semanticapi-langchain

LangChain integration for Semantic API — find the right API for any task using natural language.

What is Semantic API?

Semantic API matches natural language queries to real API endpoints. Ask "send an SMS" and get back the best provider, endpoint, parameters, and auth docs — instantly.

✨ Auto-Discover & Auto-Heal

Auto-Discover: Your agent can query any API — even ones not in the catalog. Semantic API automatically finds the docs, generates a provider config, and returns endpoint details. No manual setup.

Auto-Heal: If an API's config goes stale (endpoints move, versions change), Semantic API detects the failure, re-discovers from current docs, and retries — all transparently. Your agent never sees a stale endpoint.

Built-in Retry: All tools include exponential backoff with retry logic for rate limits and transient errors — production-ready out of the box.

Installation

pip install semanticapi-langchain

Quick Start

from semanticapi_langchain import SemanticAPIToolkit

# Create toolkit (uses SEMANTICAPI_API_KEY env var)
toolkit = SemanticAPIToolkit(api_key="sapi_your_key")
tools = toolkit.get_tools()

# Use the query tool directly
query_tool = tools[0]
result = query_tool.run("send an SMS")
print(result)

Tools

With a LangChain Agent

from langchain_openai import ChatOpenAI
from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain_core.prompts import ChatPromptTemplate
from semanticapi_langchain import SemanticAPIToolkit

toolkit = SemanticAPIToolkit()
tools = toolkit.get_tools()
llm = ChatOpenAI(model="gpt-4o-mini")

prompt = ChatPromptTemplate.from_messages([
    ("system", "You find APIs for tasks. Use semanticapi tools to discover them."),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}"),
])

agent = create_tool_calling_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

response = executor.invoke({"input": "How do I send an email programmatically?"})
print(response["output"])

Using the Wrapper Directly

from semanticapi_langchain import SemanticAPIWrapper

api = SemanticAPIWrapper(api_key="sapi_your_key")

# Query for an API
result = api.query("send an SMS")

# Search capabilities
results = api.search("weather")

# List providers
providers = api.list_providers()

# Async support
result = await api.aquery("send an SMS")

Configuration

Development

git clone https://github.com/semanticapi/semanticapi-langchain.git
cd semanticapi-langchain
pip install -e ".[dev]"
pytest

License

MIT