Build Journal
Local Fetch Improvements & Tavus Fixes — September 4, 2026
I improved local untrackedFetch and fixed Tavus in-app frames for better user experience on the agentic web platform.
What shipped
- Local untrackedFetch feature — Enhanced functionality for smoother external data fetching.
- Tavus in-app frames unblocked — Resolved access issues for AISEO board and Calendly booking.
- Lucas live persona fix — Corrected usage of OWNED Interviewer persona for enhanced radio features.
Today, I set out to enhance the local untrackedFetch feature and address some issues with the Tavus integration on my agentic web platform. I shipped a feature improvement for the local fetch functionality and resolved a couple of bugs related to the radio tools and Tavus in-app frames. In the end, it was a productive day, but not without its challenges.
The primary focus was on the untrackedFetch feature for external calls. This feature is crucial for rendering paths that rely on data from outside our core system. I implemented a solution that allows for smoother data fetching without tracking every request, which should optimize performance and user experience. This enhancement is a direct response to feedback that indicated users were experiencing delays when accessing external content. I dedicated about an hour to this, and while I thought it would be straightforward, I hit a few snags along the way.
One of the first bugs I encountered was related to the noindex purge follow-ups from my work yesterday. I had to ensure that the new fetch functionality wouldn't inadvertently cause any indexing issues with our search engine optimization. I spent a solid 30 minutes double-checking that the purge was properly configured and that our local fetch wouldn't conflict with existing SEO settings. It's these little details that can easily slip through the cracks if not carefully monitored.
Next up was the Tavus integration. I had two specific fixes to implement today. The first was to unblock the in-app frames, specifically for the AISEO board and Calendly booking. Users were reporting that they couldn't access these features seamlessly within the Tavus environment, which was frustrating for them and ultimately detrimental to user retention. I spent about 15 minutes debugging the interaction configuration and grounding the Lucas radio-stage tools in the #SPACE setup. This fix should allow users to navigate between different functionalities without any hitches.
The second fix for Tavus was more about ensuring that the Lucas live feature utilized the correct persona. Previously, it was defaulting to a stock ID without the necessary tools, which limited the capabilities of the radio feature. I had to dive into the persona settings and make sure that Lucas was using the OWNED Interviewer persona, which has a comprehensive set of 28 tools. This required some careful adjustments and testing to ensure that it was functioning as intended. After a couple of attempts and a few tweaks, I was able to get it right.
Reflecting on the day, I am proud of the progress I made, especially considering the challenges I faced. The beauty of building this platform solo with the help of AI is that I can iterate quickly, but it also means that I have to be the one to troubleshoot every little issue that arises. There’s no team to lean on, which can be daunting, but it also means every victory feels all the more rewarding.
As I work towards my goal of creating a one-man-show company with a billion-dollar valuation, days like today remind me of the importance of persistence. Each feature I ship and each bug I fix brings me closer to a product that genuinely meets user needs. I can see that these improvements to the local fetch and Tavus integration will enhance the overall experience on the platform, and that’s what keeps me motivated.
I ended the day feeling accomplished but also aware of the continuous grind ahead. Building the agentic web platform is not just about shipping features; it's about creating an ecosystem where users can thrive. As I prepare for tomorrow, I’ll be looking at user feedback and analytics to see how these changes are received and what further adjustments might be necessary. The journey continues, and I’m excited for what’s next.