Recordable Chat With Models

Unlike standard language models that process plain text, Chat Models take chat messages (think back-and-forth dialogue) as input and generate replies. LangChain supports a wide variety of Chat Model providers, including OpenAI and Hugging Face.

A static message is appended to each prompt sent to the model. This helps to establish a consistent context for responses. Also used to define other logic, format, or requirements you want the model to apply to its output.

Recorded chats

Recordable chats allow you to capture the audio and video of conversations with advanced voice capabilities. The chat history will also display a visual indicator to show which conversations have been recorded.

When a record and a chat are connected, an optional introduction note can be created using selected data fields of the record. This helps you share succinct, read-only context with your collaborators. Your collaborators who have access to a record and a license can view the full details of the record in a browser tab.

The following parameters help you customize the appearance and behavior of the Recordable Chat with Models component. name: A user-friendly name for the model spec, shown in the header dropdown menu. default: Specifies if the model spec is the default selection, to be auto-selected on every new chat.

Recorded messages

Recorded messages are independent of any model object. The messages are displayed in the Messages tab which is conversation based. A full list is also displayed when a radio user group in the Chat menu is selected. They are not displayed in the 'Recently Discussed Elements' report, which is element-based.

Le Chat reflects and responds faster than other chat assistants (up to 1000 words/sec). The Mistral models with the lowest latency combine pre-trained information from the most recent information balanced between robust search, social media, and news with high accuracy for text, audio, or visual content.

Records Chat uses private artificial intelligence, which means that your conversations remain within the boundaries of Appian Cloud and are protected against unauthorized access from any outside party.

Recorded responses

Le Chat is powered by the Mistral models that are the fastest and most reliable chat assistants. Its image and document understanding uses the best vision and optical character recognition (OCR) models in the industry.

When a participant answers a question with a generative AI model, the answer it provides may contain "hallucinations." For example, the model could return different information when answering the same question at different times.

The Recorded Responses Page displays all responses for your project except those that are in progress or those used to populate Embedded Data, such as Org Hierarchy Questions. You can find information in the Recorded responses help article about viewing, searching for, exporting, and deleting recorded answers.

You may want to use only a subset (or subsets) of the recorded responses for some analyses. You might, for example, want to only see responses from a certain geographic region or those provided by single mothers. Filtering responses is a help article that can be used to create filters to meet your needs.

Recorded conversations

In the chat history, you can see a transcription of your voice conversations along with audio and video clips. This makes it easy to find and play back your advanced voice conversations. A visual indicator in your chat history will help you identify which conversations were recorded using advanced voice mode.

While recording calls and conversations is an important tool for journalists, you must be aware of your state's laws regarding the consent required to record a call or conversation. There are a wide range of states that have different laws, and in some cases, federal law may preempt a state's laws.

You can use the custom tip view (on the left) to display a customized recording tip. To change the recording tip view, select your desired option from the Settings menu.