Issue with private python library and LangGraph Platform deployment

[majorgilles]

Hello, I already posted in langgraph-cloud-help but received no help yet.

• do you know if it's possible to trigger a langgraph-cloud (so, hosted 100% by langgraph) via an langgraph CLI call? reason being we need to generate some token dynamically, and I don't think we can write scripts in the langgraph cloud UI to do that for us as part of the deployment process (when creating a new revision, for example, it would need a new token to authorize download some private company artifcat / library)

• new deployments + revisions are having issues with Env variables:
  I have a working configuration for the dockerisation process in langgraph_studio with a script that copies an env variable in .env:

#!/bin/bash

# Install dependencies
# Make sure Leapp is configured with the shared AWS account profile
CODEARTIFACT_AUTH_TOKEN=$(aws codeartifact get-authorization-token --domain theDomain --domain-owner theOwner --query authorizationToken --output text  --profile theProfile)
export CODEARTIFACT_AUTH_TOKEN

# Remove existing CODEARTIFACT_AUTH_TOKEN line from .env
sed -i '' '/^CODEARTIFACT_AUTH_TOKEN=/d' .env

# Add the new CODEARTIFACT_AUTH_TOKEN to .env
echo "CODEARTIFACT_AUTH_TOKEN=$CODEARTIFACT_AUTH_TOKEN" >> .env

poetry config http-basic.codeartifact aws $CODEARTIFACT_AUTH_TOKEN
poetry update

{
  "dockerfile_lines": [
    "ENV PIP_EXTRA_INDEX_URL=https://aws/:$CODEARTIFACT_AUTH_TOKEN@theDomain-theOwner.d.codeartifact.eu-west-1.amazonaws.com/pypi/theRepository/simple/"
  ],
  "graphs": {
    "fine_data_extraction_v1": "./projects/fine_data_extraction/graph_v1.py:graph",
    "picture_categorization_v1": "./projects/picture_categorization/graph_v1.py:graph",
    "lizy_bot_v0": "./projects/lizy_bot/graph_v0.py:graph"
  },
  "env": "./.env",
  "python_version": "3.12",
  "dependencies": [
    "."
  ]
}

with an .env that looks like that:

LANGCHAIN_API_KEY=
LANGCHAIN_TRACING_V2=true
LANGCHAIN_ENDPOINT=https://eu.api.smith.langchain.com/
OPENAI_API_KEY=
AWS_DEFAULT_REGION=eu-west-1
AWS_ACCESS_KEY_ID=
AWS_SECRET_ACCESS_KEY=
TAVILY_API_KEY=
ANTHROPIC_API_KEY=
POETRY_HTTP_BASIC_FRONTEND_USERNAME=aws

CODEARTIFACT_AUTH_TOKEN=

LangGraph Studio is working fine with this setup and is correctly able to build the docker image and download the artifact.
Yet when I input the CODEARTIFACT_AUTH_TOKEN which is required for the build to function in LangGraph cloud, it's not able to correctly auth to my private repository (edited)

I'm setting this env variable manually just like any other env variable:

I added some lines to RUN echo 'this is the token: ' + $CODEARTIFACT_AUTH_TOKEN and it doesn't show up!
=> when I click on "New revision", despite having tried to add $CODEARTIFACT_AUTH_TOKEN env varible, the CODEARTIFICAT_AUTH_TOKEN appears not to be saved, I need to add it AGAIN via the "+ Add" button.

I have the same issue with new deployments for which I've set env variables, they don't seem to be persisted at all:

I didn't have that issue before when I created my initial deployments a few weeks ago.