pi

pi talks to Inferam over OpenAI Chat Completions.

With the CLI

inferam pi

Arguments pass through: inferam pi -p "summarise the diff", inferam pi --continue, inferam pi --session <id>.

The CLI adds an inferam provider for the run through a temporary pi extension (pi -e), deleted when pi exits. Your ~/.pi/agent settings, extensions and sessions are used as they are, and none of your files are modified. If you don’t pass --model or --provider, the CLI selects inferam/DeepSeek-V4.1-Flash-1017.

By hand

Set up before 30 October 2026?

The context window is now 262,144 tokens. Change 524288 to 262144 in your setup below, or switch to the CLI (run inferam update first). With the old value, long sessions get a 400 instead of compacting.

Add a provider to ~/.pi/agent/models.json:

{
  "providers": {
    "inferam": {
      "baseUrl": "https://api.inferam.ai/v1",
      "api": "openai-completions",
      "apiKey": "$INFERAM_API_KEY",
      "models": [
        { "id": "DeepSeek-V4.1-Flash-1017", "name": "DeepSeek-V4.1-Flash-1017", "reasoning": true, "contextWindow": 262144, "maxTokens": 32768 }
      ]
    }
  }
}

Then:

export INFERAM_API_KEY=ifr_...
pi --model inferam/DeepSeek-V4.1-Flash-1017

"$INFERAM_API_KEY" tells pi to read the key from that environment variable, so the key itself never goes in the file. (On Windows, set it with $env:INFERAM_API_KEY = "ifr_..." in PowerShell.)