| text-generation | prompt | Generates text based on a given prompt |
| sentiment-analysis | text | Determines the sentiment of a given text |
| question-answering | question | Provides an answer to a given question based on a given context |
| | context | |
| summarization | text | Generates a short summary of a given text |
| translation | text | Translates text from one language to another |
| | target | |
| language-detection | text | Detects the language of a given text |
| text-classification | text | Classifies a given text into one or more predefined categories |
| | categories | |
| named-entity-recog. | text | Identifies named entities in a given text |
| text-to-speech | text | Converts text into spoken audio |
| image-recognition | image | Identifies objects and scenes in an image |
| text-to-image | prompt | Generates an image based on a given prompt |
| semantic-search | query | Finds documents related to a given query |
| | documents | |
| chatbot | message | Generates responses to a given message |
## How To Use
Copy the folder inside the resouce folder of your server, start the script inside server.cfg. To get the Api key go to your account on https://beta.openai.com/account/api-keys and copy the key.
In this example i've used Davinci gpt 3.5 Turbo model, you can change the model and use what you need.