Surface Duo Blog

Test the latest AI features in Kotlin

November 23, 2023 Nov 23, 2023 11/23/23
Craig Dunn
Hello prompt engineers, Last week we looked at one of the new OpenAI features – Assistants – in the web playground, but good news: the OpenAI Kotlin library is already being updated with the new APIs and you can start to try them out right now in your Android codebase with snapshot package builds. With a few minor configuration ...

OpenAI Assistants

November 18, 2023 Nov 18, 2023 11/18/23
Craig Dunn
Hello prompt engineers, OpenAI held their first Dev Day on November 6th, which included a number of new product announcements, including GPT-4 Turbo with 128K context, function calling updates, JSON mode, improvements to GPT-3.5 Turbo, the Assistant API, DALL*E 3, text-to-speech, and more. This post will focus just on the Assistant ...

Chunking for citations in a document chat

November 12, 2023 Nov 12, 2023 11/12/23
Craig Dunn
Hello prompt engineers, Last week’s blog introduced a simple “chat over documents” Android implementation, using some example content from this Azure demo. However, if you take a look at the Azure sample, the output is not only summarized from the input PDFs, but it’s also able to cite which document the answer is drawn from...

Document chat with OpenAI on Android

November 3, 2023 Nov 3, 2023 11/3/23
Craig Dunn
Hello prompt engineers, In last week’s discussion on improving embedding efficiency, we mentioned the concept of “chunking”. Chunking is the process of breaking up a longer document (ie. too big to fit under a model’s token limit) into smaller pieces of text, which will be used to generate embeddings for vector similarity ...

More efficient embeddings

October 29, 2023 Oct 29, 2023 10/29/23
Craig Dunn
Hello prompt engineers, I’ve been reading about how to improve the process of reasoning over long documents by optimizing the chunking process (how to break up the text into pieces) and then summarizing before creating embeddings to achieve better responses. In this blog post we’ll try to apply that philosophy to the Jetchat ...

Responsible AI and content safety

October 22, 2023 Oct 22, 2023 10/22/23
Craig Dunn
Hello prompt engineers, This week we’re taking a break from code samples to highlight the general availability of Azure AI Content Safety. In this blog series we’ve touched briefly on the using prompt engineering to restrict the types of responses an LLM will provide, such as setting the system prompt to set boundaries on what ...

“Search the web” for up-to-date OpenAI chat responses

October 12, 2023 Oct 12, 2023 10/12/23
Craig Dunn
Hello prompt engineers, Over the course of this blog series, we have investigated different ways of augmenting the information available to an LLM when answering user queries, such as: However, there is still a challenge getting the model to answer with up-to-date “general information” (for example, if...

Android tokenizer for OpenAI

October 5, 2023 Oct 5, 2023 10/5/23
Craig Dunn
Hello prompt engineers, The past few weeks we’ve been extending JetchatAI’s sliding window which manages the size of the chat API calls to stay under the model’s token limit. The code we’ve written so far has used a VERY rough estimate for determining the number of tokens being used in our LLM requests: This very ...

Speech-to-speech conversing with OpenAI on Android

September 28, 2023 Sep 28, 2023 09/28/23
authorsCraig Dunn, Kristen Halper
Hello prompt engineers, Just this week, OpenAI announced that their chat app and website can now ‘hear and speak’. In a huge coincidence (originally inspired by this Azure OpenAI speech to speech doc), we’ve added similar functionality to our Jetpack Compose LLM chat sample based on Jetchat. The screenshot below shows ...

Infinite chat with history embeddings

September 21, 2023 Sep 21, 2023 09/21/23
Craig Dunn
Hello prompt engineers, The last few posts have been about the different ways to create an ‘infinite chat’, where the conversation between the user and an LLM model is not limited by the token size limit and as much historical context as possible can be used to answer future queries. We previously covered: ...