Skip to content

Add llm-perplexity-agent to the plugin directory - #1570

Open
zalez wants to merge 1 commit into
simonw:mainfrom
zalez:add-llm-perplexity-agent
Open

Add llm-perplexity-agent to the plugin directory#1570
zalez wants to merge 1 commit into
simonw:mainfrom
zalez:add-llm-perplexity-agent

Conversation

@zalez

@zalez zalez commented Jul 28, 2026

Copy link
Copy Markdown

Adds one line to docs/plugins/directory.md for llm-perplexity-agent, which provides a perplexity-agent model.

llm install llm-perplexity-agent
llm -m perplexity-agent 'What changed in MCP 2026-07-28?'

Why a second Perplexity plugin

It is complementary to llm-perplexity, not a replacement, and I placed it directly after that entry so the distinction is visible from adjacency.

llm-perplexity wraps Perplexity's Sonar chat modelssonar, sonar-pro, sonar-deep-research, sonar-reasoning-pro. This wraps their Agent API (POST /v1/agent), a different endpoint: a multi-step agent that runs its own searches, fetches pages and synthesises one cited answer.

There is no model-name overlap, and both use the same perplexity key alias, so having both installed works with no extra configuration — which is how I run them.

Options

llm -m perplexity-agent -o preset xhigh -o recency week 'Latest on X'
llm -m perplexity-agent -o domains 'nasa.gov,-reddit.com' 'Artemis status'

preset selects research depth and is passed through unvalidated, since Perplexity's schema declares no enum for it. Poll progress goes to stderr so the answer still pipes cleanly.

Published on PyPI as llm-perplexity-agent. BSD-3-Clause. Tested against Python 3.10–3.14.

Provides the perplexity-agent model, wrapping Perplexity's Agent API. Placed
next to llm-perplexity since the two cover different Perplexity endpoints and
the adjacency makes the distinction visible.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant