🎉2025 Gooey.AI Copilot Update

Gooey.AI’s Copilot is used by thousands of global frontline workers - including HVAC technicians, teachers, farmers, and nurses. Here are our latest features designed for organizations to empower their frontline and measure how well AI extends their mission.

TL;DR

Long Version

Support for all the latest models

We have added support for Deepseek, OpenAI o3, Gemini 2, GPT-4o, GPT-4o-mini, LLAMA 3.3, o1 preview, and o1 preview mini and smaller local fine-tuned models like Sarvam 2B, SEA-Lion, and AfroLLAMA.

Functions and Secrets

  • Gooey.AI Functions adds the power of JavaScript to any of your AI Workflows through

  • Now you can store SECRETS in your API Keys tab in the Accounts Page. This will allow you to use external services with sandboxed Functions. [add examples links]

  • Two amazing updates from this update include LLM-based code writing and execution and Web Search.

Web search & Code writing and execution

  • In this example, we use Google search via API which calls two searches:

    • The length of the Pont d’Arc

    • The speed of a cheetah

  • Based on the question and the information from Google, the LLM creates the Javascript function and runs it in our sandbox.

https://gooey.ai/copilot/base-copilot-w-search-rag-code-execution-v1xm6uhp/

WhatsApp improvements

We’ve implemented a number of improvements to our Whatsapp integration. Including:

  • Follow up questions as Button - now everyone can support perplexity like experience on WhatsApp, with auto-suggested follow-up questions. A huge win for user engagement!

  • Voice note support

  • Real-time web search

Screenshot of Voice Support with Web search
Screenshot of Auto Suggested Follow up buttons
  • Citation support

  • Thumbs Up / Down Support - Incredibly useful to help your team identify where your Copilot is not meeting user expectations.

Screenshot of Citation Support
Screenshot of Citation Support with Feedback Button
  • Location - Copilots can request the user’s location and our integration will show the native WhatsApp UI to allow the user to share their location

  • Inline Images

  • Document uploads - upload a PDF or CSV and will scan or parse it and enable the user to read it

Cache

  • Now you can set your Cache Settings for your knowledge base. Cached knowledge means our responses are under 5 seconds even with thousands of knowledge base PDFs, Google Docs, or PPTs

  • If your copilot has documents that need to be regularly updated keep “Always Check for Updates” on. However, this will come at the cost of increased latency

  • By default, we don’t keep this setting on. Find it under Settings > Cache

You can see our test run for with and without Caching.

https://gooey.ai/bulk/caching-test-gooey-bot-bulk-run-hgpr7muqblun/

Lower is better in the chart above

Wider support for knowledge-base documents

  • In addition to supporting basic documents like PDF, .txt, and CSV, we offer support for links from Google Drive (including entire folders!).

  • We’ve also added PPTX support for those in the MS Office Suite ecosystem!

  • OneDrive support comes very soon!

ASR support (for broadest language support)

We now offer ASR support for 16 models. The MMS model alone can identify 4000 languages.

  1. Whisper Large v2

  2. Whisper Large v2 Hindi fine-tuned

  3. Whisper Large v2 Telugu fine-tuned

  4. Whisper Large v3

  5. Whisper Large v3 Chichewa fine-tuned

  6. GPT-4o Audio

  7. Conformer Hindi (ai4bharat.org)

  8. Conformer English (ai4bharat.org)

  9. Vakyansh Bhojpuri (EkStep)

  10. Google Cloud v1

  11. Chirp / USM (Google Cloud v2)

  12. Deepgram

  13. Azure Speech

  14. Seamless M4T v2 (Meta AI)

  15. Massively Multilingual Speech (Meta AI)

  16. GhanaNLP ASR v2

Web widget

Don’t miss our shiny new “Sources” preview in the Gooey.AI Copilot Web Widget. Our web widget shows all this as web pages or embedded documents inside the Chat Widget. The Preview pane allows easy inspection of sources and increases the credibility of AI Copilot sources.

  • You can use the built-in Photo & Audio inputs, which allow users to share photo-based queries and voice notes.

  • Link as a stand-alone page or embed

Analysis

  • Integrate with your own analysis tools: Push copilot data to external sources in real-time (via POST functions) or on a daily schedule.

  • Create conversations into structured insights with customized analysis prompts that can analyze every conversation in real-time.

Templating

With Jinja-based templating, the AI Copilot can customize the prompt based on specified use cases. You can also learn more about templating from our guide.

Last updated

Was this helpful?