| title | description | ms.topic | ms.custom | ms.collection | ms.date | zone_pivot_groups | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Azure OpenAI text completion input binding for Azure Functions | Learn how to use the Azure OpenAI text completion input binding to access Azure OpenAI text completion APIs during function execution in Azure Functions. | reference |
|
| 07/08/2024 | programming-languages-set-functions |
[!INCLUDE preview-support]
The Azure OpenAI text completion input binding allows you to bring the results text completion APIs into your code executions. You can define the binding to use both predefined prompts with parameters or pass through an entire prompt.
For information on setup and configuration details of the Azure OpenAI extension, see Azure OpenAI extensions for Azure Functions. To learn more about Azure OpenAI completions, see Learn how to generate or manipulate text.
[!INCLUDE functions-support-notes-samples-openai]
::: zone pivot="programming-language-csharp"
This example demonstrates the templating pattern, where the HTTP trigger function takes a name parameter and embeds it into a text prompt, which is then sent to the Azure OpenAI completions API by the extension. The response to the prompt is returned in the HTTP response.
:::code language="csharp" source="~/functions-openai-extension/samples/textcompletion/csharp-ooproc/TextCompletions.cs" range="23-31":::
This example takes a prompt as input, sends it directly to the completions API, and returns the response as the output.
:::code language="csharp" source="~/functions-openai-extension/samples/textcompletion/csharp-ooproc/TextCompletions.cs" range="37-46":::
::: zone-end
::: zone pivot="programming-language-java" This example demonstrates the templating pattern, where the HTTP trigger function takes a name parameter and embeds it into a text prompt, which is then sent to the Azure OpenAI completions API by the extension. The response to the prompt is returned in the HTTP response.
:::code language="java" source="~/functions-openai-extension/samples/textcompletion/java/src/main/java/com/azfs/TextCompletions.java" range="31-46":::
This example takes a prompt as input, sends it directly to the completions API, and returns the response as the output.
:::code language="java" source="~/functions-openai-extension/samples/textcompletion/java/src/main/java/com/azfs/TextCompletions.java" range="52-65":::
::: zone-end
::: zone pivot="programming-language-javascript"
This example demonstrates the templating pattern, where the HTTP trigger function takes a name parameter and embeds it into a text prompt, which is then sent to the Azure OpenAI completions API by the extension. The response to the prompt is returned in the HTTP response.
:::code language="javascript" source="~/functions-openai-extension/samples/textcompletion/javascript/src/functions/whois.js" :::
::: zone-end
::: zone pivot="programming-language-typescript"
This example demonstrates the templating pattern, where the HTTP trigger function takes a name parameter and embeds it into a text prompt, which is then sent to the Azure OpenAI completions API by the extension. The response to the prompt is returned in the HTTP response.
:::code language="typescript" source="~/functions-openai-extension/samples/textcompletion/typescript/src/functions/whois.ts" :::
::: zone-end
::: zone pivot="programming-language-powershell"
This example demonstrates the templating pattern, where the HTTP trigger function takes a name parameter and embeds it into a text prompt, which is then sent to the Azure OpenAI completions API by the extension. The response to the prompt is returned in the HTTP response.
Here's the function.json file for TextCompletionResponse:
:::code language="json" source="~/functions-openai-extension/samples/textcompletion/powershell/WhoIs/function.json" :::
For more information about function.json file properties, see the Configuration section.
The code simply returns the text from the completion API as the response:
:::code language="powershell" source="~/functions-openai-extension/samples/textcompletion/powershell/WhoIs/run.ps1" :::
::: zone-end
::: zone pivot="programming-language-python"
This example demonstrates the templating pattern, where the HTTP trigger function takes a name parameter and embeds it into a text prompt, which is then sent to the Azure OpenAI completions API by the extension. The response to the prompt is returned in the HTTP response.
:::code language="python" source="~/functions-openai-extension/samples/textcompletion/python/function_app.py" range="7-11" :::
This example takes a prompt as input, sends it directly to the completions API, and returns the response as the output.
:::code language="python" source="~/functions-openai-extension/samples/textcompletion/python/function_app.py" range="14-18" :::
::: zone-end
::: zone pivot="programming-language-csharp"
The specific attribute you apply to define a text completion input binding depends on your C# process mode.
In the isolated worker model, apply TextCompletionInput to define a text completion input binding.
In the in-process model, apply TextCompletion to define a text completion input binding.
