langgraph/cloud/deployment/cloud/

[giscus[bot]]

langgraph/cloud/deployment/cloud/

Build language agents as graphs

https://langchain-ai.github.io/langgraph/cloud/deployment/cloud/

[marinij]

Hello!
Is there a way to create a new revision by command-line ?
Also, would it be possible to restart a deployment by command-line ?

[andrewnguonly]

Creating a new revision and restarting a deployment by command line are not supported at the moment.

Alternatively, LangGraph Cloud supports the ability to automatically update a deployment (i.e. create a new revision) on push to a branch. This can be configured in the Deployment Settings.

[vermapratyush]

When running langgraph up, I get the following build error
Seems like gcc isn't available in the base image?

#7 16.03 Downloading zipp-3.20.2-py3-none-any.whl (9.2 kB)
#7 16.04 Downloading leb128-1.0.8-py3-none-any.whl (3.8 kB)
#7 16.05 Downloading tzlocal-5.2-py3-none-any.whl (17 kB)
#7 16.06 Downloading zstd-1.5.5.1-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.8 MB)
#7 16.09    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.8/1.8 MB 67.2 MB/s eta 0:00:00
#7 16.09 Downloading mypy_extensions-1.0.0-py3-none-any.whl (4.7 kB)
#7 16.34 Building wheels for collected packages: clickhouse-sqlalchemy, ciso8601
#7 16.34   Building wheel for clickhouse-sqlalchemy (setup.py): started
#7 16.90   Building wheel for clickhouse-sqlalchemy (setup.py): finished with status 'done'
#7 16.90   Created wheel for clickhouse-sqlalchemy: filename=clickhouse_sqlalchemy-0.3.2-py3-none-any.whl size=53508 sha256=8fc7e355526c8e4b239d2861e5c1b1863b156265572843b2dc788a7ec1199a0a
#7 16.90   Stored in directory: /tmp/pip-ephem-wheel-cache-mvdwl9ae/wheels/d0/c6/3f/e92747d519bd8e6c779750883f6bd42189d71cd97a8e879c8f
#7 16.90   Building wheel for ciso8601 (pyproject.toml): started
#7 17.16   Building wheel for ciso8601 (pyproject.toml): finished with status 'error'
#7 17.17   error: subprocess-exited-with-error
#7 17.17
#7 17.17   × Building wheel for ciso8601 (pyproject.toml) did not run successfully.
#7 17.17   │ exit code: 1
#7 17.17   ╰─> [19 lines of output]
#7 17.17       /tmp/pip-build-env-m9g8ximl/overlay/lib/python3.11/site-packages/setuptools/_distutils/dist.py:261: UserWarning: Unknown distribution option: 'test_suite'
#7 17.17         warnings.warn(msg)
#7 17.17       /tmp/pip-build-env-m9g8ximl/overlay/lib/python3.11/site-packages/setuptools/_distutils/dist.py:261: UserWarning: Unknown distribution option: 'tests_require'
#7 17.17         warnings.warn(msg)
#7 17.17       running bdist_wheel
#7 17.17       running build
#7 17.17       running build_py
#7 17.17       creating build
#7 17.17       creating build/lib.linux-aarch64-cpython-311
#7 17.17       creating build/lib.linux-aarch64-cpython-311/ciso8601
#7 17.17       copying ciso8601/__init__.pyi -> build/lib.linux-aarch64-cpython-311/ciso8601
#7 17.17       copying ciso8601/py.typed -> build/lib.linux-aarch64-cpython-311/ciso8601
#7 17.17       warning: build_py: byte-compiling is disabled, skipping.
#7 17.17
#7 17.17       running build_ext
#7 17.17       building 'ciso8601' extension
#7 17.17       creating build/temp.linux-aarch64-cpython-311
#7 17.17       gcc -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -fPIC -DCISO8601_VERSION=2.3.1 -DCISO8601_CACHING_ENABLED=1 -I/usr/local/include/python3.11 -c isocalendar.c -o build/temp.linux-aarch64-cpython-311/isocalendar.o
#7 17.17       error: command 'gcc' failed: No such file or directory
#7 17.17       [end of output]
#7 17.17
#7 17.17   note: This error originates from a subprocess, and is likely not a problem with pip.
#7 17.17   ERROR: Failed building wheel for ciso8601
#7 17.17 Successfully built clickhouse-sqlalchemy
#7 17.17 Failed to build ciso8601
#7 17.17 ERROR: Could not build wheels for ciso8601, which is required to install pyproject.toml-based projects
#7 17.25
#7 17.25 [notice] A new release of pip is available: 24.0 -> 24.2
#7 17.25 [notice] To update, run: pip install --upgrade pip
#7 ERROR: process "/bin/sh -c PYTHONDONTWRITEBYTECODE=1 pip install --no-cache-dir -c /api/constraints.txt -r /deps/__outer_inference/inference/requirements.txt" did not complete successfully: exit code: 1
------
 > [langgraph-api 3/7] RUN PYTHONDONTWRITEBYTECODE=1 pip install --no-cache-dir -c /api/constraints.txt -r /deps/__outer_inference/inference/requirements.txt:
17.17       [end of output]
17.17
17.17   note: This error originates from a subprocess, and is likely not a problem with pip.
17.17   ERROR: Failed building wheel for ciso8601
17.17 Successfully built clickhouse-sqlalchemy
17.17 Failed to build ciso8601
17.17 ERROR: Could not build wheels for ciso8601, which is required to install pyproject.toml-based projects
17.25
17.25 [notice] A new release of pip is available: 24.0 -> 24.2
17.25 [notice] To update, run: pip install --upgrade pip
------
time="2024-09-13T23:13:52+01:00" level=warning msg="The \"line\" variable is not set. Defaulting to a blank string."
failed to solve: process "/bin/sh -c PYTHONDONTWRITEBYTECODE=1 pip install --no-cache-dir -c /api/constraints.txt -r /deps/__outer_inference/inference/requirements.txt" did not complete successfully: exit code: 1

