pi-ppq — PayPerQ Provider for pi

Adds PayPerQ as a model provider in pi, giving you access to 50+ LLM models including Claude, GPT, Grok, Gemini, DeepSeek, Qwen, and more — all billed per-token with no subscription.

Also includes PayPerQ's smart routing models (auto, autoclaw/auto, autoclaw/eco, autoclaw/premium) that automatically pick the best model for each prompt, cutting costs by up to 70%.

Quick Start

1. Get an API key from your PayPerQ dashboard (it starts with sk-).

2. Authenticate in pi using the /login command:

/login

Select PayPerQ from the provider list, then paste your API key when prompted. pi stores it securely and uses it for all subsequent requests.

3. Pick a model with /model — look for entries prefixed ppq/:

/model ppq/claude-sonnet-4.6
/model ppq/auto
/model ppq/autoclaw/eco

Alternative: You can also set the PPQ_API_KEY environment variable instead of using /login. See Configuration for details.

Available Models

Models are fetched live from the PayPerQ API every time pi starts. You always get the latest models and current pricing without updating the extension.

| Category | Examples | Pricing | |----------|----------|---------| | Standard chat | claude-sonnet-4.6, gpt-5.5, grok-4.20, gemini-3-flash-preview, deepseek-v4-pro | Per-token from PPQ API | | Routing models | auto, autoclaw, autoclaw/auto, autoclaw/eco, autoclaw/premium | Dynamic (depends on selected model) |

Run pi --list-models to see the full list after setup.

Routing Models

PayPerQ offers two types of smart routers that analyze your prompt and pick the right model automatically:

auto

The general-purpose AI router. It scores your prompt and routes it to the best model for the job — simple questions go to fast/cheap models, complex reasoning to powerful ones. No configuration needed.

/model ppq/auto

autoclaw (and variants)

Optimized for OpenClaw and coding agents. Available in three profiles:

| Model ID | Profile | Behavior | |----------|---------|----------| | autoclaw | Auto (balanced) | Same as autoclaw/auto — good quality at a fair price | | autoclaw/auto | Auto (balanced) | Balanced quality and cost for each prompt tier | | autoclaw/eco | Eco | Maximum cost savings — routes to the cheapest viable model | | autoclaw/premium | Premium | Best quality — routes to top-tier models at every tier |

How it works: AutoClaw classifies your prompt across four tiers (Simple → Medium → Complex → Reasoning) and maps each tier to a different underlying model depending on the profile you chose. The routing happens server-side in under 1ms with no added latency.

Configuration

Authentication

Primary method — /login:

/login

Select PayPerQ from the list and enter your API key. pi saves it to its credential store and uses it automatically.

Alternative — environment variable:

export PPQ_API_KEY="sk-..."

When both are set, the environment variable takes precedence. This is useful for CI/CD pipelines or containerized environments where interactive login isn't practical.

Model Updates

Models and pricing are fetched from https://api.ppq.ai/v1/models at startup. If the endpoint is unreachable the provider still registers (without models) and you can /reload later to pick them up. No manual updates needed.

Troubleshooting

"No models found for provider ppq"

The models endpoint was unreachable during startup. Check your network connection and run /reload.

"No API key available"

No API key has been configured. Run /login and select PayPerQ to enter your key, or set the PPQ_API_KEY environment variable.

"Unknown model" or 404 errors from the API

Some model IDs may have changed on PayPerQ's side. Run /reload to refresh the model list. If a model was removed entirely, pick another from the updated list.

Routing model cost shows $0

Routing models (auto, autoclaw, etc.) show $0 because their cost depends on which underlying model the router picks. The actual per-request cost is reported in pi's usage display after each response.