sentrial 0.1.2

Sentrial - Observability and time-travel debugging for AI agents

Sentrial Python SDK

Observability and time-travel debugging for AI agents. Track every decision, tool call, and state change in your agent workflows.

Features

• Automatic Tracking: Capture agent reasoning, tool calls, and state changes
• Git-like Branching: Fork agent sessions at any point
• Time Travel: Replay and debug past agent executions
• Framework Agnostic: Works with LangChain, custom agents, and more
• Visual UI: Beautiful web interface to explore agent behavior

Installation

From PyPI

# Standard installation
pip install sentrial

# With LangChain integration
pip install sentrial[langchain]

# All integrations
pip install sentrial[all]

From GitHub

pip install git+https://github.com/neelshar/Sentrial.git#subdirectory=packages/python-sdk

Local Development

cd packages/python-sdk
pip install -e .

Quick Start

Basic Usage

from sentrial import SentrialClient

# Initialize client
client = SentrialClient(
    api_url="https://api.sentrial.app",  # Or your self-hosted URL
    project_id="your-project-id"
)

# Create a session
session_id = client.create_session(name="Customer Support Agent")

# Track tool calls
client.track_tool_call(
    session_id=session_id,
    tool_name="search_knowledge_base",
    tool_input={"query": "password reset"},
    tool_output={"articles": ["KB-001", "KB-002"]},
    reasoning="Searching for relevant articles"
)

# Track LLM decisions
client.track_decision(
    session_id=session_id,
    reasoning="User already tried KB solutions. Escalating to human support.",
    alternatives=["Try another KB article", "Ask for more info"],
    confidence=0.85
)

# Close session
client.close_session(session_id)

LangChain Integration

from sentrial import SentrialClient, SentrialCallbackHandler
from langchain.agents import AgentExecutor, create_react_agent

# Initialize Sentrial
client = SentrialClient(api_url="...", project_id="...")
session_id = client.create_session(name="LangChain Agent")

# Create callback handler
handler = SentrialCallbackHandler(client, session_id, verbose=True)

# Use with LangChain - automatic tracking!
agent_executor = AgentExecutor(
    agent=agent,
    tools=tools,
    callbacks=[handler],  # ← That's it!
    verbose=True
)

result = agent_executor.invoke({"input": "Help user with login issue"})

The callback handler automatically tracks:

• Agent reasoning (Chain of Thought)
• Tool calls (inputs & outputs)
• Tool errors
• Agent completion

Examples

• simple_agent.py - Basic agent tracking
• langchain_agent.py - Full LangChain integration

Documentation

• Getting Started
• API Reference
• LangChain Integration

Support

• Email: [e-mail]
• Discord: Join our community
• Issues: GitHub Issues

License

MIT License - see LICENSE for details.