🍜From Bland to Brilliant AI Agents: New Agent Features for April 2024!

All the incredible additions to the best agent builder anywhere - Gooey.AI's AI Agent.

The TL;DR of our latest in Copilot

1. All the best LLMs! LLaMA3, Claude3, Mistral, Gemini Pro 1.5 & Gemma support - Use any of the best LLM (& 3 vision models!) in Copilot including GPT4 Turbo, GPT Vision and LLaMA2.

2. WhatsApp Self Serve - Get yourself a spare SIM and connect your AI agent.

3. Fast Speech Recognition in 1000+ languages- We now run Meta's MMS-Large speech recognition model "hot" meaning you can transcribe virtually any language in near real-time.

4. OpenAI + Azure Voices - great for realistic and African accents!

5. Transparent Pricing - 3 credit (~$.03) / message + your LLM, TTS and speech recognition token costs.

6. <1s RAG with 1000s of docs, pdfs and video - Add 1000s of PDFs, files or youtube videos to your knowledge base and we'll query them in milliseconds (after the first run).

7. Built-Analytics and Conversation Analysis - Dashboard and analytics for monitoring AI Agent interactions, retention, 👍🏾 👎🏽 feedback and messages categorization.

8. Bring Your Own Agent Evaluation - Get yourself a few golden questions and answers and then test how changes to your agent's LLM, instructions or knowledge docs improve its answer quality.

9. Auto-embed Documents + Compare Vectors Feature - Just add a PDF or doc to agent, and we'll visually understand its contents and vectorize it. Works for YouTube links too.

10. Built-in Synthetic Data Maker - Every video or PDF really needs a FAQ to help your agent better answer questions. Our synthetic data maker has you covered.

And the long version…

We're thrilled to announce huge strides with Gooey.AI's Copilot. Here is a low down of what we have been up to:

1. LLaMA3, Claude3, Mistral, Gemma and Gemini Pro 1.5 and GPT-4 Turbo with Vision

Our AI Agent now harnesses the best of every major LLM from opensource (Mistral Mixtral, Meta's LLaMA3, Google's Gemma) to private (OpenAI's GPT4-Turbo Vision, Anthropic's Claude3 family and Google's Gemini 1.5 Pro). Compare vision support in Claude3, GPT4 V vs Gemini Pro 1.0.

These updates allow you to choose from the best LLMs from the industry that suit your needs on cost, speed, reasoning and accuracy.

Here's a sample evaluation of the competing LLMs using Farmer.CHAT's golden questions and answers (in this case, the first one - GPT4 Turbo + Vision - still beats Mixtral, Gemma and Gemini Pro 1).

Screenshot from our Bulk and Evaluation tool, showing the difference in quality of answers from the copilot.

2. WhatsApp Self Serve

Due to popular demand, you can now integrate your own Business WhatsApp Phone Number with the AI Agent! Bring a spare number that you’d like to connect with your agent so that you can use any global number.

Check out our refreshed Integrations Tab to know more.

In the Integrations settings, you can also turn on 👍🏾 👎🏽 feedback, streaming and send broadcast messages too.

3. Fast, Massively Multilingual Speech

Adding to our arsenal of multi-lingual goodness we now support Meta's latest Massively Multilingual Speech model (along with the best of Google, OpenAI, Azure and more). This latest release from Meta can identify more than 4,000 spoken languages, 40 times more than any model. And we run it hot on our A100s so you can actually use it in a real-time chat setting.

Try it here: https://gooey.ai/speech/massively-multilingual-speech-hindi/?example_id=bxfzzimvbr99

Plus GhanaNLP translation support

Which offers high-end and accurate context-based translation for African languages.

4. Support for OpenAI Voices & Global Accents with Azure TTS

For those implementing high-scale multilingual agents in Africa, India and South America, we now offer global accent support for English allowing better communication and information to a wide range of users. We also have several languages and dialects including Hindi, Bhojpuri, Swahili, and Chichewa hosted through various open source models. This means that you can not only have agents that understand and respond in local languages or global accents but can also do lipsync!

Check out our copilot for Opportunity.org which uses Tanzania English Accent to help African farmers.

5. Transparent Pricing

We are rolling out transparent pricing starting at just 3 Credits (~$.03) + your usage of LLM, TTS, and speech recognition services. You'll have complete visibility into your costs, allowing you to plan and budget with confidence.

6. Faster RAG with VespaDB

We've integrated VespaDB to significantly enhance the AI Agent's RAG performance. This powerful database technology accelerates data retrieval and analysis, making your interactions with the Agent more efficient and responsive than ever before. We have customers with 1000+ PDFs that can now be searched in under 1 second (vs 16 seconds before).

7. Advanced Analytics with Meta-analysis

After all the building and testing is up and running, you need to make sure your analytics are in place! We’ve got you covered there too.

Once you’ve added integration to your agent, you’ll gain access to an amazing dashboard of usage data, conversations, and feedback.

8. Bring Your Own Copilot Evaluation

To understand which LLM works best for you, check out our Bulk Runner & Evaluator. All you need to do is create about 5-10 samples of “Golden Questions and Answers” that are closest to your agent’s user needs, add them to the workflow, and create a run with a difference. Here’s a Farmer.CHAT example that you can run and modify. Visit our Copilot Evaluation guide too.

9. Auto-embed documents + settings to Compare Vectors Feature

You can now automatically create vector embeddings for any document or link you add to Agent. But if you want to get into the details, you can choose among the latest and best embeddings from OpenAI and other open-source embeddings.

10. Built-in Synthetic Data Maker

Sometimes your knowledge base needs some extra love for your agent to answer users well. By creating Synthetic Data, you provide vector embeddings to retrieve texts closest to the users' questions and answer effectively.

Say goodbye to data scarcity with our built-in synthetic data maker for PDFs, Excel sheets, Google folders, and YouTube. This important addition generates high-quality, realistic synthetic data, empowering you to test, train, and develop applications with ease.

What’s coming in May

1. Web-widget support - Add a Gooey.AI Agent on your webpage with a single line of Javascript code.

2. Function Calling - call out to your servers or any data source inside an agent run.

Resources

Start building your own Agent

Check our Agent Guide

Book a Demo

Gooey's goal is to make the best of open-source and private AI accessible to everyone. We can't wait to see what you'll build with these tools so hopefully everyone on earth can benefit from the incredible innovation of this moment in time.

Cheers,

Gooey.AI Team