netweb-sdk 0.1.1

Netweb Enterprise Agent Engineering Framework

Netweb SDK

Netweb SDK is an enterprise-grade agent engineering framework for building, orchestrating, and deploying AI assistants with custom tools, workflows, and LLM providers.

• Custom tool integration with @nw_tool decorator
• Composable agents via create_agent
• Structured tool execution with JSON tool calls
• Built-in providers for OpenAI, Anthropic, Google, and open-source models
• Workflow orchestration for multi-agent and handoff patterns

Why Netweb SDK

Netweb SDK is designed for production AI applications that need:

• deterministic tool invocation from generated outputs
• reusable agent components and prompt configuration
• enterprise workflows for multi-step tasks
• provider-agnostic architecture for LLM integration

Installation

Install from PyPI:

pip install netweb-sdk

Or install from source:

pip install .

Quick Start

1. Define a custom tool

Use @nw_tool to expose an async Python function as a structured tool that agents can call.

from sdk import nw_tool

@nw_tool("add", description="Add two integers together")
async def add(a: int, b: int) -> int:
    return a + b

2. Create an agent

Use create_agent to instantiate an agent with a role, goal, LLM provider, and tool set.

from sdk import create_agent, nw_tool, OpenAIProvider

@nw_tool("add", description="Add two integers together")
async def add(a: int, b: int) -> int:
    return a + b

llm = OpenAIProvider(model="gpt-4o-mini")

agent = create_agent(
    name="calculator_agent",
    role="calculator",
    goal="perform accurate arithmetic using available tools",
    llm=llm,
    tools=[add],
    extra_instructions="You are a precise calculator. Use the add tool for addition operations."
)

3. Run the agent

response = await agent.nw_run("Add 5 and 3 together")
print(response.output)

Output:

8

API Overview

nw_tool

nw_tool registers a function as a structured LangChain tool and wraps it for SDK execution.

• name: tool identifier
• description: user-facing tool description
• accepts async callables
• exposes execute(**kwargs) and schema()

Example:

from sdk import nw_tool

@nw_tool("summarize", description="Summarize text into concise bullet points")
async def summarize(text: str) -> str:
    return "..."

create_agent

Creates an SDK agent instance with tools, system prompts, and memory.

from sdk import create_agent, OpenAIProvider

agent = create_agent(
    name="research_agent",
    role="technical researcher",
    goal="produce high-quality technical research articles",
    llm=OpenAIProvider(model="gpt-4o-mini"),
    tools=[add],
    extra_instructions="Always return valid JSON when invoking tools."
)

agent.nw_run(input)

Executes the agent on a user prompt and returns a structured AgentResponse.

response = await agent.nw_run("Calculate 12 + 34")
print(response.output)

Advanced Use

Combining tools and memory

The SDK supports conversation memory and tool pipelines so agents can remember context and execute tool calls consistently.

Workflow orchestration

For advanced multi-agent workflows, use WorkflowBuilder, SupervisorWorkflow, and HandoffWorkflow from the SDK.

Publishing to PyPI

Ensure the package metadata is up to date in pyproject.toml and README.md is present.

Build and upload:

python -m pip install --upgrade build twine
python -m build
python -m twine upload dist/*

Notes

• Set your LLM provider credentials using environment variables for OpenAI, Anthropic, or Google before running agents.
• The package is built for Python 3.12 and above.
• Use structured JSON tool outputs to keep LLM tool calls reliable.