Ask about your screen or conversation, or ⌘⇧⏎ for Smart assist
Transcription history
Conversation history will appear here when listening.
Settings
Cerebras
OpenAI
Anthropic
Gemini
Deepgram
Groq
Custom
MiniMax
Base URL
Your Custom key and LLM requests are sent to this URL. To also use it for speech-to-text, pick Custom on the Audio tab — otherwise transcription uses OpenAI or Groq.
Not connected
Why it costs money
Ollama URL
DeepSeek
Azure AI Foundry
Endpoint
Fast
Smart
Vision
Local mode keeps microphone and meeting audio on this computer. It never silently falls back to a cloud transcription provider. Custom sends audio to the Base URL and Custom key set on the API Keys tab (requires both).
whisper.cpp runtimeChecking…
0%
Language
CPU threads
Use 0 for whisper.cpp's default. English-only models always use English regardless of the language setting.
Track slides
Check every (ms)
Captures a tiny hash every few seconds, captions only on stable change with your chat model. No images stored; clear with transcript.
0 / 2000
These rules tell SulyaraV2 how to write, not what to say. Use one imperative sentence per line (e.g. Never use em-dashes.). They apply to every mode except coding-problem solves (LeetCode), which stays strict. Stored locally; never sent anywhere except the AI provider when SulyaraV2 answers a question.
Stored locally and sent to your AI provider only as context when SulyaraV2 answers. Saved when you press Done.
0 / 600
Keep it short: role, company and key requirements. Only the first 600 characters are used (300 for technical questions).