Skip to content
AggregaAggrega
Select language: English

Import a LangGraph application — one graph and its checkpointer

The langgraph.json of a LangGraph application — one graph, keeping its state in PostgreSQL — drafted into Aggrega files by aggrega import, with the model the graph calls and the application’s image answered once in an answers file, and built for Docker Compose.

Import a LangGraph application walks it step by step: the command, what it asks for, the files it writes and why each line is there, the build, and the graph running. This folder holds each stage of that walkthrough:

folder or filewhat it holds
source/the application’s langgraph.json, the .aggrega-config.yaml that makes the folder a workspace, and import-answers.yaml, the values no configuration holds: the model, the application’s image, and the names of the Kubernetes Secrets a build for Kubernetes reads the key and the database’s password from
draft/the files aggrega import . --target compose wrote, unchanged, release.lock included, and what it printed: stdout.txt on standard output, import.txt on standard error

The configuration is written for this example. The application’s code is not part of it: the import never reads it, and the Release names the image it runs in.

In a checkout of the documentation repository, check.sh runs the import again, compares what it writes and prints with draft/, compares the blocks the walkthrough quotes with these files, and validates and builds what the import wrote. It needs the standard-library catalog (where to get one) checked out beside this repository, or its directory named in AGGREGA_CATALOG_DIRS.

Scaffold the example

You can run this example’s import with the commands below, or have your AI assistant run it for you.

Write the example into a new folder named after it:

Terminal window
aggrega scaffold --example import-langgraph

It then prints the commands to run next. The first changes into the new folder:

Terminal window
cd import-langgraph

Build

The import drafts the Aggrega files beside the project, in source/. It weighs the project against the standard-library catalog, which the workspace’s default registry publishes as a signed snapshot: the first import fetches it, verifies it, keeps it in the workspace’s content cache and names it by its digest in the record it writes beside the drafts. A workspace that names its own catalog folders — in AGGREGA_CATALOG_DIRS, or under workspace.catalog_dirs — imports against those instead. The import reads the model and the image from import-answers.yaml, then pulls what the drafted Release depends on, so the build follows at once:

Terminal window
cd source
aggrega import . --target compose
aggrega build --release release.yaml

The image the example names stands for one you build from your project; building it and running the stack are in the walkthrough linked above.

With an AI assistant

An AI assistant set up with Aggrega’s tools can do the same: paste this prompt into it.

Use the Aggrega tools only, and create no folder yourself. If the Workspace folder holds no `import-langgraph` folder yet, call the `scaffold` tool once with `example` set to `import-langgraph` and no `dir`: it writes the example into a new `import-langgraph` folder, whose `source` folder holds what this example imports. Then call the `import` tool with `workspace_root` set to `import-langgraph/source`, `source` set to `.` and `target` set to `compose`: it reads the answers it needs from `import-answers.yaml` beside `langgraph.json`, writes its drafts into `import-langgraph/source` and pulls what they need. Then call the `validate` tool with `workspace_root` set to `import-langgraph/source` and `release` set to `release.yaml`, and the `build` tool with `workspace_root` set to `import-langgraph/source`, `release` set to `release.yaml` and `write` set to true. Tell me which files the build wrote; they are under `import-langgraph/source/dist/`.

Test it

With the stack running as the walkthrough runs it, one request reaches the graph. The route below is the walkthrough’s own program’s, a small HTTP server on port 8000; your own program decides its own. The Release publishes the graph on 127.0.0.1:8000, so the request goes from your machine:

Terminal window
curl -sS -X POST http://127.0.0.1:8000/threads/order-1234/runs -H 'Content-Type: application/json' \
-d '{"message": "Where is my order 1234? It has not arrived yet."}'
{
"thread": "order-1234",
"error": "the run failed: OpenAIAuthenticationError: Error code: 401 - Incorrect API key provided: sk-place************-key. [… omitted: where to find an API key]"
}

The run’s key was a placeholder, so the answer is OpenAI’s refusal of it. On Kubernetes, forward a local port to the Deployment, which has no Service, and send the same request to it from another terminal:

Terminal window
kubectl port-forward deploy/support 8000:8000
Terminal window
curl -sS -X POST http://127.0.0.1:8000/threads/order-1234/runs -H 'Content-Type: application/json' \
-d '{"message": "Where is my order 1234? It has not arrived yet."}'

The answer is the same.