Dashtech

Raghvi first saw the problem while doing a consulting case in logistics: companies operate under strict SLAs, where delays or breakdowns trigger huge fines. Yet drivers are often randomly assigned unfamiliar trucks; hence, this is the vehicle type most prone to breakdowns.

Jake, with family in vehicle parts manufacturing, learned that while mechanics use diagnostic computers, the tech is outdated, lacks reasoning, and is unusable by non-mechanics. For trucks breaking down far from service centers, this leaves drivers stranded and fines mounting.

Dashtech is a local AI agent built for the truck dashboard:

Diagnostics: Tap Run diagnostics to generate a prioritised checklist tailored to the truck’s make, model, and configuration Free-form Q&A: Ask questions, and Dashtech reasons through likely causes, points to the right manual section, and suggests safe next steps Safety-first design: Hands-free voice input and spoken guidance, with large, high-contrast UI for glanceable use

Dashtech was built with a Tauri and Rust backend for speed, and a React/TypeScript + Tailwind frontend for a clean, responsive UI. This wraps a GPT-oss-20b model running locally, with manuals from major truck brands loaded via RAG for precise retrieval. The pipeline involves polling signals from the truck, detecting issues, running the reasoning models, and then generating instructions through a structured workflow. We prototyped UI layouts quickly, then built both frontend and backend ourselves, integrating AI reasoning with live truck data feeds

While building, we ran into a few challenges. Firstly, different makes and models of truck have different codes and procedures. We solved this with a manual ingestion and RAG pipeline. Secondly, designing an interface that works hands-free, with minimal distraction, took several iterations, and trialling different voice input and output tools that didn’t require permissions. Thirdly, running a large model offline required optimising memory and response time, since drivers can’t depend on connectivity.

We’re proud of the fact that Dashtech is easy for truckers to use, and hands-free, given that they likely have no technical knowledge and would want to use the tool anywhere and in any situation. We also built a full end-to-end product in just a few weeks - from data processing and multi-step AI analysis to a polished desktop UI and PDF export. This was Raghvi’s first time taking a project all the way through the engineering stack, and our first virtual hackathon together, collaborating seamlessly to combine our experience and expertise.

Through this process we learnt how to structure manuals and error codes into RAG pipelines that return actionable instructions. We also learnt the importance of designing for users with limited connectivity and high-pressure conditions.

In the near term, Dashtech runs on a laptop or portable device connected to the truck. The longer-term vision is to embed it directly into vehicle dashboards. That would require close collaboration with manufacturers to integrate with existing systems. Running gpt-oss would need dedicated GPU hardware. Beyond trucks, the same workflow can extend to other commercial vehicles such as buses, vans, and heavy machinery, and eventually into predictive maintenance and fleet management integrations.

Built With

• python
• rag-pipeline-for-manuals-**other**:-voice-i/o
• rust
• tailwind
• tailwind-**ai**:-gpt-oss-20b-running-locally-with-ollama
• tauri
• typescript