Implementation:Langchain ai Langchain Partner Directory Layout

Overview

Concrete reference template for a LangChain partner integration package, provided by the langchain-deepseek package at libs/partners/deepseek/.

Description

The libs/partners/deepseek/ directory serves as the canonical reference for how a LangChain partner integration package is laid out. It demonstrates the full set of required directories and files that every partner package must include to function correctly within the monorepo's build, test, lint, and release pipelines.

Required directories:

Directory / File	Purpose
langchain_deepseek/	Python package containing integration source code
langchain_deepseek/__init__.py	Public API surface with __all__ and __version__
langchain_deepseek/chat_models.py	Chat model implementation (ChatDeepSeek)
langchain_deepseek/py.typed	PEP 561 type-checking marker
langchain_deepseek/data/	Static data (model profiles)
tests/unit_tests/	Unit tests (no network access)
tests/integration_tests/	Integration tests (real API calls)
scripts/	Package automation utilities
Makefile	Standardized make targets
pyproject.toml	Build system, dependencies, tool configuration
README.md	Package description and usage
LICENSE	MIT license
uv.lock	Locked dependency versions

Usage

Use this layout as a starting template when creating any new partner integration. Copy the directory structure from libs/partners/deepseek/ and replace deepseek with the new integration name throughout.

Code Reference

Source Location: libs/partners/deepseek/ (entire directory)

Directory tree:

libs/partners/deepseek/
  LICENSE
  Makefile
  README.md
  pyproject.toml
  uv.lock
  langchain_deepseek/
    __init__.py
    chat_models.py
    py.typed
    data/
      _profiles.py
  tests/
    __init__.py
    unit_tests/
      __init__.py
      test_chat_models.py
    integration_tests/
      __init__.py
      test_chat_models.py
  scripts/
    check_imports.py

Makefile targets:

# Run unit tests (network disabled)
test tests:
	uv run --group test pytest --disable-socket --allow-unix-socket $(TEST_FILE)

# Run integration tests (network enabled)
integration_test integration_tests:
	uv run --group test --group test_integration pytest --timeout=30 $(TEST_FILE)

# Lint: ruff check + ruff format --diff + mypy
lint lint_diff lint_package lint_tests:
	[ "$(PYTHON_FILES)" = "" ] || uv run --all-groups ruff check $(PYTHON_FILES)
	[ "$(PYTHON_FILES)" = "" ] || uv run --all-groups ruff format $(PYTHON_FILES) --diff
	[ "$(PYTHON_FILES)" = "" ] || mkdir -p $(MYPY_CACHE) && uv run --all-groups mypy $(PYTHON_FILES) --cache-dir $(MYPY_CACHE)

# Format code
format format_diff:
	[ "$(PYTHON_FILES)" = "" ] || uv run --all-groups ruff format $(PYTHON_FILES)
	[ "$(PYTHON_FILES)" = "" ] || uv run --all-groups ruff check --fix $(PYTHON_FILES)

I/O Contract

Input	Description
Directory structure	Must match the canonical layout above
pyproject.toml	Must declare langchain-core as a dependency
Makefile	Must provide test, lint, format, integration_tests targets

Output	Description
Installable package	pip install langchain-<name> from the built wheel
Passing CI checks	Lint, unit tests, and integration tests all pass

Usage Examples

Creating a new partner package from the template:

# Copy the deepseek package as a starting point
cp -r libs/partners/deepseek libs/partners/myservice

# Rename the Python package directory
mv libs/partners/myservice/langchain_deepseek libs/partners/myservice/langchain_myservice

# Update all references from "deepseek" to "myservice" in:
#   - pyproject.toml (name, description, URLs)
#   - __init__.py (imports)
#   - chat_models.py (class name, env vars)
#   - Makefile (lint_diff --relative path)
#   - tests/ (imports, class names)

Running tests in the partner package:

cd libs/partners/deepseek

# Install dependencies
uv sync --group test

# Run unit tests
make test

# Run integration tests (requires DEEPSEEK_API_KEY)
make integration_tests

Related Pages

• Principle:Langchain_ai_Langchain_Package_Scaffolding