pi-yolo-auto
Pi provider extension for the Yolo-Auto flat-rate Qwen3.8-27B API (OpenAI-compatible, stale-while-revalidate model catalog sync, subscription usage).
Package details
Install pi-yolo-auto from npm and Pi will load the resources declared by the package manifest.
$ pi install npm:pi-yolo-auto- Package
pi-yolo-auto- Version
0.1.6- Published
- Aug 28, 2026
- Downloads
- 657/mo · 40/wk
- Author
- ryanjoserbrosas
- License
- MIT
- Types
- extension
- Size
- 52.4 KB
- Dependencies
- 0 dependencies · 1 peer
Pi manifest JSON
{
"extensions": [
"./index.ts"
],
"image": "https://raw.githubusercontent.com/ryan-brosas/pi-yolo-auto/main/assets/yolo-auto-logo-mark-128.png"
}Security note
Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.
README
pi-yolo-auto
A pi provider extension for Yolo-Auto's flat-rate Qwen3.8-27B API
Adds an OpenAI-compatible provider with auto model-catalog sync and per-session subscription (Free/Builder/Pro) usage to pi.
Run
pi --model yolo-auto/qwen3.8-27b "hello"
Loads the provider from the pi package registry and runs one turn against the
Yolo-Auto endpoint. The extension serves the embedded model catalog immediately
and refreshes it from GET /models on session start, so the first real prompt
already has a resolved model id.
Why pi-yolo-auto?
| Capability | What it unlocks | |
|---|---|---|
| ⚡ | Flat-rate LLM access | A paid Qwen3.8-27B endpoint integrated like any native pi provider, with no per-token surprise. |
| 🔄 | Self-updating model catalog | Stale-while-revalidate sync keeps models.json current without downtime or manual edits. |
| 📊 | Subscription visibility | The status footer shows your tier (Free/Builder/Pro) and request counters straight from /usage. |
| 🪟 | Plan-aware context | Qwen3.8-27B reports 128K context on Free/Builder and 256K on Pro, matching the site tiers; the catalog hot-swaps when the detected plan changes. |
How it fits
flowchart LR
API1[/GET /models/] --> Sync[update-models.js]
API2[/GET /usage/] --> Footer[(status footer)]
Sync --> Models[models.json]
Models --> Patch[patch.json]
Patch --> Custom[custom-models.json]
Custom --> Ext[index.ts provider]
Ext --> Footer
The extension entry point (index.ts) is the composition root: it registers the
provider, owns the stale-while-revalidate loop over the embedded catalog, and
feeds the subscription footer. The pure pipeline (models.ts, usage.ts) stays
dependency-free so it is unit-testable without pi-ai.
Available Models
| Model | Context | Vision | Reasoning | Input $/M | Cache Read $/M | Output $/M |
|---|---|---|---|---|---|---|
| Qwen3.8 27B | 131K | ✅ | ✅ | $0.00 | $0.00 | $0.00 |
Install
Install it through pi's package registry:
pi install npm:pi-yolo-auto
Then load the extension and a yolo-auto provider appears in /model:
pi update --extensions # or restart pi
Auth
Get your API key from your Yolo-Auto account, then log in from inside pi (recommended):
/login yolo-auto
You'll be prompted to paste the key; it is stored in ~/.pi/agent/auth.json
(key in the access slot) and never printed back.
The /login path is the one and only auth path. No environment variable.
Usage
pi --list-models yolo-auto
pi --model yolo-auto/qwen3.8-27b "hello"
Credentials are read from ~/.pi/agent/auth.json (written by /login) and
resolved at request time; they are never stored by this package or printed.
Documentation
- Configuration & data ownership: AGENTS.md
- Model catalog pipeline: models.ts
- Subscription parser: usage.ts
- Model cache & sync: scripts/update-models.js
License
MIT, © 2026 Ryan Brosas. Third-party licenses are listed in THIRD_PARTY_NOTICES.md when they exist.
