OpenAI API

Basic API handling for openAI.

Uses the specification from: https://beta.openai.com/docs/api-reference/introduction

Changelog

v1.0.0

After GTP4 the API has breaking changes. This package still supports the legacy calls, but the calling proc has changed.

Endpoint Default openAI endpoint changed to support GPT4. All call-procs now includes an optional openAIendpoint parameter.

const
  urlCompletionLegacy = "https://api.openai.com/v1/completions"
  urlCompletion = "https://api.openai.com/v1/chat/completions"

proc aiPrompt*(apiKey, prompt: string, maxTokens = 30, openAIendpoint = urlCompletion): JsonNode =

Prompt options The main aiCreateRequest which formats the API-call has breaking changes. If you still use < GTP4, you need to specify the openAIendpoint = "legacy-call-url".

If you are using >= GTP4 you can rely on the default settings.

Authorization

You need an API key to use the API. The API key is used in the headers of the request.

let headers = aiHeaders(apiKey)
# Authorization: Bearer sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Environment variables

OpenAI is apparently in favor of saving keys to environmental variables.

If that's also your cup of tea, you can use the following:

# On linux:
export openAIKey="sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"

import std/os
echo getEnv("openAIKey")
# sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Results

All results are returned as a JsonNode object. That requires you to import nim's standard json library.

import std/json

Basic request

import
  std/json,
  openai

let resp = aiPrompt(apiKey, "Why is Nim-lang the best programming language?", maxTokens = 50)

echo resp["choices"][0]["message"]["content"]

# Legacy:
# echo resp["choices"][0]["text"]

# Nim-lang is the best programming language because it is a powerful, statically typed, compiled language that is designed to be fast, efficient, and expressive. It has a simple syntax that is easy to learn and understand, and it is

Custom request

Current

import
  std/json,
  openai

let
  envKey = getEnv("openAIKey")
  question = "Why is Nim-lang the best programming language?"
  max_tokens = 100
  n = 3       # number of choices to return

let req = aiCreateRequest(prompt = question, max_tokens = max_tokens, n = n)
check req == """{"model":"gpt-4","messages":[{"role":"system","content":"Why is Nim-lang the best programming language?"}],"temperature":0.0,"top_p":1,"n":3,"max_tokens":100,"presence_penalty":0,"frequency_penalty":0}"""

let resp = aiGetSync(envKey, req)

check resp["choices"].len() == 3

echo resp["choices"][1]["message"]["content"]

# Whether a language is \"the best\" is subjective and depends largely on the task at hand, personal preference, or specific project requirements. However, Nim-lang possesses some qualities that can make it stand out for certain situations:\n\n1. Efficiency: Nim compiles to C, C++, and JavaScript, offering efficient performance close to what you would get from these languages.\n\n2. Expressiveness: Nim allows programmers to write high-level code that is both human-understandable and machine-optimized. This balances readability

Legacy

import
  std/json,
  openai

let
  envKey = getEnv("openAIKey")
  question = "Why is Nim-lang the best programming language?"
  max_tokens = 100
  n = 3       # number of choices to return
  best_of = 5 # number of completion (must be higher than n)

let req = aiCreateRequest(envKey, prompt = question, max_tokens = max_tokens, n = n, best_of = best_of)
check req == """{"model":"text-davinci-003","prompt":"Why is Nim-lang the best programming language?","temperature":0,"max_tokens":100,"top_p":1,"n":3,"presence_penalty":0,"frequency_penalty":0,"best_of":5}"""

let resp = aiGetSync(envKey, req)
check resp["choices"].len() == 3

echo resp["choices"][1]["text"]

# Nim-lang is the best programming language because it is a powerful, statically typed, compiled language that is designed to be fast, efficient, and expressive. It has a simple syntax, a powerful macro system, and a modern type system. Nim-lang also has a great community of developers who are constantly working to improve the language and make it even better. Additionally, Nim-lang is open source and free to use, making it an attractive option for developers.

Public procedures

aiHeaders

proc aiHeaders*(apiKey: string): HttpHeaders =

Authorization: Bearer

---

aiGetAsync*

proc aiGetAsync*(client: AsyncHttpClient, url: string, headers: HttpHeaders, body: string): Future[JsonNode] {.async.} =

API: Async with custom client

---

aiGetAsync*

proc aiGetAsync*(apiKey, body: string, openAIendpoint = urlCompletion): Future[JsonNode] {.async.} =

API: Async with one-time client

---

aiGetSync*

proc aiGetSync*(client: HttpClient, url: string, headers: HttpHeaders, body: string): JsonNode =

API: Sync with custom client

---

aiGetSync*

proc aiGetSync*(apiKey, body: string, openAIendpoint = urlCompletion): JsonNode =

API: Sync with one-time client

---

aiCreateRequest*

proc aiCreateRequestLegacy*(
    prompt = "What is nim-lang",
    model = "gpt-4",
    role = "system",
    temperature = 0.0,
    maxTokens = 30,
    top_p = 1,
    n = 1,
    stop = "",
    presence_penalty = 0,
    frequency_penalty = 0,
    user = ""
): string =

Create request body

---

aiPrompt*

proc aiPrompt*(apiKey, prompt: string, maxTokens = 30, openAIendpoint = urlCompletion): JsonNode =

Basic prompt

---