AI Integrations
AI assistant for faster energy insights
Key Achievements

Client
Energy aspects is a leading global provider of real-time market insights and data analytics, providing comprehensive market intelligence across the entire energy complex.



i860gwuwq4tl3itcbrga0og8vb18vb-small.webp&w=3840&q=75)
Case study
The need
The customer needed a simpler way to consume data spread across multiple applications and a powerful search to link insights across research and time-series data. The challenge was to build an AI-ready search engine and reliably route each query to the right agent, whether data, documentation, or user-specific information.
Our solution
DIMI is the AI assistant for Energy Aspects that delivers precise answers and charts from research and data, and routes each request to the right module.
Search module
To provide the AI agent with the most relevant results, we defined specific characteristics for each type of content (research publications, time-series data, API documentation) and built a tailored agentic search module that accounted for those characteristics. To ensure its quality, we compiled an extensive evaluation set covering both common and advanced queries.
Routing module
Choosing the appropriate agent to handle a query is a crucial part of our agentic flow, which we developed and refined based on real user queries. To verify the correctness of routing, we compiled another extensive evaluation set with the help of domain experts and user feedback.
Key results
- Unified scattered data into one AI assistant, improving UX by letting customers consume content in one place
- Advanced search across research, time series, and API docs
- Tailored agentic search per content type
- Smart routing to the right agent for each query
- Filters by commodity and time for precise answers
- Instant charts with jump to Data Explorer
- License-aware access control that shows only permitted content
- Quality validated with evaluation sets and real user feedback, delivering over 90% positive response ratings
- Faster analysis and higher client satisfaction
Tech stack
- Vercel AI SDK - powers all AI agents
- Elasticsearch - powers the knowledge base used for search.
- UI - NextJS, TypeScript, React
- Backend - NodeJS, Vercel Fluid compute
- Models - OpenAI GPT-5 Chat, Anthropic Claude Sonnet 4.5
Team size
- 2 Full-Stack Developers


%20(1)k8vyter8d61g7mrpl3uip4fqsiew76-small.webp&w=3840&q=75)















