Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/manusapis/Agix/llms.txt

Use this file to discover all available pages before exploring further.

Agix’s OpenAI integration sends every chat request to OpenAI’s /chat/completions REST endpoint using a standard Bearer token for authentication. The same OpenAIProvider class also powers the OpenRouter and Custom Endpoint options, because all three share the same OpenAI-compatible API surface. Once you enter your API key in the sidebar and save, your chosen model is ready to answer questions about your spreadsheet data.

​
Default configuration

The default configuration registered in providers.config.ts is: 
{
  id: "openai",
  name: "OpenAI",
  apiKey: "",                          // Supplied at runtime in the sidebar
  baseUrl: "https://api.openai.com/v1",
  defaultModel: "gpt-4o-mini",
  enabled: true,
}
gpt-4o-mini is the default because it offers a strong balance of speed and cost for the everyday spreadsheet tasks Agix handles. You can type a different model name into the Model field at any time without changing any default configuration.

​
Setting up OpenAI in the sidebar

1

Open the Agix sidebar

In Excel, click Home → Agix. In Google Sheets, open the Agix add-on from the Extensions menu.
2

Select the OpenAI provider

In the Provider dropdown, choose OpenAI.
3

Enter your API key

Paste your sk-... key into the API Key field. Keys are stored in the Office document settings and never leave your device in plaintext.
4

Choose a model (optional)

The default model is gpt-4o-mini. Change it to gpt-4o or any other model name supported by your account.
5

Save

Click Save. Agix calls settingsStore.save() and your configuration persists for the current document.

​
AIProviderConfig shape for OpenAI

{
  id: "openai",
  name: "OpenAI",
  apiKey: "sk-...",
  baseUrl: "https://api.openai.com/v1",
  defaultModel: "gpt-4o-mini",
  enabled: true
}

​
Request parameters

OpenAIProvider.chat() forwards the following fields to the /chat/completions endpoint:
​
model
string
required
The model to use, e.g. gpt-4o-mini or gpt-4o. Taken from the ChatRequest.
​
messages
ChatMessage[]
required
The full conversation history, including any system, user, and assistant turns.
​
temperature
number
default:"0.7"
Controls response randomness. Agix uses 0.7 when the caller does not supply a value.
​
max_tokens
number
Maximum tokens in the response. Omitted from the request when not specified, which lets OpenAI use its own default.
​
stream
boolean
default:"false"
Always set to false — Agix uses non-streaming responses.

​
Response shape

A successful call returns a ChatResponse: 
interface ChatResponse {
  provider: "openai";
  model: string;
  content: string;          // Text of the first choice
  finishReason?: string;    // e.g. "stop", "length"
  usage?: {
    promptTokens: number;
    completionTokens: number;
    totalTokens: number;
  };
}
Token usage is mapped from the OpenAI usage object (prompt_tokens, completion_tokens, total_tokens).

​
Shared implementation with OpenRouter

OpenAI and OpenRouter both resolve to the same OpenAIProvider class in the factory: 
case "openai":
case "openrouter":
case "custom":
  return new OpenAIProvider(config);
The only difference between them is the baseUrl and the model naming convention. This means any fix or improvement to OpenAIProvider applies to all three provider IDs automatically.
Use gpt-4o for complex multi-step analysis, formula generation, or tasks where accuracy matters most. Switch to gpt-4o-mini for quick data lookups, cell summaries, and high-volume operations where speed and cost efficiency are the priority.

Build docs developers (and LLMs) love

Get started for freeTalk to us