[vermapratyush]

Fixed it by adding the following to langgraph.json. Shouldn't this be install in the base image itself.

"dockerfile_lines": [
    "RUN apt-get update && apt-get install -y gcc"
  ],

[suyashkumar2409]

Hey, I am actively using langgraph and langgraph studio to build my AI agents. I would love to see what you are doing for langgraph cloud, and whether it makes my use cases simpler to execute.

What is the procedure for gaining beta access?

[hwchase17]

its currently available for all users with a Plus account

[vbarda]

Hi @suyashkumar2409! We just added more deployment options for LangGraph Platform (formerly LangGraph Cloud), including "Self-hosted Lite" - a limited free version. It is available to anyone with a free (developer) LangSmith account. Please check out the documentation to learn more: https://langchain-ai.github.io/langgraph/concepts/deployment_options/#self-hosted-lite. let me know if you have any questions!

[antoremin]

Seems like "interrupt" button is grayed out on my instance, and there's no delete button altogether. Would love to have a way to stop all current threads/redeploy from the latest commit.

[andrewnguonly]

You can create a new revision to redeploy from the latest commit of the specified branch for your deployment. There's no way to delete an individual revision. If the Interrupt button is grayed out, then the revision is in a terminal state (i.e. it can't be interrupted).

If you want to delete the entire deployment (including all revisions for the deployment), you can delete the deployment from the deployments list view. See Delete Deployment in the documentation.

[alchemistgo87]

I don't see any "New Deployment" button on deployments tab. How can I deploy a github repo?

[andrewnguonly]

After clicking on the Deployments tab in the left-hand navigation, do you see the following screenshot at the top of your screen?

If not, can you share a screenshot of what you see after clicking on the Deployments tab?

[alchemistgo87]

Hi Andrew,
Thanks for the reply!
Sharing the screenshot of what I see on my deployments page:
https://drive.google.com/file/d/1-gLQ4Wsm9B3OXxkVF_7x4axhE0scbhcn/view?usp=sharing

[andrewnguonly]

LangGraph Cloud is only available for paid LangSmith plans (Plus, Premier, Startup Enterprise). Based on your screenshot, it looks like your LangSmith plan is a free Developer plan. The free Developer plan is not able to access LangGraph Cloud. You'll need to upgrade your LangSmith plan to a paid plan in order to access LangGraph Cloud.

[alchemistgo87]

Ok, got it. Thanks for the help!

[achrafmam2]

I am using https://www.pantsbuild.org for building our python binaries. Pants can output a docker image. Is it possible to use a custom docker image for deployment?

[andrewnguonly]

At this moment, LangGraph Cloud does not support deploying from a custom Docker image.

[belmai]

It is possible to connect gitlab repo?

[andrewnguonly]

Currently, it's not possible to connect and deploy from a GitLab repository. As a workaround, you can mirror a GitLab repository to a GitHub repository.

Repository mirroring: https://docs.gitlab.com/ee/user/project/repository/mirror/

[yerx]

Hi, I created a new deployment and added environment variables. I want to add more environment variables to my project, but I can't find a button to update environment variables. Could you point me in the right direction?

