How to run Langflow with a self-hosted model on a cloud GPU

Q: How do the docs point Langflow at a local or OpenAI-compatible server?

The Ollama bundle page says the Base URL defaults to http://127.0.0.1:11434. Its text-generation example is llama3.2:latest. The OpenAI Compatible page says to set Base URL, and the local section prints http://localhost:8000/v1 after export LANGFLOW_SSRF_ALLOWED_HOSTS=localhost. It discovers models from /v1/models. That page does not name vLLM. The agent tutorial says LANGFLOW_SERVER_ADDRESS defaults to 127.0.0.1:7860. The files do not print an SSH command.

Rent a GPU on Deploy, then open a terminal in JupyterLab or over SSH. Install Langflow with the uv command the v1.13.2 README prints, run uv run langflow run, and point the Ollama component or the OpenAI Compatible provider at a server on that machine. Export the flow before you destroy the instance, because destroying the instance deletes its disk.

An agent can rent the GPU and run this application on it. The agent pages are GPU compute for AI agents and the agent integration guide. Those pages are how an agent discovers a GPU and runs work on it. This page is the Langflow install. A Docker Compose app whose self-host page opens http://localhost/install is How to run Dify with a self-hosted model on a cloud GPU. A canvas whose ChatOllama page names a base URL is How to run Flowise with a self-hosted model on a cloud GPU. A CLI whose connect help sets --base-url to http://localhost:8000/v1 for openai-compatible is How to run Letta agents with a self-hosted model on a cloud GPU. A browser chat whose serve command listens on port 8080 is How to run Open WebUI on a cloud GPU. A Python agent library whose CodeAgent can take TransformersModel or an OpenAI-compatible server is How to run smolagents with a self-hosted model on a cloud GPU. A crew whose LLM class takes base_url for Ollama or an OpenAI-compatible server is How to run CrewAI with a self-hosted model on a cloud GPU. A graph runtime whose overview example does not name a model is How to run LangGraph agents with a self-hosted model on a cloud GPU. An AgentChat library whose Ollama example sets model to llama3.2 is How to run AutoGen agents with a self-hosted model on a cloud GPU. A typed agent whose vLLM example sets base_url to http://localhost:8000/v1 is How to run Pydantic AI agents with a self-hosted model on a cloud GPU. A framework whose README sets Ollama and HuggingFaceEmbedding is How to build LlamaIndex agents with a self-hosted model on a cloud GPU. A browser agent whose models page sets base_url to http://localhost:11434/v1 is How to run browser-use with a self-hosted model on a cloud GPU. A pipeline whose OllamaChatGenerator sets url to http://localhost:11434 is How to run Haystack agents with a self-hosted model on a cloud GPU. A program whose language-models page sets dspy.LM to ollama_chat/llama3.2:1b is How to run DSPy with a self-hosted model on a cloud GPU. An agent whose Ollama example sets id to llama3.1:8b is How to run Agno agents with a self-hosted model on a cloud GPU. A TypeScript agent whose local-models section sets url on the model object is How to run Mastra agents with a self-hosted model on a cloud GPU. A ChatAgent whose vLLM example sets ModelPlatformType.VLLM is How to run CAMEL-AI agents with a self-hosted model on a cloud GPU. A general vLLM install is How to serve an OpenAI-compatible API with vLLM on a cloud GPU. A general SGLang install is How to run SGLang on a cloud GPU. Ollama’s install and port 11434 are How to run Ollama on a cloud GPU. An embedding server is How to serve embeddings and rerankers with Text Embeddings Inference on a cloud GPU. CloudGPU’s hosted gateway is a different base URL: How do I use an OpenAI-compatible API for open models?. How published weight sizes are read: How much VRAM do I need to run an LLM?. The tunnel is How to use SSH port forwarding on a cloud GPU. If the process runs out of memory: How to fix CUDA out of memory on a cloud GPU. If the process raises no kernel image on an RTX 5090: How to fix CUDA no kernel image on an RTX 5090.

The commands below are copied from the upstream files and docs pages named in each step. They were not run on a CloudGPU instance for this page. Anything this page adds around them is marked. The install page and the model page name vendors in example replies. This page does not copy those names.

What the docs print about memory

The v1.13.2 README prints no GPU GB figure. The install page prints “Minimum: Dual-core CPU and 2 GB RAM” and “Recommended: Multi-core CPU and at least 4 GB RAM.” Those lines are system RAM. They are not vram_gb. The Ollama bundle page says large models require significant system memory, and it quotes an error that the model requires more system memory than is available. That page prints no GB figure. The model information table on unsloth/Llama-3.1-8B-Instruct prints parameter counts 8B, 70B, and 405B, and a context length of 128k. This page does not treat those parameter counts as vram_gb. A later section of that card names training hardware. This page does not copy that section and does not use it as an inference memory figure. This page does not mark a catalog row as enough.

