Chat Tips & Best Practices

How Your Questions Get Processed

When you ask Omi a question, it goes through an intelligent routing system that determines the best way to answer:

Understanding these paths helps you frame questions that get the richest, most relevant answers.

Simple Chat
Agentic (Context-Rich)
Persona

For: Greetings, general advice, brainstormingQuestions like “Hi!”, “What’s a good book to read?”, or “Tips for productivity” get fast, direct responses without searching your memories.Examples:

“Hello!”
“What’s a good productivity tip?”
“Help me brainstorm ideas for…”

Prompting Strategies

Be Specific with Time

Omi excels at temporal queries. The system converts relative time references to precise date ranges.

Use natural time expressions - the system understands them perfectly.

Great examples:

“What did I discuss yesterday?”
“Summarize my last 3 days”
“What happened this morning?”
“Conversations from last week”
“What did I talk about on Monday?”

Less effective:

Vague timeframes without anchors (“a while ago”)
Very broad ranges (“everything from this year”)

Use Natural Language for Topics

Omi uses semantic search - it finds conversations by meaning, not just keywords. Ask naturally!Great examples:

“What have I discussed about my health?”
“Conversations about career growth”
“Times I talked about feeling stressed”
“Discussions involving my manager”
“What have I said about the project launch?”

Why it works: The system embeds your question and finds conversations with similar meaning, even if they use different words.

Ask About Yourself

Omi builds a knowledge base of facts about you across all conversations. Tap into it!Great examples:

“What do you know about me?”
“What are my goals?”
“Who do I talk to most?”
“What are my preferences?”
“What hobbies have I mentioned?”

Why it works: This triggers the memories system, which stores structured facts extracted from your conversations.

Leverage Your Integrations

If you’ve connected services, ask about them directly.

Google Calendar

“What meetings do I have today?”
“When is my next meeting with Sarah?”
“Schedule a meeting with John at 3 PM tomorrow”

Gmail

“Show my recent emails about the project”
“Emails from John this week”

Whoop (Health)

“How was my sleep last night?”
“What’s my recovery score today?”
“Show my workouts this week”

GitHub

“My open pull requests”
“Issues assigned to me”

Example Questions by Category

Category	Great Questions	Why They Work
Time-Based	”What did I do yesterday?”, “Summarize my week”	Clear time bounds trigger precise date filtering
Topic Search	”What have I said about AI?”, “My health discussions”	Semantic search finds conceptually related conversations
People	”Conversations with John”, “What did Sarah mention?”	People filter in search + vector similarity
Action Items	”What tasks do I have?”, “What do I need to do?”	Direct access to action item system
Personal Facts	”What’s my favorite food?”, “Where do I work?”	Memory retrieval of stored facts
Analysis	”How productive was I this month?”, “Patterns in my week”	Multi-tool orchestration for comprehensive answers
Calendar	”My meetings tomorrow”, “When am I free this week?”	Calendar integration for schedule awareness

Pro Tips

These advanced strategies are based on how the system is designed internally.

Combine Time + Topic

Narrow results by specifying both when and what:

“What did I discuss about the project last week?”
“Conversations about health this month”
“What did John say yesterday?”

Ask Follow-ups

The system remembers your last 10 messages:

First: “What did I discuss yesterday?”
Follow-up: “Tell me more about the second one”
Follow-up: “What action items came from that?”

Be Direct

Skip pleasantries when you need information:

Better: “My tasks for today”
Slower: “Hey, could you maybe show me…”

Use Relative Dates

Relative dates work better than absolute:

Better: “yesterday”, “last 3 days”, “this morning”
Slower: “January 15th, 2024”

Request Summaries

Trigger comprehensive retrieval:

“Summarize my week”
“Overview of yesterday’s conversations”
“Key themes from last month”

Ask for Specific Counts

When you want thorough results:

“Show me my last 10 conversations”
“What are my top 5 priorities?”

What NOT to Ask

Some question patterns don’t work well due to system design.

