Multi-provider AI CLI

pplx-cli

One command surface for the models and knowledge you already use.

Choose Perplexity, NVIDIA NIM, or OpenRouter for chat. Keep your RAG knowledge base fast, local, and under your control.

Explore providers View source

Multi-provider chat / local RAG / LLM wiki

pplx-cli / session-042

$ install

pip install pplx-cli

$ configure default

perplexity setup --provider nvidia

✓ NVIDIA NIM key stored securely

$ ask with an override

perplexity ask --provider openrouter --model openrouter/free "Summarize the release"

Using provider: OpenRouter

Response ready · history saved locally

chat / hostedrag / localwiki / visual

provider: routed

01 / providers

Pick the best route for every question.

The CLI keeps provider setup uncomplicated: save a default once, or select a provider and native model per command when the task calls for it.

Perplexity

Sonar models with the familiar aliases for fast answers, reasoning, and deep research.

Default model

sonar

Use small, large, or huge

NVIDIA NIM

Route requests to NVIDIA-hosted models while keeping the same concise CLI workflow.

Default model

meta/llama-3.3-70b-instruct

Pass a native NVIDIA model ID

OpenRouter

Reach a broad model catalog through one provider, including the default free route.

Default model

openrouter/free

Pass any OpenRouter model slug

Credential flow

Environment variables take priority for CI and shell-specific credentials; otherwise perplexity setup stores the selected provider key and default locally.

02 / LLM wiki

Your LLM wiki, mapped—not buried.

Turn an interlinked Markdown vault into a living knowledge graph. See how notes, research, prompts, and open questions actually connect.

vault / relationships

active note / llm wiki — Your connected Markdown vault

Build the graph

$ perplexity knowledge-graph --dir ~/my-vault

Understands the links you write

Parses wikilinks, aliases, headings, Markdown links, relative paths, and external URLs.

Explore instead of browse

Drag, zoom, and inspect a force-directed graph that reveals clusters and high-connection notes.

Keep a portable artifact

Open a local graph immediately or save a standalone HTML file for offline sharing and review.

Save it offline

$ perplexity knowledge-graph --dir ~/my-vault --output ~/Desktop/wiki.html

Bring the context into RAG

03 / local context

Your knowledge has a home field advantage.

Hosted model choice and local retrieval are separate on purpose. Use the provider that fits the answer; keep the context you’ve built close.

See the four-step setup

local-rag / queryready

$ perplexity rag "How did we handle auth retries?"

mode hybrid  sources notes + chats
matches 12      latency 89ms

Hybrid retrieval

Vector similarity and keyword matching are fused so exact names and related ideas both surface.

One local knowledge base

Notes and chat history stay together in a portable SQLite-backed index built for fast recall.

Private by default

RAG indexing and retrieval happen on your machine; route only the answer generation you choose.

04 / quickstart

Four commands. Then make it yours.

Set a default once. Search locally, then make the connections in your Markdown wiki visible.

Install the CLI

$ pip install pplx-cli

Choose a default provider

$ perplexity setup --provider openrouter

Ask, then search locally

$ perplexity rag "what did we decide about retries?"

Map your LLM wiki

$ perplexity knowledge-graph --dir ~/my-vault

copy / paste / go

$ pip install pplx-cli$ perplexity setup --provider openrouter$ perplexity rag "what did we decide about retries?"$ perplexity knowledge-graph --dir ~/my-vault

No platform lock-in. Use the provider that fits.

Open source / MIT licensed / cross-platform

Read the docs on GitHub