perplexity-search-sdk

Python SDK for the Perplexity Search API: live web search and query-relevant page snippets. A thin Python wrapper over a compiled core built from Perplexity-internal source.

Install

pip install pplx-srch-sdk

CPython 3.12+ on Linux (x86_64, aarch64) and macOS. The import name matches: import pplx_srch_sdk.

Quick start

Get an API key at perplexity.ai/account/api and export it as PERPLEXITY_API_KEY.

import pplx_srch_sdk

hits = pplx_srch_sdk.search.web("rust async runtimes", limit=5)
for hit in hits:
    print(hit.title, hit.url)

snippets = pplx_srch_sdk.content.snippets(
    query="rust async runtimes",
    urls=[hit.url for hit in hits],
)
for s in snippets:
    print(s.url, s.error or (s.text or "")[:200])

An async client with the same surface is available as pplx_srch_sdk.AsyncPplxClient.

Search filters and token budgets

hits = pplx_srch_sdk.search.web(
    "rust async runtimes",
    published_after_date="3/1/2026",
    published_before_date="3/5/2026",
    updated_after_date="3/1/2026",
    max_tokens=8000,
    max_tokens_per_page=4000,
)

Publication and last-updated bounds use MM/DD/YYYY; updated_before_date sets an upper last-updated bound. Use recency_filter="hour" (or "day", "week", "month", "year") for relative publication recency instead of published_after_date / published_before_date.

max_tokens caps total search-result content tokens; max_tokens_per_page caps content tokens per result page. Omit search_context_size when passing explicit token budgets. Inspect help(pplx_srch_sdk.search.web) for all options supported by the installed version.

Development

The Python wrapper (src/pplx_srch_sdk/) is maintained in this repository; PRs welcome. bin/ holds the compiled core, delivered prebuilt by Perplexity's internal release pipeline — never edited by hand. pip install . packs the host platform's .so via scripts/hatch_build.py; wheels for any supported platform build on any host with PPLX_SDK_PLATFORM=<tag> python -m build --wheel.

Apache-2.0 — see LICENSE.