> For the complete documentation index, see [llms.txt](https://ai-docs.fptcloud.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://ai-docs.fptcloud.com/~/revisions/GkZVtBK6yRi7YYdjkSnZ/ai-marketplace/ai-inference/tutorials/playground.md).
# Playground
### How to get started with Playground?
The AI Marketplace Playground lets you test and interact with AI models without coding — including text generation, text-to-speech, speech-to-text, vision tasks, and embeddings.
* On the homepage, click **“Playground”** in the top navigation bar (between Pricing and Support Center).
* Click **Get API Key** to go to the My API Key page, where you can create or manage your API keys.
* Click **View Code** to open a pop-up showing sample code snippets (available in **cURL**, **Python**, and **Node.js**) along with an example response for your request.
***
### 1. Large Language Model (LLM)
**Description:**\
A text-based assistant trained on a huge amount of information (books, websites, articles, conversations). It can understand your request and reply with natural, human-like text.
**Quick Action Templates Available:**
* Create a blog outline
* Create a go-to-market plan
* Summarize this report in bullet points
* Help me prepare talking points for my meeting
* Professionalize my email
* Write a to-do list for a personal project
**Steps:**
1. Click on the task that best fits your needs. The system will automatically prepare a response based on your input.
2. Type your question, topic, or specific request related to the selected task into the text box.
3. Click **Send** → The result will appear on the right.
**Advanced Settings (left panel):**
* **System prompt** – set style/behavior of the assistant.
* **Temperature**, **Max output tokens**, **Top P**, **Top K**, **Penalties** – fine-tune creativity, length, and randomness.
***
### 2. Text to Speech Model (TTS)
**Description:**\
Converts written text into spoken audio.
**Steps:**
1. Select the type of voice in the **Voice** dropdown.
2. Select the speed of the voice in the **Speed** dropdown.
3. Select the language in the **Language** dropdown.
4. Enter up to 500 characters in **Text** input.
5. Click **Run** → The system will generate the audio file, which can be played or downloaded.
***
### 3. Speech to Text Model (STT)
**Description:**\
Converts spoken audio into written text.
**Steps:**
1. Select the **Model** from the dropdown.
2. Choose **Response format** (e.g., json or text).
3. Set the **Source language**.
4. Select the **Task**:
* **Transcription** – converts speech to text.
* **Translation** – converts speech to text in another language.
5. Upload or record audio (supported formats: **MP3/WAV**, max length **15s**).
6. Click **Run** → The transcribed or translated text will appear in the Result section.
***
### 4. Vision Language Model (VLM)
**Description:**\
An assistant that combines computer vision (understanding images) and natural language processing (understanding text) to make sense of the world in a more human-like way.
**Steps:**
1. Upload your image (**JPEG**, **PNG**, **JPG**, max 2 files, 20 MB each).
2. Ask a question about your image in the text box.
3. Click **Run** → The model will respond based on the image content.
***
### 5. Embedding Model
**Description:**\
Embedding models are transformer-based neural networks that transform chunks of documents (passages of text) into a numeric representation or vector.\
These vectors capture the semantic meaning of text and are used for tasks like search, recommendation, and similarity comparison.
**Steps:**
1. Select the type of input you want to test in the **Input type** dropdown.
2. Select the desired output format for the embedding in the **Encoding format** dropdown.
3. Choose how to handle input that exceeds the token limit in the **Truncate** dropdown.
4. Enter up to 500 characters in **Text** input.
5. Click **Run** → The system will display the embedding vector (numerical representation) in the Result section.
**Note:**\
Guardrail feature is coming soon to enhance content safety.
***
### 6. Rerank Model
#### **Description**
Rerank models are transformer-based neural networks designed to **reorder a list of candidate passages** based on their **relevance to a given query**.\
Unlike embedding models that output vector representations, rerank models directly produce a **relevance score** for each passage relative to the query.\
They are typically used in **retrieval pipelines**, **semantic search**, and **question-answering systems** to refine initial search results and improve ranking accuracy.
***
#### **Steps**
**1. Select the model**
From the **Model** dropdown, choose a rerank model such as `bge-reranker-v2-m3`.
**2. Enter your query**
In the **Query** field, type the question or information request that defines what you want to find.\
Example: *“What is FPT Smart Cloud and what are its main business areas?”*
**3. Add passages**
Input one or more text passages in the **Passages** section.\
Each passage represents a potential answer or piece of content to be evaluated against the query.
**4. Run the model**
Click **Run** → The model computes a **logit score** for each passage, indicating how relevant it is to the given query.
**5. View the output**
In the **Output** panel:
* Each passage is displayed with a **logit score**.
* Passages are automatically **sorted by relevance**, with higher scores indicating stronger semantic alignment with the query.
***
### 7. How to use Extra Controls in the Playground?
* **System prompt**: Guide the model’s behaviour and style.
* **Enable streaming**: Show partial results as they are generated.
* **Stop sequences**: Cut off the model’s response at specific words or phrases.
* **Fine-tune this model**: Link to start customizing the model for your own use case.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://ai-docs.fptcloud.com/~/revisions/GkZVtBK6yRi7YYdjkSnZ/ai-marketplace/ai-inference/tutorials/playground.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.