[andrewnguonly]

To add/change environment variables, you'll have to create a new revision.

Create new revision: https://langchain-ai.github.io/langgraph/cloud/deployment/cloud/#create-new-revision

[phfifofum]

Trying to deploy to cloud and getting a timeout on the the server starting up. Build passes fine, seems to be a server issue but the logs dont give me any insight....
Revision ID - 02a1678d35-bbf6-4770-8b3d-00ae2800ee3d
Any ideas team?
06/02/2025, 11:19:21 Using auth of type=langsmith
06/02/2025, 11:19:21 Started server process [1]
06/02/2025, 11:19:21 Waiting for application startup.
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 Applied database migration
06/02/2025, 11:19:21 No LANGGRAPH_STORE configuration found, using default configuration
06/02/2025, 11:19:34 Using auth of type=langsmith
06/02/2025, 11:19:34 Started server process [1]
06/02/2025, 11:19:34 Waiting for application startup.
06/02/2025, 11:19:34 No LANGGRAPH_STORE configuration found, using default configuration
06/02/2025, 11:19:51 Using auth of type=langsmith
06/02/2025, 11:19:52 Started server process [1]
06/02/2025, 11:19:52 Waiting for application startup.
06/02/2025, 11:19:52 No LANGGRAPH_STORE configuration found, using default configuration
06/02/2025, 11:20:27 Using auth of type=langsmith
06/02/2025, 11:20:27 Started server process [1]
06/02/2025, 11:20:27 Waiting for application startup.
06/02/2025, 11:20:27 No LANGGRAPH_STORE configuration found, using default configuration
06/02/2025, 11:21:10 Using auth of type=langsmith
06/02/2025, 11:21:10 Started server process [1]
06/02/2025, 11:21:10 Waiting for application startup.
06/02/2025, 11:21:10 No LANGGRAPH_STORE configuration found, using default configuration

[andrewnguonly]

I looked at the deployment. I'm not sure what the issue is, but I think the issue might be a fluke. Can you try creating a new revision for the deployment and see if the issue persists?

[phfifofum]

just replied on slack, i tried it again with fresh deployment and no beuno - 33f980f5-4ff7-4829-af95-6b3e04f04679

Again this works fine locally and in studio

[kihumban]

I previously hosted my Langgraph backend API locally using a Docker container, and everything worked as long as the container was running. The frontend was hosted on Vercel.
Recently, I migrated the backend API to Langgraph Cloud and redeployed the frontend on Vercel after removing the previous deployment. I updated the environment variable NEXT_PUBLIC_API_URL to point to the new cloud API server and modified the frontend code accordingly.

However, the application is now failing to work, and I keep receiving error messages - "Unable to find the current thread ID" in the client application. This normally happens the client application cannot connect to the backend API server. Langgraph cloud automatically create a new API key when you deploy, however, I am unable to get the value of that API key. I am suspecting the issue is the API key that I have in the environmental variable is the issue.

I updated the environment variable NEXT_PUBLIC_API_URL to point to the new cloud API server and modified the frontend code accordingly:

export const createClient = () => {
const apiUrl = process.env.NEXT_PUBLIC_API_URL;
const apiKey = process.env.LANGSMITH_API_KEY;
return new Client({
apiUrl: apiUrl,
apiKey: apiKey
});
};

Here is error from the browser console:

Error creating thread HTTPError: HTTP 403: {"detail":"Missing authentication headers"}
at HTTPError.fromResponse (7933-37d17a5b62912814.js:256:24595)
at async 7933-37d17a5b62912814.js:256:25713
at async RetryOperation._fn (7933-37d17a5b62912814.js:204:6647)

Could you please advise on what additional authentication details or configuration might be required?

[andrewnguonly]

However, the application is now failing to work, and I keep receiving error messages - "Unable to find the current thread ID" in the client application. They're simply separate deployments.

Clarifying: Any existing thread IDs that were created from a local deployment will not exist in a new cloud deployment.

Langgraph cloud automatically create a new API key when you deploy, however, I am unable to get the value of that API key.

This is expected. You won't be able to access the API key that's automatically created for the deployment. You'll need to create another key (a second one), which will be used to authenticate with the API server.

Authentication: https://langchain-ai.github.io/langgraph/cloud/reference/api/api_ref/#authentication

I am suspecting the issue is the API key that I have in the environmental variable is the issue.