Avoid	Why	Better Alternative
”Tell me everything”	Too broad, hits context limits	”Summarize my last week"
"Search all my data”	No specific intent to optimize for	”What topics do I discuss most?”
Extremely old queries	Vectors may not exist for old conversations	Focus on recent conversations
Breaking persona character	Confuses persona apps	Stay in character with persona apps
Questions about other users	Privacy boundaries	Focus on your own data

Integration-Specific Tips

Google Calendar
Gmail
Whoop
GitHub
Notion

Best practices for calendar queries:

Action	Example Query
Check schedule	”What meetings do I have today?” or “Am I free at 3 PM?”
Find meetings	”When is my next meeting with [name]?”
Create events	”Schedule a meeting with John tomorrow at 2 PM about project review”
Modify events	”Move my 3 PM meeting to 4 PM”

Best practices for email queries:

Action	Example Query
Search by sender	”Emails from John this week”
Search by topic	”Recent emails about the contract”
Combine filters	”Emails from Sarah about the project in the last 3 days”

Best practices for health queries:

Data Type	Example Query
Sleep	”How was my sleep last night?” or “Sleep quality this week”
Recovery	”What’s my recovery score today?”
Workouts	”Show my workouts this week” or “How much strain yesterday?”
Trends	”How has my sleep been this month?”

Best practices for development queries:

Action	Example Query
PRs	”My open pull requests” or “PRs I need to review”
Issues	”Issues assigned to me” or “Open issues in [repo]“
Create	”Create an issue for the login bug in [repo]”

Action	Example Query
Search	”Search my notes about [topic]“
Find pages	”Find my page about product roadmap”

Understanding Citations

When Omi answers questions using your conversations, it includes numbered citations like [1] and [2]. These link to the source conversations.

Tap citations in the app to view the full conversation
Multiple citations mean information came from multiple conversations
No citations means the answer came from general knowledge or your stored memories (facts about you)

Quick Reference Card

Best Question Patterns:

Time-based:    "What did I [verb] [time period]?"
Topic-based:   "What have I discussed about [topic]?"
Person-based:  "Conversations with [name]" or "What did [name] say?"
Action items:  "What tasks do I have?" or "My to-dos"
Personal:      "What do you know about my [aspect]?"
Calendar:      "My meetings [time period]"
Summary:       "Summarize my [time period]"

Power Combinations:

Time + Topic:   "What did I discuss about [topic] [time period]?"
Time + Person:  "Conversations with [name] [time period]"
Topic + Person: "What did [name] say about [topic]?"
All three:      "What did [name] and I discuss about [topic] [time period]?"

Chat System Architecture

Technical deep dive into how the chat system works

Building Chat Tools

Create custom tools for your Omi apps

Storing Conversations

How your conversations and memories are stored

Real-time Transcription

How live transcription works under the hood

Get Started

Core Development

DIY Guide

Info

Chat Tips & Best Practices

How Your Questions Get Processed

Prompting Strategies

Google Calendar

Gmail

Whoop (Health)

GitHub

Example Questions by Category

Pro Tips

Combine Time + Topic

Ask Follow-ups

Be Direct

Use Relative Dates

Request Summaries

Ask for Specific Counts

What NOT to Ask

Integration-Specific Tips

Understanding Citations

Quick Reference Card

Best Question Patterns:

Power Combinations:

Chat System Architecture

Building Chat Tools

Storing Conversations

Real-time Transcription

​How Your Questions Get Processed

​Prompting Strategies

Google Calendar

Gmail

Whoop (Health)

GitHub

​Example Questions by Category

​Pro Tips

Combine Time + Topic

Ask Follow-ups

Be Direct

Use Relative Dates

Request Summaries

Ask for Specific Counts

​What NOT to Ask

​Integration-Specific Tips

​Understanding Citations

​Quick Reference Card

​Best Question Patterns:

​Power Combinations:

​Related Documentation

Chat System Architecture

Building Chat Tools

Storing Conversations

Real-time Transcription

How Your Questions Get Processed

Prompting Strategies

Example Questions by Category

Pro Tips

What NOT to Ask

Integration-Specific Tips

Understanding Citations

Quick Reference Card

Best Question Patterns:

Power Combinations:

Related Documentation