Feature
AI Chat
Agent
Ask questions and get AI-generated insights grounded in real events. The chat agent uses RAG over the full event corpus to deliver verifiable answers with source citations.
From your question
to a grounded answer
Embed query
Your question is converted into a 768-dimensional vector using the same Gemini embedding model as the event corpus.
Retrieve context
The query vector is matched against events, raw events, and original sources using cosine similarity via pgvector.
Generate answer
The retrieved context is passed to your selected LLM, which generates a response grounded in real event data.
Cite sources
Every claim in the response links back to specific events and original articles for full verifiability.
Retrieval-augmented generation
Three layers of context
in every answer
The chat agent doesn't just search for keywords — it traverses the full event graph. Starting from your query, it retrieves unified events, drills down into the raw events that compose them, and traces back to the original source articles. This multi-layer retrieval ensures comprehensive, verifiable answers.
- Query embedding in the same 768-dim vector space
- Cosine similarity retrieval via pgvector
- Event → raw events → original sources traversal
- Context-aware prompt construction for the LLM
- Inline citations linking to specific sources
Provider agnostic
Choose your model
The same retrieval pipeline feeds into any LLM. Switch providers at any time via the model selector — the grounding and citations remain consistent regardless of which model generates the response.
Anthropic
Claude
Advanced reasoning and nuanced analysis of complex geopolitical scenarios.
OpenAI
GPT
Broad general knowledge with strong summarization and structured output.
Gemini
Multimodal understanding and fast response times for real-time queries.
xAI
Grok
Direct, concise answers with real-time awareness and minimal hedging.
DeepSeek
DeepSeek
Cost-effective inference with strong multilingual and reasoning capabilities.
What's the latest on EU sanctions and their market impact?
The EU adopted its 14th sanctions package on March 18, targeting energy exports. European energy stocks dropped 4.2% within hours...
Which sectors were hit hardest?
Energy and defense sectors saw the largest declines. Natural gas futures rose 6.1% as supply concerns...
Natural language
Ask anything about
what's happening
The chat agent supports multi-turn conversations with full context retention. Ask follow-up questions, request summaries of specific regions, compare events over time, or explore emerging trends — all through natural language. Each response is grounded in the live event corpus.
- Multi-turn conversations with context memory
- Follow-up questions that build on previous answers
- Region and topic-specific summaries on demand
- Trend exploration across time periods
- Conversation history saved and resumable
Capabilities
Everything that powers the AI chat agent.
Semantic retrieval
Questions are matched against the event corpus using vector similarity — finding relevant context by meaning, not just keyword overlap.
Source citations
Every answer includes inline citations linking back to original source articles, so you can verify and explore any claim.
Multi-provider AI
Choose between Anthropic, OpenAI, Google, xAI, or DeepSeek models. Switch at any time — the retrieval pipeline stays the same.
Event-aware context
The agent understands the full event graph — from unified events to raw events to original sources — providing deep contextual answers.
Real-time knowledge
Answers draw from the latest ingested and correlated data, providing insights grounded in current events rather than static training data.
Fact-check pipeline
A dedicated retrieval path cross-references claims against multiple source articles, surfacing supporting and contradicting evidence.
Ask, don't search
Get instant, sourced answers about global events — powered by your choice of AI model.