Skip to content
Fast Inference
DOCS / Pi
Integration

Pi

Register an Inference.net model for Pi Agent and route chat completions through the gateway.

The fast pi adapter adds Inference.net's OpenAI-compatible provider to Pi Agent. It updates settings.json, models.json, and auth.json under ~/.pi/agent while preserving a snapshot of all three files.

Connect Pi

  1. Install and sign in

    bash
    npm install -g @inference/fast
fast login

  2. Enable Pi routing

    bash
    fast pi on

    The default selection is glm-5.2. To start with another callable model:

    bash
    fast pi on --model <catalog-id>

  3. Verify the adapter

    bash
    fast pi status

    Restart Pi and send a prompt. The request should appear in Usage.

Files and model

models.json receives the provider endpoint and selected model. settings.json makes that provider and model active. auth.json stores the machine-scoped project key with an @inference/fast ownership marker.

The API surface is OpenAI-compatible Chat Completions at https://api.inference.net/v1.

Restore your configuration

bash
fast pi off
fast pi status

The adapter restores the original snapshots of all three files. Restart Pi afterward.

Troubleshooting

SymptomCheck
Inference.net is not listedRestart Pi after on and inspect models.json.
Wrong model startsRe-run fast pi on --model <catalog-id>.
Status says offConfirm settings.json selects inference-net.
Requests do not appearRun fast status and confirm the selected project.