n8n Chat Workflow: Route Prompts to Free LLMs (NVIDIA, Ollama)

Route n8n Chat prompts to free LLMs—then return the reply

This n8n workflow automatically takes messages from n8n Chat, sends the prompt to multiple free LLM options (NVIDIA Integrate, Agnes AI, and Ollama), and returns responses back to the chat—plus an optional Ollama Cloud LangChain agent run with memory.

What this workflow does

• Receives a chat message (and optional file uploads) via the n8n Chat trigger.
• Calls NVIDIA Integrate using the /v1/chat/completions endpoint, then returns the first completion to the chat.
• Calls Agnes AI using the same /v1/chat/completions pattern, then returns the first completion to the chat.
• Calls a local Ollama model via the /api/chat endpoint and returns the model’s message content to the chat.
• Runs a LangChain agent with Ollama Cloud: uses an Ollama Cloud chat model plus buffer-window memory, then returns the agent output to the chat.

Use cases

• Multi-model chat testing: compare responses from NVIDIA Integrate, Agnes AI, and your local Ollama setup.
• Cost-control and flexibility: route prompts to local inference (Ollama) or Ollama Cloud while keeping the chat experience consistent in n8n.
• Agentic assistance: use the Ollama Cloud LangChain agent with memory for iterative, context-aware replies.
• SaaS operators: offer users a single n8n Chat entry point while experimenting with free LLM backends.

Technical details

• Trigger/Chat: n8n-nodes-langchainchat trigger
• HTTP calls: HTTP Request steps for NVIDIA Integrate and Agnes AI (/v1/chat/completions)
• Local inference: Ollama chat via /api/chat (optionally update base URL like http://host.docker.internal:11434)
• Agent mode: n8nn8n-nodes-langchainagent using Ollama Cloud chat with buffer-window memory
• Included nodes/components: sticky note, LangChain chat/agent nodes (e.g., n8nn8n-nodes-langchainchat, n8nn8n-nodes-langchainlm chat ollama)

Setup notes: add an NVIDIA API bearer token and an Agnes AI bearer token, install Ollama and pull the referenced model (e.g., gemma4:latest), then configure Ollama Cloud (Ollama API) credentials and ensure gemma4:31b is available.