Core Concepts
How Viventium thinks—the key ideas behind the system.
Dual-Process Architecture
Viventium runs two layers simultaneously:
| Layer | What It Does | Speed |
|---|---|---|
| Main Agent | Responds to you in real-time | Instant |
| Background Agents | Analyze, research, verify in parallel | Seconds to minutes |
Why: Single-model AI must choose between fast or deep. Viventium gives you both. The main agent keeps conversation flowing while background agents handle heavy analysis.
Background Agents
Specialized processes that activate based on what you're discussing.
Each agent has:
- Trigger conditions – When it activates (e.g., overconfident statements, research requests)
- Specialized capability – What it's good at (e.g., fact-checking, emotional read)
- Non-blocking execution – Runs in parallel, never delays your response
Why: Different tasks need different approaches. A fast model handles dialogue; a reasoning model handles math; a research model handles deep dives. Viventium routes automatically.
See Background Agents for the full list.
Activation
How background agents decide to engage.
When you send a message, each background agent evaluates:
- Does this message match my trigger conditions?
- How confident am I that I should activate?
- Has enough time passed since my last activation? (cooldown)
If conditions are met, the agent runs.
Why: Not every message needs deep analysis. Activation filtering keeps the system efficient and avoids noise.
Brewing
What happens when background agents are processing.
You'll see "Thinking..." status indicators. This means:
- One or more background agents activated
- They're analyzing, researching, or reasoning
- Results will surface as follow-up insights
Why: Transparency. You know when deeper work is happening versus a simple response.
Insights
Follow-up realizations that appear after the main response.
Insights might include:
- A pattern noticed across your conversation
- A blind spot in your reasoning
- Research findings from multiple sources
- An emotional read you didn't ask for but needed
Why: The main response handles your immediate need. Insights add depth without blocking your flow.
Memory
How Viventium remembers context.
- Session memory – Full context of current conversation
- Long-term memory – Key facts, preferences, and context across sessions
- Natural recall – Viv references past context like a friend would
Why: Every conversation shouldn't start from zero. Memory makes Viv a thinking partner that knows your context.
Proactive Scheduling
Viventium can check in on your behalf.
Schedule recurring prompts:
- "Check my inbox every morning at 9am"
- "Remind me about my goals every Sunday"
- "Research competitor news daily"
Viv runs these autonomously and delivers results via chat or Telegram.
Why: An AI that only responds when prompted is limited. Proactive engagement keeps you on track without manual effort.
See Scheduling for setup details.
Multi-Surface Parity
Same brain across all interfaces.
| Interface | Best For |
|---|---|
| Chat | Deep work, rich formatting, visual feedback |
| Voice | Thinking out loud, hands-free, interruption-friendly |
| Telegram | Mobile access, quick check-ins, notifications |
All share the same:
- Background agents
- Memory
- Integrations
- Scheduling
Why: Your thinking partner shouldn't change based on how you access it.