Chat & Conversations
The Conversation mode is the core of Kognar Platform. It's a full-featured AI chat interface supporting multi-turn dialogue, file attachments, rich content rendering, and real-time tool use.
Sending messages
Type your message in the input bar at the bottom of the screen and press Enter to send. Use Shift+Enter to add a new line without sending.
While the assistant is generating a response, the Send button turns into a Stop button (red). Click it to cancel the response at any time.
Streaming responses
Responses appear in real time as the model generates them — character by character. You'll see a blinking cursor at the end of the current output while the model is still writing.
Message actions
On the last assistant message, two action buttons appear:
- Retry — Regenerates the response using the same message
- Export — Downloads the response as a
.md(Markdown) file
Context window meter
The header shows a context meter — a progress bar indicating how much of the model's context window is currently in use. As conversations grow, this meter fills up.
When the context is nearly full, Kognar will warn you before sending. You can choose to compress the conversation (reduces size while preserving meaning) or cancel.
Starting a new conversation
Click + in the sidebar or press Ctrl/Cmd+N. Conversations are automatically titled with a timestamp and can be renamed at any time.
Quick start prompts
When a conversation is empty, a set of prompt buttons appears in the center of the screen. Click any of them to pre-fill a commonly used starting message:
| Prompt | Pre-filled message |
|---|---|
| Code | Review code for readability and performance |
| Research | Search the web for current information on a topic |
| Analysis | Analyze and summarize the main points of a text |
| Summarize | Summarize the following text objectively |
| Writing | Help draft a professional email |
Response metadata
Below each assistant message, Kognar shows:
- P: — Prompt tokens used
- C: — Completion tokens generated
- Latency — Response time in milliseconds
Markdown rendering
Assistant responses render full GitHub-flavored Markdown:
- Headings, bold, italic, strikethrough
- Ordered and unordered lists
- Code blocks with syntax highlighting
- Tables with borders and alternating rows
- Blockquotes
- Inline links (open in a new browser tab)
Tool execution
When an agent uses a tool (like web search or code analysis), a tool card appears in the response thread. It shows:
- The tool name and status (running / complete)
- The result returned by the tool
- Which agent invoked the tool
Tool cards are collapsible to save space.
Voice input
Click the microphone icon in the chat input bar to dictate your message instead of typing:
- Click the mic icon — recording starts immediately
- Speak your message
- Stop recording — the text is transcribed and placed in the input field
- Review and press Enter to send
See Audio Transcription for details on on-device vs. cloud transcription modes.
Web search sources
When the Web Search agent is active and returns results, a sources accordion appears below the response:
- The header shows how many sources were consulted (e.g. "4 sources")
- Click the header to expand and see the full list
- Each source shows the page title and URL
- Long URLs are truncated on screen — hover to see the full address
- Click a source to open it in your browser
If no sources were found, the accordion does not appear.
Conversation context menu
Right-click any conversation in the sidebar to access:
| Option | Description |
|---|---|
| Edit title | Rename the conversation |
| Delete | Remove the conversation permanently |
| Details | View creation date, last update, model used, and estimated token usage |
Sound notifications
Kognar can play a sound when the assistant finishes responding. Enable or disable this in Settings > General > Preferences > Play sounds.