | Source availability | Minimum CPU cores | Minimum memory MB |
|---|---|---|---|
anything-llm | Available | 1 | 2048 |
bolt-diy | Available | 2 | 2048 |
comfyui | Coming Soon | 2 | 4096 |
contextforge | Available | 1 | 1024 |
crewai | Coming Soon | 1 | 1024 |
deerflow | Available | 2 | 4096 |
dify | Available | 2 | 6144 |
flowise | Available | 1 | 1024 |
hermes-agent | Available | 1 | 1024 |
ironclaw | Available | 2 | 2048 |
jan | Coming Soon | 2 | 2048 |
langflow | Available | 1 | 2048 |
librechat | Available | 1 | 2048 |
localai | Available | 2 | 4096 |
nanoclaw | Coming Soon | 1 | 512 |
nemoclaw | Coming Soon | 4 | 8192 |
ollama | Available | 2 | 4096 |
open-webui | Available | 1 | 1024 |
openai-compatible-proxy | Available | 1 | 512 |
openclaw | Available | 2 | 2048 |
paperclip | Available | 1 | 1024 |
perplexica | Available | 2 | 2048 |
rag-api | Available | 1 | 2048 |
searxng | Available | 1 | 512 |
stable-diffusion-webui | Coming Soon | 2 | 4096 |
text-generation-webui | Available | 2 | 4096 |
vllm | Available | 4 | 8192 |
whisper | Available | 2 | 4096 |
Before deploying
Select an available card and review its current variables, image/source version, resource limits and application instructions. Coming Soon entries cannot be deployed. Minimum memory is a declared requirement, not a capacity guarantee. Required password inputs and generated system secrets are different: a blank password field does not automatically generate a value.
Use Template Deploy for the deployment steps and Template Authoring for the manifest schema. Container health checks report readiness; they do not by themselves restart a hung application or provide automatic image/security updates.
AI configuration
Choose the source supported by the template's aiInjection mapping, or configure the application itself when no mapping is declared. BYOK, shared Provider AI, Local AI and Metered AI have different storage and billing behavior; see How AI works.
OpenClaw: the shipped template disables Control UI device authentication. Password authentication remains configured. Read the prominent hardening warning before deployment; the catalog's availability marker is not approval of that security default.
Hermes Agent: new deployments install Hermes v2026.9.24 (the hermes-agent row above reflects the current source, newer than the v2.12.8 snapshot). The dashboard has two locks. Panel single sign-on comes first; Hermes then asks for its own sign-in: username admin and the Dashboard Password shown under Credentials on the service page. Messaging platforms are connected and paired on the dashboard's Channels and Pairing pages; no terminal is needed. Hermes follows the AI source chosen at deploy, or the first provider key set (OpenRouter, OpenAI, Anthropic, then Gemini); a model chosen on the dashboard's Models page is kept. A Hermes service deployed before this change keeps v2026.6.5, which has no sign-in of its own, so its Dashboard Password row stays empty. The declared minimum (1 core, 1024 MB) covers the dashboard, gateway and chat; heavy tool use needs more.