GET /v1/public/compute-capabilities does not list GPUs. It returns feature flags. The memory column is vram_gb from GET /v1/public/gpu-catalog (generated_at 2026-11-14T09:10:32.863Z). The catalog response also includes an availability field. This page does not copy it. Every catalog row below has min_gpu_count 1. Which listing you can start is on Deploy.

The catalog template list had no Langflow template. It did have Ollama templates (jupyterlab-ollama, ollama-qwen3, ollama). Those templates are a different install. This page does not treat them as Langflow.

Catalog name	vram_gb
RTX 2080	8
RTX 3080	10
RTX 2080 Ti	11
RTX 3080 Ti	12
V100 16G	16
A30	24
RTX 3090	24
RTX 4090	24
RTX 4090D	24
RTX 4080 SUPER 32G	32
RTX 5090	32
RTX 5090D	32
V100 32G	32
A100 40G	40
A800 40G	40
A40	48
RTX 4090 48G	48
RTX 4090D 48G	48
RTX A6000	48
Ascend 910B	64
NVIDIA RTX PRO 5000 72G	72
RTX PRO 5000 72G	72
A100 80G	80
A800 80G	80
H800	80
RTX 6000D	83
H20 96G	96
RTX PRO 6000 96G	96
RTX PRO 6000 Max-Q	96
H20-3e 141G	141

Install

Tag v1.13.2 is commit 8b42952d734de18e34c2838ff6c634efb1815878, committed 2026-11-10T21:30:51Z, published 2026-11-10T22:26:00Z. Main on 2026-11-14 was commit 260cbf347cc5563cc48c79094c175ca31532ea2b, committed 2026-11-13T19:37:04Z. pushed_at was 2026-11-14T06:23:46Z. Those commits are not the same. The README quickstart lines in this section are on the tag v1.13.2 README. The same lines are on main. This page did not diff the rest of the two READMEs. The GitHub license field was MIT. PyPI langflow info.version on 2026-11-14 was 1.13.2, requires_python <3.15,>=3.10, license_expression MIT, uploaded 2026-11-10T22:25:46Z. info.license was empty. The README says Python 3.10–3.14. The install page says Python version 3.10 to 3.14. Those strings match the PyPI range. This page did not install the package.

The install page prints a virtual environment before the package install. For Linux or macOS it prints these two lines. VENV_NAME is the placeholder on that page. Copied:

uv venv VENV_NAME
source VENV_NAME/bin/activate

The same page then prints uv pip install langflow inside that environment, and uv run langflow run. It also prints a Windows activation line, VENV_NAME\Scripts\activate. This page copies the Linux line. The tag README does not print uv venv. Its install and run lines are the ones to run after that environment is active. Copied:

uv pip install langflow -U

uv run langflow run

That README says Langflow starts at http://127.0.0.1:7860. It also says official release packages and tagged source archives disable automatic login by default, to set LANGFLOW_SUPERUSER_PASSWORD before starting, and to sign in as langflow (or LANGFLOW_SUPERUSER). It says development branches retain automatic login by default. Main is not the tag.

The install page’s first install line is uv pip install langflow without -U. The README’s first install line includes -U. Those commands are not the same. Run one of them after source VENV_NAME/bin/activate. The install page’s upgrade line is uv pip install langflow -U. Its pin example is uv pip install langflow==1.4.22. That pin example and PyPI 1.13.2 are not the same version. The page also prints uv pip install langflow-base for the UI and API without provider extensions. The page confirms the URL http://127.0.0.1:7860.

The tag README Docker lines, copied:

read -rsp "Langflow administrator password: " LANGFLOW_SUPERUSER_PASSWORD; echo
export LANGFLOW_SUPERUSER_PASSWORD
docker run -p 7860:7860 -e LANGFLOW_SUPERUSER_PASSWORD langflowai/langflow:latest

That README says Langflow is available at http://localhost:7860/ and to sign in as langflow. The install page’s Docker lines are different. Copied:

docker run -p 7860:7860 \
  -e LANGFLOW_AUTO_LOGIN=false \
  -e LANGFLOW_SUPERUSER_PASSWORD=SUPERUSER_PASSWORD \
  langflowai/langflow:latest

That page says to replace SUPERUSER_PASSWORD with a password, then open http://localhost:7860/. The two Docker commands are not the same. The files this page read do not print an SSH command. For a server you start on the instance, the tunnel is How to use SSH port forwarding on a cloud GPU.

