AgentDiff Coordination
AgentDiff is a lightweight coordination library that prevents common concurrency issues in multi-agent systems - such as agents starting before dependencies complete, multiple agents writing to shared resources, or concurrent API calls exceeding rate limits.
Simply add @coordinate decorators for resource locks and @when for event-driven chaining. AgentDiff integrates with existing agent frameworks like LangChain, CrewAI, or pure Python implementations.
Provides coordination primitives without requiring framework migration or architectural changes.
What AgentDiff Coordination Solves
- Race conditions between agents accessing shared resources.
- Corrupted state when multiple agents write to the same keys.
- API rate limit chaos from concurrent LLM calls.
- "Two nodes writing to same key" bugs that require extensive debugging.
- Framework complexity that gets in the way of actually building agents.
Core Features
@coordinate- Resource locks + automatic lifecycle events.@when- Event-driven agent chaining (no manual orchestration).emit()- Custom events for complex workflows.- Zero Configuration - Works immediately, configure only what you need.
- Framework Agnostic - Works with any agent framework or pure Python.
Use Cases: Concurrency Issues and Race Conditions
Concurrent State Updates
# Before: Race conditions in shared state
def process_customer_data():
customer_state["status"] = "processing" # Race condition
result = process_data()
customer_state["result"] = result # Overwrites other agent
def update_customer_profile():
customer_state["status"] = "updating" # Conflicts with processor
customer_state["profile"] = new_profile # State corruption
# After: Resource locks prevent conflicts
@coordinate("data_processor", lock_name="customer_123")
def process_customer_data():
customer_state["status"] = "processing" # Exclusive access
result = process_data()
customer_state["result"] = result # Safe update
@coordinate("profile_updater", lock_name="customer_123")
def update_customer_profile():
customer_state["status"] = "updating" # Waits for processor
customer_state["profile"] = new_profile # No conflicts
API Rate Limit Management
# Before: Multiple agents hitting APIs simultaneously
def research_agent():
response = openai.chat.completions.create(...) # Rate limited
def analysis_agent():
response = openai.chat.completions.create(...) # Rate limited
def summary_agent():
response = openai.chat.completions.create(...) # Rate limited
# Running in parallel creates debugging challenges
threading.Thread(target=research_agent).start()
threading.Thread(target=analysis_agent).start()
threading.Thread(target=summary_agent).start()
# After: Resource locks queue API calls safely
@coordinate("researcher", lock_name="openai_api")
def research_agent():
response = openai.chat.completions.create(...) # Queued safely
@coordinate("analyzer", lock_name="openai_api")
def analysis_agent():
response = openai.chat.completions.create(...) # Waits for researcher
@coordinate("summarizer", lock_name="openai_api")
def summary_agent():
response = openai.chat.completions.create(...) # Waits for analyzer
Manual Orchestration Complexity
# Before: Complex manual coordination
def run_workflow():
research_result = research_agent()
if research_result:
analysis_result = analysis_agent(research_result)
if analysis_result:
summary_result = summary_agent(analysis_result)
if summary_result:
final_report = editor_agent(summary_result)
# Error handling, retries, parallel flows increase complexity
# After: Event-driven coordination
@coordinate("researcher")
def research_agent():
return research_data
@when("researcher_complete")
def start_analysis(event_data):
analysis_agent(event_data['result']) # Auto-triggered
@when("analyzer_complete")
def start_summary(event_data):
summary_agent(event_data['result']) # Auto-chained
# Just start the workflow - coordination happens automatically
research_agent() # Everything else flows automatically
Installation & Requirements
Python Support: 3.9+ (tested on 3.9, 3.10, 3.11, 3.12)
pip install agentdiff-coordination
Documentation
- Quick Start Guide - Get up and running in 5 minutes.
- API Reference - Complete function documentation.
- Use cases - Example use cases.
- Configuration Guide - Environment variables and settings.
- Examples - Example agent coordination patterns.
Contributing
We welcome contributions! See CONTRIBUTING.md for guidelines.
License
MIT License - see LICENSE file for details.
About AgentDiff
AgentDiff provides practical tools for AI developers working with multi-agent systems. This coordination library addresses common concurrency challenges encountered in production agent workflows.
- GitHub: https://github.com/AgentDiff
- Issues: Report bugs and request features
- Community: Share coordination patterns and production experiences
Metadata
Release files for agentdiff-coordination 0.1.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| agentdiff_coordination-0.1.1.tar.gz | 330.2 kB | Without provenance | Details for agentdiff_coordination-0.1.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| agentdiff_coordination-0.1.1-py3-none-any.whl | Python 3 | none | any | Details for agentdiff_coordination-0.1.1-py3-none-any.whl |
Total release size: 360.3 kB
Release files / agentdiff_coordination-0.1.1.tar.gz
| Download URL | agentdiff_coordination-0.1.1.tar.gz |
|---|---|
| Size | 330.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
bd2de39482197865cd3563b8f68632c19ceb467e922fc36b456e91e8c54eb8cf
|
|
BLAKE2b-256 checksum How to use checksums |
3f3957833a0177b3b46c1a412017c820fa61ff6bf3fdd97341fbddbc2f2f4b4a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.0
|
Release files / agentdiff_coordination-0.1.1-py3-none-any.whl
| Download URL | agentdiff_coordination-0.1.1-py3-none-any.whl |
|---|---|
| Size | 30.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
26023af0a5dc3f2da9454f77478efae32ba87cc2d97c24cfb0e8d313be5c5a39
|
|
BLAKE2b-256 checksum How to use checksums |
fbc8d9d38bf5355ddf91c768cf7565c1500e065b80bf00b85da11514559f8efb
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.0
|
Security for agentdiff-coordination 0.1.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.