Is there any way to run this on windows?

[bracesproul]

We're working on adding better support for running LangGraph Cloud apps on Windows/Linux in the coming weeks! I'll post an update in this repo & update the Readme once that's ready

[lanesky]

Hi, I know the thread is closed, but I wanted to share the approach I successfully used on Windows. I referenced the two links below:

• https://langchain-ai.github.io/langgraph/how-tos/deploy-self-hosted/#build-the-docker-image
• https://langchain-ai.github.io/langgraph/cloud/reference/cli/#dockerfile

Here are the steps:

1. Install the langgraph CLI:

   pip install -U langgraph-cli

2. Use the langgraph CLI to create a Dockerfile:

   langgraph dockerfile -c langgraph.json Dockerfile

   In the directory containing langgraph.json, a Dockerfile will be created with the following content:

   FROM langchain/langgraphjs-api:20
   
   ADD . /deps/open-canvas
   
   RUN cd /deps/open-canvas && yarn install --frozen-lockfile
   
   ENV LANGSERVE_GRAPHS='{"agent": "./src/agent/open-canvas/index.ts:graph", "reflection": "./src/agent/reflection/index.ts:graph", "thread_title": "./src/agent/thread-title/index.ts:graph"}'
   
   WORKDIR /deps/open-canvas

3. Build the Docker image with the command below:

   docker build -t open-canvas .

4. Run the Docker container with the command below:

   docker run `
      --env-file .env `
      -p 57318:8000 `
      -e REDIS_URI="redis://host.docker.internal:6379/0" `
      -e DATABASE_URI="postgresql://myuser:mypassword@host.docker.internal:5432/mydatabase" `
      -e LANGSMITH_API_KEY="xxxxxxxxxxxxxxxxxxxxxxxxxxx" `
      open-canvas

   Note:

   • 57318 is the port number, which must match the one in constants.ts.
   • host.docker.internal is used to access the host from within the Docker container.
   • Redis can be installed in WSL.
   • A LANGSMITH_API_KEY is required.

Additionally, I noticed there are three places where the model name is hardcoded and cannot be selected from the UI:

• src/agent/open-canvas/nodes/updateArtifact.ts and src/agent/open-canvas/nodes/updateHighlightedText.ts are hardcoded to use gpt-4o.
• src/agent/reflection/index.ts is hardcoded to use claude-3-5-sonnet-20240620.

[Julsgaard]

@lanesky Do you know why the LangSmith API key is needed if LangGraph is open source and can be run locally?

I'm working on a project where privacy is important, so we need to run everything locally, including the LLM (hopefully this is possible with open-canvas). Any insights would be greatly appreciated!

[lanesky]

While LangGraph is indeed open source, it appears that not all components are included in the public codebase. Here's what I found:

Looking at their documentation (https://langchain-ai.github.io/langgraph/concepts/deployment_options/), they provide a Self-Hosted solution that uses Redis as a realtime message broker. However, when examining the code in their public GitHub repository (https://github.com/langchain-ai/langgraph), I noticed that the Redis-related code isn't included. Interestingly, if you look at the Dockerfile for the Self-Hosted plan's Docker image langchain/langgraphjs-api:20 (https://hub.docker.com/layers/langchain/langgraphjs-api/20/images/sha256-25ff0937c4803c271430376c66d9f4eae4e17b448efb917797431fe22220486f?context=explore), you'll see:

COPY /api /api # buildkit
COPY /storage /storage # buildkit

I think these parts are licensed codes.

[IsraelShok]

@Julsgaard I also trying to find a way to run this locally to save privacy. If you find a way, please share. Thanks!