perplexity

name	perplexity
description	Deep web research via OpenRouter web_search server tool. Uses any model + grounded search with citations.
version	0.1.1
type	extension
runtime	python3
entry	plugin.py
permissions	["net","tool","route","widget","read_settings"]
env_from_settings	["OPENROUTER_API_KEY"]
when_to_use	User asks for deep research, fact-checking with sources, or web search with LLM synthesis and citations.
timeout_sec	120
ui_tab	{"tab_id":"research","title":"Perplexity Research","icon":"search","render":{"kind":"declarative","schema_version":1,"components":[{"type":"form","route":"research","method":"POST","target":"research_results","fields":[{"name":"query","label":"Research question","type":"text","placeholder":"Ask anything — get a grounded answer with sources...","required":true},{"name":"max_results","label":"Max search results","type":"text","placeholder":"5"}],"submit_label":"Research"},{"type":"markdown","target":"research_results","path":"answer"},{"type":"json","target":"research_results","path":"citations"}]}}

Perplexity Research

An LLM-grounded web search extension using OpenRouter's openrouter:web_search server tool. Sends your question to a model that can search the web in real-time and returns a synthesized answer with citations.

Requirements

• OPENROUTER_API_KEY configured in Settings

How it works

1. Your query is sent to OpenRouter with the openrouter:web_search tool enabled
2. The model (Claude Sonnet by default) decides when to search and synthesizes results
3. You get a grounded answer plus a list of source URLs

Capabilities

• Agent tool: ext_..._perplexity_research — deep research with citations
• HTTP route: POST /api/extensions/perplexity/research — JSON body {"query": "..."}
• Widget: declarative form on the Widgets page

Network policy

Connects to OpenRouter API (openrouter.ai) only. The actual web search is performed server-side by OpenRouter's infrastructure.