Can you confirm that process.env.LANGSMITH_API_KEY is set to a valid API key that was created from the same LangSmith organization where the LangGraph deployment was created? For example, just print out the value of process.env.LANGSMITH_API_KEY to double check.

[kihumban]

@andrewnguonly Thank you for the feedback.

I created a new API key which is what I have used in the environmental variable and I have confirmed it's correct. I have also tried with one of the existing API key. However, I am still getting the same error. The application is working perfectly in Langgraph Studio

[kihumban]

@andrewnguonly I am successfully connecting to the API through the Jupyter notebook. For instance I run this code successfully. where the URL is my langgraph cloud API URL.

`client = get_sync_client(url= url)

for chunk in client.runs.stream(
None, # Threadless run
"chat", # Name of assistant. Defined in langgraph.json.
input={
"messages": [{
"role": "human",
"content": "What is hope forum?",
}],
},
stream_mode="updates",
):
print(f"Receiving new event of type: {chunk.event}...")
print(chunk.data)
print("\n\n")`

The app is also running successfully on Langgraph Studio and on the browser on a locally hosted API server (http://localhost:8123). The issue come when I connect to the langgraph cloud server and run the client on the browser. The server is throwing a 403 error - {"Missing authentication headers"} . See the image below. 

What I am missing? I have the paid Startup account.

[romanh24]

@kihumban hi,
I have the same error. How did you fix it?

[kihumban]

@romanh24 I reconfigured a couple of the environmental variables. I am presuming you have your backend API on langgraph cloud. Where is your client - local or on the cloud e.g. Vercel? Mine was on Vercel

[romanh24]

@kihumban
After deploying to the Platform, I updated the .env file on my local machine with the API URL(is it correct? screen) from Platform and am using the useStream() React hook in my Next.js app. However, I'm seeing the following error in the console:
HTTPError: HTTP 403: {"detail":"Missing authentication headers"}.

I'm currently using use client. If there's a shorter or better way to handle this with use server, I'd like to try it.

My next step is to deploy to Vercel, but I want to test it locally first.

[vfzale]

If I may ask, what were the environmental variables that you reconfigured? Also are the values for API_BASE_URL and NEXT_PUBLIC_API_URL the same?

[romanh24]

If I may ask, what were the environmental variables that you reconfigured? Also are the values for API_BASE_URL and NEXT_PUBLIC_API_URL the same?

Not sure what you mean under API_BASE_URL.
Check my last post with the solution that worked for me.

[kihumban]

@vfzale if your backend is hosted in Langgraph cloud, the API_BASE_URL is your Langgraph cloud API URL. NEXT_PUBLIC_API_URL is frontend react site URL or localhost:[port]/api if locally hosted.

That is the way I have mine configured and it works.

[vfzale]

It works now, thank you for your help!

[kihumban]

What does your NEXT_PUBLIC_API_URL or equivalent variable reference? Can you try http://localhost:8123/ for local deployment

[romanh24]

For local development, I use in the useStream() hook:
NEXT_PUBLIC_LANGGRAPH_API_URL=http://127.0.0.1:2024

Code:

const thread = useStream<{ messages: Message[] }>({
   apiUrl: process.env.NEXT_PUBLIC_LANGGRAPH_API_URL,
   assistantId: process.env.NEXT_PUBLIC_LANGGRAPH_ASSISTANT_ID || "agentic_rag",
   messagesKey: "messages",
   threadId: threadId,
   onThreadId: setThreadId
 });

[kihumban]

I have not used the useStream method the way you have it, so I don't know how it's supposed to work. For me I have a langgraph cloud server client object which to make the API calls including creating threads.

export const createClient = () => { const apiUrl = process.env.NEXT_PUBLIC_API_URL; return new Client({ apiUrl, }); };

[romanh24]

Thanks for the answer.

The solution is to pass the apiKey value with the process.env.NEXT_PUBLIC_LANGSMITH_API_KEY or process.env.NEXT_PUBLIC_LANGRAPH_API_KEY.
In the apiUrl can be process.env.NEXT_PUBLIC_LANGSMITH_API_URL or process.env.NEXT_PUBLIC_LANGGRAPH_API_URL.

My current code:

const thread = useStream<{ messages: Message[] }>({
   apiUrl: process.env.NEXT_PUBLIC_LANGSMITH_API_URL,
   apiKey: process.env.NEXT_PUBLIC_LANGSMITH_API_KEY,
   assistantId: process.env.NEXT_PUBLIC_LANGGRAPH_ASSISTANT_ID || "agentic_rag",
   messagesKey: "messages",
   threadId: threadId,
   onThreadId: setThreadId
 });