langchain-octen

LangChain integration for Octen Web Search. Use Octen's fast, real-time web search inside LangChain agents and chains.

Installation

pip install langchain-octen

Setup

Get an API key from octen.ai and set it as an environment variable:

export OCTEN_API_KEY="your-api-key"

Or pass it directly when creating the tool:

from langchain_octen import OctenSearchResults

tool = OctenSearchResults(octen_api_key="your-api-key")

Usage

Basic Search

from langchain_octen import OctenSearchResults

tool = OctenSearchResults(max_results=5)
results = tool.invoke({"query": "latest AI research papers"})
for r in results:
    print(r["title"], r["url"])

With Domain Filtering

# Only include results from specific domains
results = tool.invoke({
    "query": "machine learning",
    "include_domains": ["arxiv.org", "papers.nips.cc"],
})

# Exclude results from specific domains
results = tool.invoke({
    "query": "Python tutorial",
    "exclude_domains": ["youtube.com", "reddit.com"],
})

In a LangChain Agent

from langchain_octen import OctenSearchResults
from langgraph.prebuilt import create_react_agent

tool = OctenSearchResults(max_results=5)
agent = create_react_agent(llm, [tool])
response = agent.invoke({"messages": [("user", "What happened in AI today?")]})

Async Support

import asyncio
from langchain_octen import OctenSearchResults

tool = OctenSearchResults(max_results=3)

async def main():
    results = await tool.ainvoke({"query": "Python 3.13 new features"})
    print(results)

asyncio.run(main())

Advanced Options

from langchain_octen import OctenSearchResults

# Time-filtered search
tool = OctenSearchResults(
    max_results=5,
    time_basis="published",
    start_time="2025-01-01",
    end_time="2026-03-30",
)
results = tool.invoke({"query": "Python release notes"})

# Search with highlighted snippets
tool = OctenSearchResults(
    max_results=5,
    highlight_enabled=True,
    highlight_max_tokens=200,
)
results = tool.invoke({"query": "LangChain tutorial"})

# Search with full page content
tool = OctenSearchResults(
    max_results=3,
    full_content_enabled=True,
    full_content_max_tokens=1000,
)
results = tool.invoke({"query": "Python documentation"})

Configuration

Development

# Install dependencies
poetry install --with test,dev

# Run unit tests (no API key needed)
make test

# Run integration tests (requires OCTEN_API_KEY)
export OCTEN_API_KEY="your-api-key"
make integration_tests

# Run all tests with coverage
poetry run pytest tests/ --cov=langchain_octen --cov-report=term-missing

# Lint and format
make lint
make format

License

MIT