mem0ai
Protobuf version constraint prevents deployment with LangGraph Cloud and langgraph-api
🐛 Describe the bug
Description
The current protobuf version constraint in mem0ai (protobuf>=5.29.0,<6.0.0) creates an incompatibility with LangGraph deployments, which require protobuf>=6.32.1. This prevents mem0ai from being used in production LangGraph Cloud deployments.
Environment
- mem0ai version:
1.0.0 - Python version:
3.11.9 - Deployment platform: LangGraph Cloud
- Conflicting package:
langgraph-api 0.5.7
The Problem
Current Constraint (mem0ai 1.0.0)
protobuf>=5.29.0,<6.0.0
Required by LangGraph
langgraph-api==0.5.7 requires protobuf>=6.32.1,<7.0.0
grpcio-tools>=1.75.0 requires protobuf>=6.31.1,<7.0.0
Result
Dependency resolution fails during deployment:
× No solution found when resolving dependencies:
╰─▶ Because only mem0ai<=1.0.0 is available and mem0ai==1.0.0 depends on
protobuf>=5.29.0,<6.0.0, we can conclude that mem0ai>=1.0.0 depends on
protobuf>=5.29.0,<6.0.0.
And because stacey==0.3.1 depends on mem0ai>=1.0.0 and
protobuf>=6.32.1,<7.0.0, we can conclude that stacey==0.3.1 cannot be
used.
Impact
Who Is Affected
- LangGraph users: Cannot use mem0ai in LangGraph Cloud deployments
- LangChain users: Similar issues in production environments requiring protobuf 6.x
- Any async framework: Using langgraph-api, grpcio-tools, or other packages requiring protobuf 6.x
Deployment Environments
- ❌ LangGraph Cloud - Strict dependency resolution fails build
- ❌ Docker containers - Dependency conflicts prevent installation
- ⚠️ Local development - Works but shows warnings
- ❌ CI/CD pipelines - Automated deployments fail
Why This Matters
1. LangGraph is Growing
LangGraph is becoming the standard for production AI agent deployments. The incompatibility locks out a significant user base.
2. Protobuf 6.x is the New Standard
Google's protobuf library has moved to version 6.x (current: 6.33.0). Most modern gRPC-based tools now require it:
grpcio>=1.75.0→ requires protobuf 6.xlanggraph-api>=0.5.0→ requires protobuf 6.x- Future versions will increasingly adopt this
3. The Constraint is Unnecessary
Mem0ai doesn't actually use protobuf in its core functionality. It's only a transitive dependency through:
qdrant-client(vector database)- Possibly
grpcio(if used)
The HTTP API client (which is what users actually interact with) has no protobuf dependency.
Verification That Protobuf 6.x Works
We've tested mem0ai with protobuf 6.33.0 in local development:
$ pip list | grep -E "(mem0|protobuf)"
mem0ai 1.0.0
protobuf 6.33.0
$ python -c "from mem0 import AsyncMemoryClient; print('✓ Works')"
✓ Works
Result: Everything works perfectly. The HTTP API functions correctly despite the version warning.
Suggested Fix
Update the protobuf constraint in mem0ai to support both major versions:
Option 1: Wide Range (Recommended)
# In mem0ai's pyproject.toml or setup.py
"protobuf>=5.29.0,<7.0.0" # Support both 5.x and 6.x
Option 2: Protobuf 6+ Only
"protobuf>=6.31.1,<7.0.0" # Move to protobuf 6.x
Testing
Please test with:
- protobuf 5.29.0 (current minimum)
- protobuf 6.33.0 (latest 6.x)
- protobuf 6.31.1 (minimum for LangGraph)
All should work fine since mem0 uses HTTP APIs, not protobuf serialization.
Workarounds (Current)
Since mem0ai can't be deployed, users have to:
1. Use HTTP API Directly
Bypass the SDK entirely and call the REST API using httpx:
import httpx
class Mem0HttpClient:
def __init__(self, api_key: str):
self.client = httpx.AsyncClient(base_url="https://api.mem0.ai")
self.headers = {"Authorization": f"Token {api_key}"}
async def add(self, messages, user_id, **kwargs):
response = await self.client.post(
"/v1/memories/",
json={"messages": messages, "user_id": user_id, **kwargs},
headers=self.headers
)
return response.json()
async def search(self, query, filters, **kwargs):
response = await self.client.post(
"/v2/memories/search/",
json={"query": query, "filters": filters, **kwargs},
headers=self.headers
)
return response.json()
This works but defeats the purpose of having an SDK.
2. Force Install (Not Production-Safe)
pip install --no-deps mem0ai
pip install <all mem0 dependencies manually>
This is fragile and not suitable for automated deployments.
3. Fork and Patch
Fork mem0ai and modify the protobuf constraint locally. Not sustainable.
References
- LangGraph Deployments: https://docs.langchain.com/langgraph-platform/
- Protobuf 6.x Migration: https://protobuf.dev/support/cross-version-runtime-guarantee/
- Related Issue: (Link to any existing protobuf-related issues if found)
Related Dependencies
For reference, here are mem0ai's dependencies that might be affected:
# Current mem0ai dependencies
openai>=1.90.0
posthog>=3.5.0
protobuf>=5.29.0,<6.0.0 # ← THE ISSUE
pydantic>=2.7.3
pytz>=2024.1
qdrant-client>=1.9.1
sqlalchemy>=2.0.31
Only qdrant-client potentially uses protobuf (for gRPC vector search). All other functionality is HTTP-based.
Proposed Timeline
This is blocking production deployments for LangGraph users. A patch release with updated protobuf constraints would be greatly appreciated.
Suggested version: mem0ai==1.0.1 with protobuf>=5.29.0,<7.0.0
Additional Context
Why We Need This
We're building an AI agent platform on LangGraph that needs:
- Long-term memory across conversations (mem0's core value prop)
- Production deployment on LangGraph Cloud
- Session-scoped and user-scoped memories
Currently, we've had to implement a custom HTTP client to work around this limitation. We'd love to use the official SDK once the protobuf constraint is updated.
Test Case
To verify the fix works, test this installation sequence:
# Fresh environment
python -m venv test_env
source test_env/bin/activate
# Install langgraph-api first (gets protobuf 6.x)
pip install langgraph-api==0.5.7
# Then install mem0ai (should work with updated constraint)
pip install mem0ai==1.0.1 # Future version
# Verify no conflicts
pip check
# Test functionality
python -c "from mem0 import AsyncMemoryClient; print('✓ Compatible!')"
Thank You!
Thank you for considering this fix. Mem0 is a fantastic tool and updating this constraint will make it accessible to the growing LangGraph ecosystem. Happy to provide more details or test pre-release versions if helpful!
