Chat and Search

Search and Chat form Kapinote's dual-engine knowledge retrieval system: Local Search retrieves deterministic facts from your archive; Meeting Chat leverages your active AI provider to synthesize, reason, and compare across selected conversations.

1. Search your local meeting archive

Use the global search bar on the home screen. Kapinote matches your keywords across three kinds of content in your local SQLite database:

• Meeting titles and topics
• Structured AI summaries, key decisions, and action items
• Time-stamped, verbatim transcript segments

Search Tip: Use distinctive keywords such as customer names, project codenames, metrics, or technical terms. Selecting a search result opens the meeting directly and jumps to the exact spoken quote.

2. Chat with a single meeting

Open any meeting and use the right-side chat panel to ask targeted questions:

• "What decision was finalized regarding our target launch date?"
• "List all action items with owners and deadlines. Mark missing assignees as 'Unassigned'."
• "Which recurring objections did the customer emphasize?"
• "Quote the exact spoken dialogue when pricing was negotiated."

Evidence Grounding: Every AI response cites specific transcript timestamps. Click any citation to inspect the spoken source.

3. Chat across multiple meetings

When your question spans several syncs, user interviews, or weekly status updates, open the global chat panel:

• Filter Scope: Select only the relevant meetings to minimize context noise.
• Reference Shortcut (@): Type @ in the composer to search and attach specific past meetings as context.
• Template Shortcut (/): Type / to select multi-meeting analysis templates (e.g., /insights, /risks, /brief).

Example Prompt:
"Review these three customer feedback calls. Group recurring pain points and rank them by customer urgency."

4. Privacy and resource usage

• Search: Runs 100% locally on your Mac's storage with zero network requests.
• Chat: Transmits the selected meeting context and your prompt to the active AI provider. Cloud providers consume your configured API quota; local Ollama models process everything strictly on-device. See Privacy and model choices.