Search & Chat · June 16, 2026

Why context matters in AI conversations

AI conversations without turn memory are frustratingly limited: you need to repeat context with every message, reference “as I said before” and lose the thread of investigation.

CleverNote keeps the last 8 conversation exchanges active, allowing natural follow-up questions without repetition.

How multi-turn works

The system uses a stateless approach: the client (web app or mobile) resends recent history with each new message. From the server’s perspective, each request has complete necessary context. From the user’s perspective, the experience is a continuous conversation.

This means you can:

Refine a question:

You: What expenses did I have for services in May?
CleverNote: [lists results]
You: Of those, which were over $200?

Dig deeper into an answer:

You: Who is the plumber who serviced me?
CleverNote: "John Smith, (555) 123-4567, referred by your neighbor Peter"
You: What dates did he come?
CleverNote: [searches the plumber note for service dates]

Compare periods:

You: How much did I spend on food in March?
CleverNote: $1,240.00 (12 notes)
You: What about April?
CleverNote: $1,390.00 (14 notes), 12% increase vs March

Investigate an anomaly:

You: What was my biggest expense in February?
CleverNote: $890.00, Souza Auto Shop, Feb 14
You: Do I have more expenses with that shop?
CleverNote: [searches by entity "Souza Auto Shop"]

Limits to know

8-turn context: long conversations that exceed 8 exchanges lose the oldest messages from context. For long investigations, it helps to summarize conclusions before continuing.

No persistence between sessions: closing the app or browser ends the conversation. There’s no saved chat history on the server. If you want to keep something important from a conversation, save it as a note.

Each question triggers new search: the system doesn’t assume previous searches covered everything. Each turn may surface additional results if the question has evolved.

Tip: use context for financial investigations

Multi-turn shines in financial investigations:

1. “What’s my total spending this month?”
2. “Which category represents the most?”
3. “What were the merchants in that category?”
4. “Are there any recurring expenses I could reduce?”

Each question refines the previous one without needing to re-contextualize.

See also: Chat with citations, Entity-scoped chat, Finance dashboard