The attribute supports these parameters:
| Parameter | Description |
|---|---|
| Prompt | Gets or sets the prompt to generate completions for, encoded as a string. |
| Model | Gets or sets the ID of the model to use as a string, with a default value of gpt-3.5-turbo. |
| Temperature | Optional. Gets or sets the sampling temperature to use, as a string between 0 and 2. Higher values (0.8) make the output more random, while lower values like (0.2) make output more focused and deterministic. You should use either Temperature or TopP, but not both. |
| TopP | Optional. Gets or sets an alternative to sampling with temperature, called nucleus sampling, as a string. In this sampling method, the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. You should use either Temperature or TopP, but not both. |
| MaxTokens | Optional. Gets or sets the maximum number of tokens to generate in the completion, as a string with a default of 100. The token count of your prompt plus max_tokens can't exceed the model's context length. Most models have a context length of 2,048 tokens (except for the newest models, which support 4096). |
::: zone-end ::: zone pivot="programming-language-java"
The TextCompletion annotation enables you to define a text completion input binding, which supports these parameters:
| Element | Description |
|---|---|
| name | Gets or sets the name of the input binding. |
| prompt | Gets or sets the prompt to generate completions for, encoded as a string. |
| model | Gets or sets the ID of the model to use as a string, with a default value of gpt-3.5-turbo. |
| temperature | Optional. Gets or sets the sampling temperature to use, as a string between 0 and 2. Higher values (0.8) make the output more random, while lower values like (0.2) make output more focused and deterministic. You should use either Temperature or TopP, but not both. |
| topP | Optional. Gets or sets an alternative to sampling with temperature, called nucleus sampling, as a string. In this sampling method, the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. You should use either Temperature or TopP, but not both. |
| maxTokens | Optional. Gets or sets the maximum number of tokens to generate in the completion, as a string with a default of 100. The token count of your prompt plus max_tokens can't exceed the model's context length. Most models have a context length of 2,048 tokens (except for the newest models, which support 4096). |
::: zone-end
::: zone pivot="programming-language-python"
During the preview, define the input binding as a generic_input_binding binding of type textCompletion, which supports these parameters:
| Parameter | Description |
|---|---|
| arg_name | The name of the variable that represents the binding parameter. |
| prompt | Gets or sets the prompt to generate completions for, encoded as a string. |
| model | Gets or sets the ID of the model to use as a string, with a default value of gpt-3.5-turbo. |
| temperature | Optional. Gets or sets the sampling temperature to use, as a string between 0 and 2. Higher values (0.8) make the output more random, while lower values like (0.2) make output more focused and deterministic. You should use either Temperature or TopP, but not both. |
| top_p | Optional. Gets or sets an alternative to sampling with temperature, called nucleus sampling, as a string. In this sampling method, the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. You should use either Temperature or TopP, but not both. |
| max_tokens | Optional. Gets or sets the maximum number of tokens to generate in the completion, as a string with a default of 100. The token count of your prompt plus max_tokens can't exceed the model's context length. Most models have a context length of 2,048 tokens (except for the newest models, which support 4096). |
::: zone-end
::: zone pivot="programming-language-powershell"
The binding supports these configuration properties that you set in the function.json file.
| Property | Description |
|---|---|
| type | Must be textCompletion. |
| direction | Must be in. |
| name | The name of the input binding. |
| prompt | Gets or sets the prompt to generate completions for, encoded as a string. |
| model | Gets or sets the ID of the model to use as a string, with a default value of gpt-3.5-turbo. |
| temperature | Optional. Gets or sets the sampling temperature to use, as a string between 0 and 2. Higher values (0.8) make the output more random, while lower values like (0.2) make output more focused and deterministic. You should use either Temperature or TopP, but not both. |
| topP | Optional. Gets or sets an alternative to sampling with temperature, called nucleus sampling, as a string. In this sampling method, the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. You should use either Temperature or TopP, but not both. |
| maxTokens | Optional. Gets or sets the maximum number of tokens to generate in the completion, as a string with a default of 100. The token count of your prompt plus max_tokens can't exceed the model's context length. Most models have a context length of 2,048 tokens (except for the newest models, which support 4096). |
::: zone-end
::: zone pivot="programming-language-javascript,programming-language-typescript"
The binding supports these properties, which are defined in your code:
| Property | Description |
|---|---|
| prompt | Gets or sets the prompt to generate completions for, encoded as a string. |
| model | Gets or sets the ID of the model to use as a string, with a default value of gpt-3.5-turbo. |
| temperature | Optional. Gets or sets the sampling temperature to use, as a string between 0 and 2. Higher values (0.8) make the output more random, while lower values like (0.2) make output more focused and deterministic. You should use either Temperature or TopP, but not both. |
| topP | Optional. Gets or sets an alternative to sampling with temperature, called nucleus sampling, as a string. In this sampling method, the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. You should use either Temperature or TopP, but not both. |
| maxTokens | Optional. Gets or sets the maximum number of tokens to generate in the completion, as a string with a default of 100. The token count of your prompt plus max_tokens can't exceed the model's context length. Most models have a context length of 2,048 tokens (except for the newest models, which support 4096). |
::: zone-end
See the Example section for complete examples.
