Portainer Templates

Cognee - the memory engine for AI agents: ingest documents, chats, and data; cognify them into a knowledge graph + vector index; query the result as long-term memory.

Dify - the LLM application platform: a visual builder for chatbots, agents, and workflows, with built-in RAG, model management, and observability.

Flowise - the visual builder for LLM apps: drag-and-drop chatflows and agentflows over 100+ integrations (models, vector stores, tools), with APIs and embeddable chat widgets. One app container + a managed Postgres.

Langfuse v3 - open-source LLM engineering platform: tracing, evaluations, prompt management, and usage dashboards. This template mirrors the official self-host topology.

Letta (the MemGPT lineage) - an agent server with memory as the core primitive: agents live server-side with self-editing persistent memory, and clients (SDKs, the hosted ADE at app.letta.com) connect to them over the REST API.

LiteLLM proxy - one OpenAI-compatible endpoint in front of every LLM provider (Anthropic, OpenAI, Gemini, Bedrock, Mistral, local, …) with virtual keys, per-key budgets, spend tracking, fallbacks and retries.

Ollama - run open large language models locally behind a simple HTTP API. Pull a model (Llama, Mistral, Qwen, Gemma, Phi, and many more) and serve it from your own infrastructure, with OpenAI-compatible endpoints.

Open WebUI - the self-hosted AI chat workspace: ChatGPT-style UI over any OpenAI-compatible API, with RAG over uploaded documents, multi-user accounts, RBAC, and prompt/model presets. Single container + volume.

Phoenix - LLM tracing, evaluations, datasets, and a prompt playground in ONE container, OpenTelemetry-native.

TEI - Hugging Face's Rust embeddings server: small models like BAAI/bge-small-en-v1.5 (384-dim) serve real RAG workloads from CPU in ~512 MB, over both TEI's native API and an OpenAI-compatible /v1/embeddings endpoint.