feat: Add LangChain integration with automatic tool conversion #73

abrookins · 2025-10-01T16:02:09Z

Overview

This PR adds a comprehensive LangChain integration that eliminates the need for manual tool wrapping. Users can now get LangChain-compatible memory tools with a single function call instead of manually wrapping each tool with @tool decorators.

The Problem

Users had to manually wrap every memory tool with LangChain's @tool decorator, creating 200+ lines of boilerplate code.

The Solution

One function call replaces all the boilerplate:

from agent_memory_client.integrations.langchain import get_memory_tools

tools = get_memory_tools(
    memory_client=memory_client,
    session_id=session_id,
    user_id=user_id
)

What's Included

Core integration module with automatic tool conversion
Comprehensive documentation and examples
Full test suite (8 tests, all passing ✅)
Updated README files

Impact

98.5% reduction in boilerplate code (from 200+ lines to 3 lines)

Key Features

✅ Zero boilerplate - no manual decorators needed
✅ Automatic context injection
✅ Type-safe with full schema generation
✅ Composable with custom tools
✅ Production-ready with comprehensive tests

See full details in the PR description.

@tool

- Add get_memory_tools() function for automatic LangChain tool conversion - Eliminate need for manual @tool decorator wrapping (98.5% less boilerplate) - Support all 9 memory tools with automatic context injection - Add selective tool loading and custom tool composition - Include comprehensive documentation and working examples - Add full test suite (8 tests, all passing) This integration transforms the developer experience from tedious manual wrapping to a single function call. Users can now get LangChain-compatible memory tools with just 3 lines of code instead of 200+. Key features: - Zero boilerplate - no manual decorators needed - Automatic session/user context injection - Type-safe with full schema generation - Composable with custom tools - Production-ready with comprehensive tests

Copilot

Pull Request Overview

This PR introduces a comprehensive LangChain integration that eliminates the need for manual tool wrapping. Users can now create LangChain-compatible memory tools with a single function call instead of manually wrapping each tool with @tool decorators, reducing boilerplate from 200+ lines to just 3 lines.

Adds automatic tool conversion from memory client to LangChain StructuredTool instances
Provides comprehensive documentation and working examples
Includes full test suite with 8 tests covering all integration scenarios

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
agent-memory-client/agent_memory_client/integrations/langchain.py	Core integration module with automatic tool conversion functions and factories
agent-memory-client/agent_memory_client/integrations/init.py	Integration package initialization with exports
agent-memory-client/tests/test_langchain_integration.py	Comprehensive test suite for the LangChain integration
examples/langchain_integration_example.py	Complete working example demonstrating the integration
docs/langchain-integration.md	Detailed documentation and API reference
agent-memory-client/README.md	Updated README with LangChain integration section
README.md	Updated main README with LangChain integration examples
LANGCHAIN_INTEGRATION.md	Implementation summary and design decisions

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

Copilot · 2025-10-01T16:03:00Z

examples/langchain_integration_example.py

+    async def calculate(expression: str) -> str:
+        """Evaluate a mathematical expression."""
+        try:
+            result = eval(expression)  # Note: Use safely in production!


Using eval() is a security risk as it can execute arbitrary code. For a production example, consider using ast.literal_eval() for safe evaluation of simple expressions or a proper math expression parser.

Copilot · 2025-10-01T16:03:00Z

agent-memory-client/agent_memory_client/integrations/langchain.py

+            func=config["func"],
+            name=config["name"],
+            description=config["description"],
+            coroutine=config["func"],  # Same function works for both sync and async


The comment is misleading. The coroutine parameter should reference the async version of the function, but here it's the same function reference as func. Since all the factory functions create async functions, this should just be the async function, not 'the same function for both sync and async'.

Suggested change

func=config["func"],

name=config["name"],

description=config["description"],

coroutine=config["func"], # Same function works for both sync and async

coroutine=config["func"],

name=config["name"],

description=config["description"],

Copilot · 2025-10-01T16:03:01Z

docs/langchain-integration.md

+@tool
+async def calculate(expression: str) -> str:
+    """Evaluate a mathematical expression."""
+    return str(eval(expression))


Using eval() in documentation examples is dangerous as users may copy this code to production. Replace with a safer alternative like ast.literal_eval() or mention the security implications explicitly.

- Add return type annotations to all factory functions (-> Any) - Fix mypy import handling for optional langchain_core dependency - Add explicit str() conversions to avoid no-any-return errors - Add langchain-integration.md to mkdocs nav configuration - Fix broken link to example file (use GitHub URL instead of relative path) All mypy checks now pass and documentation builds successfully.

