API

When using the Fireaura API, specific versions can be targeted.

Using custom versions of Fireaura are prone the effects of cold starts

Base API Endpoints

https://fireaura.microart.cf/api/v3/
https://fireaura.cloud.microart.cf/quf/api/v3
https://fapi.microart.cf/v3/

These three endpoints are provided based on the underlying hardware and Fireaura subscription status.

For normal Fireaura users, the first endpoint is more than enough for almost all scenarios. The second endpoint provides extremely low-latency responses, however these are limited to 800ms compute times. Anything higher will be rejected to compute. The last endpoint is for Fireaura Ultimate subscription plans. These plans are given access to low-latency high performance compute nodes which have no time limits.

Using API's out of your tier will result in and error

Using custom versions

If a custom version is required, say for example v3 build 0b20bc9 (last build to include quick commands), the custom endpoint api can be utilized

https://fireaura.microart.cf/api/v3_0b20bc9/

At the moment, alternative API endpoints are not supported

API Specification

/info

Returns info about the current Fireaura build.

Example response

{
    "success": true,
    "data": {
        "build": "0f50c74b423417d63701369b0589f7",
        "version": 3,
        "date": "3/2/2021",
        "parameters": {
            "nIndex": 14,
            "weights": [
                0.11735386032341924,
                0.3124503535212076,
                0.13957716542690332,
                0.05008076902106477,
                ... <truncated>
            ],
            ... <truncated>
        },
        "compute": {
            "cpu": "intel/core/i7/6700k",
            "cpu": "amd/ryzen/9/5950x",
            "gpu": "nvidia/geforce/rtx/3090ti",
            "memory": 268435456,
            "allowed_time": 800
        }
    }
}

/compute/alternatives

Computes alternative ways to write the given text. Outputs are ordered by importance and how much better the AI thinks the alternative is.

POST /compute/alternatives

{ "text": "The quick brown lamb hopped over the chickens pen" }

Example response

{
    "success": true,
    "duration": 342.18905,
    "data": {
        "computed": {
            "alternatives": [
                "The quick brown fox jumped over the lazy dog",
                "The quick brown lamb hopped over the pigs pen",
                "The quick brown lamb hopped over the hens house",
                "The furry brown lamb hopped over the chickens pen"
            ]
        }
    },
    "compute": { COMPUTE NODE SPECS }
}

/compute/completion

Compute a completion to the given text. An optional background property can also be set which tells the AI how to understand the prompt better

POST /compute/completion

{

"text": "I got a",

"background": "a music lyric of a famous song"

}

Example Response

{
    "success": true,
    "duration": 203.329,
    "data": {
        "computed": {
            "completion": {
                "full": "I got a pocket, got a pocketful of sunshine",
                "added": " pocket, got a pocketful of sunshine",
                "original": "I got a",
                "accuracy": 0.9874,
                "thought_tree": [
                    "famous",
                    "music",
                    "lyrics",
                    "i got a",
                    "2007",
                    "Natasha Bedingfield",
                    "Pocketful of Sunshine",
                    "I got a pocket, got a pocketful of sunshine"
                ]
            }
        }
    }
}

Last updated