LangChain Reserp

A minimal LangChain tool for the Reserp Google Search API v2.

The tool accepts a complete Google Search URL, makes one request to POST /v2/serp/search, and returns the public JSON payload unchanged. Its results[] array is flat, page ordered, and deduplicated, with optional visible text for each URL. The package adds no retries, timeout policy, concurrency management, caching, queues, result conversion, or automatic pagination.

Installation

pip install langchain-reserp

Usage

import os

from langchain_reserp import ReserpSearchTool

os.environ["RESERP_API_KEY"] = "your-api-key"

tool = ReserpSearchTool()
result = tool.invoke(
    {"url": "https://www.google.com/search?q=photonic+computing&gl=us&hl=en"}
)

if result["ok"]:
    for item in result["results"]:
        print(item.get("text"), item["url"])

Every successful response includes pagination.next_url. Submit that value in a later tool call to advance; its presence does not prove another page contains results. Do not infer pagination from len(result["results"]).

The surrounding application owns retries, timeouts, task queues, concurrency, observability, and pagination. When the API reports retryable, this package exposes that field and takes no action itself.

Migrating

From 0.2, replace /v2/serp/urls with /v2/serp/search and urls[] with results[]. When migrating directly from v1, other notable field renames are url → request.url, finalUrl → page.url, pagination.nextUrl → pagination.next_url, and billingSource → billing_source.

For typed, page-ordered SERP blocks and explicit positions, call the structured v2 endpoint directly or use an official Reserp SDK.

API contract

• Reserp API documentation
• OpenAPI definition
• Postman collection

License

MIT