Retrieve ranked raw web search results (no LLM-generated answer) via Perplexity's standalone Search API for retrieval-only agent pipelines.

Documented steps

1. POST to https://api.perplexity.ai/search with header Authorization: Bearer $PERPLEXITY_API_KEY (https://docs.perplexity.ai/api-reference/search-post).
2. Body: {"query": "<string or array of up to 5 strings>", "max_results": 10} — max_results ranges 1-20, default 10.
3. Optionally tune search_context_size ("low"/"medium"/"high", default "high"), max_tokens (content budget across all results, up to 1,000,000), and max_tokens_per_page (per-page extraction cap).
4. Apply the same filter parameters as Sonar: search_domain_filter (max 20), country (ISO 3166-1 alpha-2), search_language_filter (ISO 639-1, max 20), search_recency_filter, and the *_date_filter/last_updated_*_filter fields.
5. Read back {"results": [{title, url, snippet, date, last_updated}], "id", "server_time"} — there is no choices/message field because this endpoint returns raw ranked results, not a generated answer (https://docs.perplexity.ai/docs/getting-started/overview).
6. Use this endpoint instead of chat completions when you only need retrieval (e.g. to feed your own RAG pipeline) — it is billed flat per request with no token costs, unlike Sonar chat completions (https://docs.perplexity.ai/docs/getting-started/pricing).