Agent flow

The agent tutorial, fetched 2026-11-14, says to click New Flow, then select the Simple Agent template. It says to remove the URL and Calculator tools, add Read File and Web Search, enable Tool Mode, and connect their Toolset ports to the Agent component’s Tools port. The Agent step tells you to enter a vendor API key. This page does not copy that vendor name. The same sentence says that for a different provider you edit Model Provider, Model Name, and API Key. The tutorial says LANGFLOW_SERVER_ADDRESS defaults to 127.0.0.1:7860, from the flow’s API access pane. The example response on that page names sites this page does not copy.

The README says you can deploy a flow as an API or export it as JSON.

Ollama and the OpenAI Compatible provider

The Ollama bundle page, fetched 2026-11-14, says the Ollama text generation component’s Base URL defaults to http://127.0.0.1:11434, and that this value is the OLLAMA_HOST environment variable in Ollama. Its model example is llama3.2:latest. The embeddings component uses the same default base URL. Its model example is all-minilm:latest. The page says uv pip install langflow already includes the bundle, and that a separate lfx install uses uv pip install lfx-ollama.

The OpenAI Compatible page says to point the model provider at any OpenAI-compatible endpoint, and that Langflow discovers models from /v1/models. The fields are Base URL (OPENAI_COMPATIBLE_BASE_URL, required) and API Key (OPENAI_COMPATIBLE_API_KEY, optional). The Base URL description includes a remote example. This page does not copy that URL. The local section prints http://localhost:8000/v1 and says loopback addresses are blocked unless you allow them:

export LANGFLOW_SSRF_ALLOWED_HOSTS=localhost

That page also says you can allowlist 127.0.0.1 or 10.0.0.0/8. After save, select OpenAI Compatible and a discovered model in a Language Model, Embedding Model, or Agent field. The page does not name vLLM. How to serve an OpenAI-compatible API with vLLM on a cloud GPU is the general server install. How to run Ollama on a cloud GPU is the Ollama install.

The Ollama bundle page does not print hf download. The block below is added on this page. How to download a Hugging Face model prints pip install -U "huggingface_hub" and hf download. This page did not download the repository. The Langflow files this page read do not print this repository id.

pip install -U "huggingface_hub"
hf download unsloth/Llama-3.1-8B-Instruct

On 2026-11-14, GET https://huggingface.co/api/models/unsloth/Llama-3.1-8B-Instruct reported gated false, cardData.license llama3.1, pipeline_tag text-generation, lastModified 2025-05-09T19:32:57.000Z, and sha 4699cc75b550f9c6f3173fb80f4703b62d946aa5. The card YAML also includes extra_gated_prompt. Those two fields are not the same fact. This page did not run hf download.

When it fails

CUDA out of memory. The Langflow files and docs pages this page read print no GPU GB figure. If the model server raises CUDA out of memory, use how to fix CUDA out of memory on a cloud GPU. This page does not pick a smaller model id for you.

“CUDA error: no kernel image is available for execution on the device” on an RTX 5090. The Langflow README does not name compute capability 12.0. The RTX 5090 guide says that card is compute capability 12.0 (sm_120) and that PyTorch 2.7 and later, on a build for CUDA 12.8 or newer, is the line that announced support. This page did not install a torch wheel and did not re-test this program on an RTX 5090. Catalog RTX 5090 and RTX 5090D are 32. That is the catalog number only.

What to copy off, then destroy

The README says a flow can be exported as JSON. Copy that export before you destroy the instance. Save your work before you destroy a cloud GPU is the checklist. The terms say data on local instance storage is permanently deleted when the instance is destroyed.

The billing page states the billing rule in these words:

Per-minute options have a 1-minute minimum; unused prepaid minutes are refunded when you destroy the instance. Some configurations use whole-hour billing with a 1-hour minimum; unused time within a charged hour is not refunded on voluntary stop. The billing period, minimum charge and initial prepayment are shown before confirmation. No subscription.

Amounts are not copied here.

Frequently asked questions

What memory figure do the Langflow docs print?

The v1.13.2 README prints no GPU GB figure. The install page prints 2 GB RAM minimum and at least 4 GB RAM recommended. The Ollama bundle page prints no GB figure. The unsloth/Llama-3.1-8B-Instruct model information table prints parameter counts 8B, 70B, and 405B, and context length 128k. This page does not treat those figures as vram_gb. This page does not mark a catalog row as enough. Which listing you can start is on https://cloudgpu.app/deploy.

Which install lines do the docs print?