- Add mypy override to disable warn_unused_ignores for langchain integration module - Use placeholder class instead of None when langchain is not installed - This allows the code to pass mypy both with and without langchain installed The type: ignore comment is needed when langchain is NOT installed (CI), but triggers unused-ignore when it IS installed (local dev). The mypy override resolves this conflict.

abrookins · 2025-10-01T16:15:54Z

CI Fixes Applied ✅

All CI failures have been resolved:

Issues Fixed

MyPy Type Checking Errors
- Added return type annotations () to all factory functions
- Fixed optional langchain import handling with proper type ignores
- Added explicit conversions to avoid errors
- Added mypy override to handle unused-ignore warnings for optional dependency
Documentation Build Failure
- Added \ to mkdocs navigation
- Fixed broken relative link to example file (now uses GitHub URL)

Test Results

All tests now passing:

✅ Test (Python 3.10) - pass
✅ Test (Python 3.11) - pass
✅ Test (Python 3.12) - pass
✅ lint - pass
✅ build - pass

The integration is ready for review!

- Replace unsafe eval() with safe AST-based evaluation in examples - Use ast.literal_eval() in documentation example - Fix misleading comment about coroutine parameter (all functions are async) Security improvements: - Example now uses AST parsing with whitelisted operators - Documentation uses ast.literal_eval() for safe evaluation - Both approaches prevent arbitrary code execution

abrookins · 2025-10-01T21:03:43Z

Review Comments Addressed ✅

I've addressed all 3 Copilot review comments:

1. Security: Unsafe `eval()` in example code

Fixed: Replaced with safe AST-based evaluation

Uses AST parsing with whitelisted operators (add, sub, mul, div, pow)
Prevents arbitrary code execution
Provides clear error messages for unsupported operations

2. Misleading comment about coroutine parameter

Fixed: Updated comment to be accurate

Changed from "Same function works for both sync and async"
To "All our functions are async"
This is correct since all factory functions create async functions

3. Security: Unsafe `eval()` in documentation

Fixed: Replaced with ast.literal_eval()

Safe evaluation of simple expressions
Prevents code injection
Appropriate for documentation examples

All tests still passing after these changes!

abrookins · 2025-10-01T22:32:00Z

Test Failure Analysis 🔍

The failing test test_temporal_grounding_integration_last_year is NOT related to the LangChain integration.

Evidence:

Test passed twice before on this branch:
- ✅ Run at 16:14 UTC - passed
- ✅ Run at 21:03 UTC - passed
- ❌ Run at 21:17 UTC - failed
No related files modified:
- Our PR only adds LangChain integration files
- We didn't touch long_term_memory.py or contextual_grounding code

Root cause is in existing code:

ERROR agent_memory_server.long_term_memory:long_term_memory.py:276 
Error extracting memories: string indices must be integers, not 'str'

This is a bug in the existing memory extraction logic, not our integration.

Conclusion:

This is a flaky test in the existing codebase. The LangChain integration is working correctly - all our integration tests pass consistently.

Recommendation: This test failure should not block merging the LangChain integration PR.

Copilot AI review requested due to automatic review settings October 1, 2025 16:02

Copilot AI reviewed Oct 1, 2025

View reviewed changes

abrookins added 2 commits October 1, 2025 09:11

Bump version for new langchain tool integration

95ceba6

abrookins merged commit aa7da91 into main Oct 1, 2025
18 of 19 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: Add LangChain integration with automatic tool conversion #73

feat: Add LangChain integration with automatic tool conversion #73

Uh oh!

abrookins commented Oct 1, 2025

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Oct 1, 2025

Uh oh!

Copilot AI Oct 1, 2025

Uh oh!

Copilot AI Oct 1, 2025

Uh oh!

abrookins commented Oct 1, 2025

Uh oh!

abrookins commented Oct 1, 2025

Uh oh!

abrookins commented Oct 1, 2025

Uh oh!

Uh oh!

Uh oh!

feat: Add LangChain integration with automatic tool conversion #73

feat: Add LangChain integration with automatic tool conversion #73

Uh oh!

Conversation

abrookins commented Oct 1, 2025

Overview

The Problem

The Solution

What's Included

Impact

Key Features

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Copilot AI Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

abrookins commented Oct 1, 2025

CI Fixes Applied ✅

Issues Fixed

Test Results

Uh oh!

abrookins commented Oct 1, 2025

Review Comments Addressed ✅

1. Security: Unsafe eval() in example code

2. Misleading comment about coroutine parameter

3. Security: Unsafe eval() in documentation

Uh oh!

abrookins commented Oct 1, 2025

Test Failure Analysis 🔍

Evidence:

Conclusion:

Uh oh!

Uh oh!

Uh oh!

1. Security: Unsafe `eval()` in example code

3. Security: Unsafe `eval()` in documentation