Chat is the best way to get cited, detailed answers to your internal questions. For most queries, Butler will use sources as context. Simple queries such as follow-ons will not. There are three filters/toggleables:
  1. Sources
    • Pick which apps Butler will search for relevant sources
    • By default, searches will be performed on all data from connected apps
  2. Model
    • The LLM used to provide the response
  3. AI Only toggle
    • Allows the chat to be used like a regular LLM (no app context used)
  4. Web Search toggle
    • Sets Model to Perplexity Sonar to enable web search
  5. AI Only toggle
    • Forces Butler answer without fetching data from connectors; allows the chat to be used as a regular LLM

All available models are listed below:
ModelContext Window
GPT-4o128k tokens
GPT-4o Mini128k tokens
Claude 3.5 Sonnet200k tokens
Claude 3.0 Opus200k tokens
Perplexity Sonar32k tokens
DeepSeek Chat64k tokens

​
Optimization

Tips for getting better responses:
  1. Use relevant keywords: Include specific terms related to your query. For example, instead of “how do I deploy”, try “what is our AWS deployment process for staging”.
  2. Be specific: Frame questions as complete sentences rather than keywords. Example: “What were the key decisions from the Q1 product roadmap meeting?” instead of “roadmap updates”.
  3. Specify source: Mention which apps or documents to search. Example: “Find the security guidelines document in Notion about API authentication”.
  4. Request format: Clarify the type of response you want - whether it’s a summary, link, or detailed explanation.
Think of Butler as a real person - the more context you provide, the better it will perform.