Implementation:Langchain ai Langgraph Config Graph Paths

Property	Value
API	`_update_graph_paths` and `_assemble_local_deps`
Type	API Doc
Source	`libs/cli/langgraph_cli/config.py`
Library	langgraph-cli
Related Workflow	CLI_Deployment

Overview

The `_update_graph_paths` and `_assemble_local_deps` functions are internal components of the LangGraph CLI that handle the mapping of graph import paths from the host filesystem to container-internal paths during Docker image building. `_assemble_local_deps` catalogues all local dependencies, classifying them as real packages or faux packages and determining their container destinations. `_update_graph_paths` then rewrites each graph's import string so it resolves correctly inside the Docker container.

Description

_assemble_local_deps

This function iterates over all entries in the `dependencies` list that start with `"."` (indicating a local path) and classifies each:

• Real packages: Directories containing `pyproject.toml` or `setup.py`. These are mapped to `/deps/<directory_name>` in the container and installed with `pip install -e .`.
• Faux packages: Directories without standard packaging files. These are mapped to `/deps/outer-<directory_name>/...` and receive a generated `pyproject.toml`.
• Requirements files: If a faux package directory contains a `requirements.txt`, it is tracked for installation before the package itself.

The function also handles:

• Reserved package name detection (prevents conflicts with `langgraph-api`, `pydantic`, `fastapi`, etc.).
• Duplicate directory name resolution via a counter.
• Working directory assignment when `"."` is in the dependencies.
• Additional Docker build contexts for dependencies in parent directories.

The result is a `LocalDeps` named tuple containing all the information needed to generate Dockerfile instructions and rewrite paths.

_update_graph_paths

This function iterates over each entry in the `graphs` dictionary and:

1. Extracts the import string (either from a plain string value or a dictionary's `"path"` key).
2. Splits the import string into module path and attribute name at the `":"` separator.
3. If the module path contains `/` or `\` (indicating a file path), resolves it to an absolute host path.
4. Searches through real packages and faux packages in `local_deps` to find the matching container path.
5. Rewrites the graph entry in the config to use the container-internal POSIX path.

LocalDeps Named Tuple

Field	Type	Description
`pip_reqs`	`list[tuple[pathlib.Path, str]]`	Local `requirements.txt` files and their container destinations
`real_pkgs`	`dict[pathlib.Path, tuple[str, str]]`	Real packages: host path to (dependency string, container name)
`faux_pkgs`	`dict[pathlib.Path, tuple[str, str]]`	Faux packages: host path to (dependency string, container path)
`working_dir`	None`	Container working directory if `"."` is in dependencies
`additional_contexts`	`list[pathlib.Path]`	Parent directory dependencies needing additional Docker build contexts

Code Reference

Source Location

Function	File	Line
`_update_graph_paths`	`libs/cli/langgraph_cli/config.py`	L492
`_assemble_local_deps`	`libs/cli/langgraph_cli/config.py`	L374
`LocalDeps`	`libs/cli/langgraph_cli/config.py`	L316

Signature

def _update_graph_paths(
    config_path: pathlib.Path, config: Config, local_deps: LocalDeps
) -> None:
    """Remap each graph's import path to the correct in-container path."""
    ...

def _assemble_local_deps(config_path: pathlib.Path, config: Config) -> LocalDeps:
    ...

Import

These are internal functions within the `langgraph_cli` package:

# Internal usage only (not part of public API)
from langgraph_cli.config import _assemble_local_deps, _update_graph_paths

I/O Contract

_assemble_local_deps

Direction	Name	Type	Description
Input	`config_path`	`pathlib.Path`	Resolved path to the configuration file
Input	`config`	`Config`	Validated configuration dictionary
Output	return	`LocalDeps`	Named tuple containing classified local dependencies
Raises	`FileNotFoundError`		If a local dependency path does not exist
Raises	`NotADirectoryError`		If a local dependency path is not a directory
Raises	`ValueError`		If a reserved package name is used or a hyphenated flat-layout name is detected

_update_graph_paths

Direction	Name	Type	Description
Input	`config_path`	`pathlib.Path`	Path to the configuration file
Input	`config`	`Config`	Configuration dictionary (mutated in place)
Input	`local_deps`	`LocalDeps`	Classified local dependencies from `_assemble_local_deps`
Output	return	`None`	The `config["graphs"]` dictionary is modified in place
Raises	`ValueError`		If the import string format is invalid or the module is not found in dependencies
Raises	`FileNotFoundError`		If the referenced local module file does not exist
Raises	`IsADirectoryError`		If the module path points to a directory

Usage Examples

Assembling Local Dependencies

import pathlib
from langgraph_cli.config import validate_config_file, _assemble_local_deps

config_path = pathlib.Path("langgraph.json")
config = validate_config_file(config_path)
local_deps = _assemble_local_deps(config_path, config)

# Inspect classified dependencies
for path, (dep_str, container_name) in local_deps.real_pkgs.items():
    print(f"Real package: {path} -> /deps/{container_name}")

for path, (dep_str, container_path) in local_deps.faux_pkgs.items():
    print(f"Faux package: {path} -> {container_path}")

Updating Graph Paths for Container

import pathlib
from langgraph_cli.config import (
    validate_config_file,
    _assemble_local_deps,
    _update_graph_paths,
)

config_path = pathlib.Path("langgraph.json")
config = validate_config_file(config_path)
local_deps = _assemble_local_deps(config_path, config)

# Before: config["graphs"]["agent"] == "./my_pkg/agent.py:graph"
_update_graph_paths(config_path, config, local_deps)
# After:  config["graphs"]["agent"] == "/deps/my_pkg/agent.py:graph"

Related Pages

• Langchain_ai_Langgraph_Graph_Module_Definition
• Langchain_ai_Langgraph_Validate_Config_File
• Langchain_ai_Langgraph_Langgraph_Build_Command