The install page prints uv venv VENV_NAME and source VENV_NAME/bin/activate before uv pip install langflow. VENV_NAME is the placeholder on that page. The v1.13.2 README does not print those two lines. It prints uv pip install langflow -U, then uv run langflow run, and says Langflow starts at http://127.0.0.1:7860. Run that install after the environment is active. It says Python 3.10 to 3.14. PyPI langflow on 2026-11-14 was 1.13.2, requires_python >=3.10 and <3.15. Tag v1.13.2 is commit 8b42952d734de18e34c2838ff6c634efb1815878. Main is a later commit. The install page’s first install line omits -U. Its pin example is uv pip install langflow==1.4.22. That pin and 1.13.2 are not the same version. The README Docker command and the install page Docker command are not the same. This page did not install the package.

How do the docs point Langflow at a local or OpenAI-compatible server?

The Ollama bundle page says the Base URL defaults to http://127.0.0.1:11434. Its text-generation example is llama3.2:latest. The OpenAI Compatible page says to set Base URL, and the local section prints http://localhost:8000/v1 after export LANGFLOW_SSRF_ALLOWED_HOSTS=localhost. It discovers models from /v1/models. That page does not name vLLM. The agent tutorial says LANGFLOW_SERVER_ADDRESS defaults to 127.0.0.1:7860. The files do not print an SSH command.

What license does the file print?

The LICENSE file is the MIT License, and its copyright line is Copyright © 2024 Langflow. The GitHub license field was MIT. PyPI license_expression for langflow 1.13.2 was MIT. info.license was empty. That license text is not a CloudGPU rate.

Sources (read 2026-11-14, Asia/Shanghai)

langflow-ai/langflow tag v1.13.2, commit 8b42952d734de18e34c2838ff6c634efb1815878, committed 2026-11-10T21:30:51Z, published 2026-11-10T22:26:00Z. README quickstart: uv pip install langflow -U, uv run langflow run, http://127.0.0.1:7860, Python 3.10–3.14, Docker read/export then docker run -p 7860:7860 -e LANGFLOW_SUPERUSER_PASSWORD langflowai/langflow:latest, sign in as langflow. Main commit 260cbf347cc5563cc48c79094c175ca31532ea2b, committed 2026-11-13T19:37:04Z. The quickstart lines match. This page did not diff the rest. GitHub license field MIT. LICENSE copyright Copyright © 2024 Langflow
PyPI langflow 1.13.2, uploaded 2026-11-10T22:25:46Z, requires_python <3.15,>=3.10, license_expression MIT, info.license empty
Install page https://docs.langflow.org/get-started-installation, fetched 2026-11-14. Python 3.10 to 3.14. 2 GB RAM minimum. 4 GB RAM recommended. Linux lines uv venv VENV_NAME and source VENV_NAME/bin/activate before uv pip install langflow. First install omits -U. Pin example uv pip install langflow==1.4.22. Docker command sets LANGFLOW_AUTO_LOGIN=false and LANGFLOW_SUPERUSER_PASSWORD=SUPERUSER_PASSWORD. http://localhost:7860/ and http://127.0.0.1:7860
Ollama bundle https://docs.langflow.org/bundles-ollama, fetched 2026-11-14. Default base URL http://127.0.0.1:11434. Example llama3.2:latest. Embeddings example all-minilm:latest. No GB figure
OpenAI Compatible https://docs.langflow.org/bundles-openai-compatible, fetched 2026-11-14. OPENAI_COMPATIBLE_BASE_URL, OPENAI_COMPATIBLE_API_KEY, /v1/models, local example http://localhost:8000/v1, export LANGFLOW_SSRF_ALLOWED_HOSTS=localhost. The remote example URL is not copied. The page does not name vLLM
Agent tutorial https://docs.langflow.org/agent-tutorial, fetched 2026-11-14. New Flow, Simple Agent template. LANGFLOW_SERVER_ADDRESS default 127.0.0.1:7860. The vendor key sentence and the example response URLs are not copied
Hugging Face model API unsloth/Llama-3.1-8B-Instruct: gated false, cardData.license llama3.1, pipeline_tag text-generation, lastModified 2025-05-09T19:32:57.000Z, sha 4699cc75b550f9c6f3173fb80f4703b62d946aa5. The card YAML includes extra_gated_prompt. The model information table prints 8B, 70B, 405B, and 128k. hf download is added on this page and was not run. The training-hardware section is not copied
Public GPU memory: https://cloudgpu.app/v1/public/gpu-catalog (generated_at 2026-11-14T09:10:32.863Z). GET /v1/public/compute-capabilities returned feature flags and no GPU list. The catalog had no Langflow template

Cards you can start are on the Deploy page. Amounts are not copied here.

Last updated: 2026-11-14 (Asia/Shanghai).