LangChain Integration

Integrate Zubbl with LangChain agents.

Setup

pip install zubbl-sdk langchain

Wrapping a LangChain Agent

from langchain.agents import create_openai_tools_agent
from zubbl import ZubblClient

zubbl = ZubblClient(api_key="zubbl_xxx")

# Create your LangChain agent
agent = create_openai_tools_agent(llm, tools, prompt)

# Wrap it with Zubbl
smart_agent = zubbl.wrap(agent)

# Use as normal — Zubbl records and learns
result = smart_agent.invoke({"input": "Analyze this codebase"})

Manual Integration

# Get recommendations before running
policies = zubbl.query(task_type="code_analysis")

# Inject into agent prompt
enhanced_prompt = f"{base_prompt}\n\nRecommended strategies:\n"
for p in policies:
    enhanced_prompt += f"- {p.pattern}: {p.description}\n"

# Run and record
result = agent.invoke({"input": task})
zubbl.ingest(
    trajectory_id=str(uuid4()),
    task_description=task,
    steps=extract_steps(result),
    status="success" if result else "failure",
)