PromptOwl + ContextNest Docs

Conversations & Chat History

This guide explains how to manage conversations in PromptOwl, including chat history, memory settings, sharing, and feedback.


Table of Contents

  1. Understanding Conversations
  2. Conversation Sidebar
  3. Starting New Conversations
  4. Continuing Conversations
  5. Memory and Context
  6. Sharing Conversations
  7. Annotations and Feedback
  8. Managing Conversations
  9. Conversation Branching
  10. Best Practices
  11. Troubleshooting

Understanding Conversations

A conversation in PromptOwl is a chat session with an AI prompt. Each conversation:

  • Is tied to a specific prompt
  • Maintains message history
  • Can be named with a topic
  • Supports feedback and annotations
  • Can be shared with team members

Conversation Components

Component Description
Topic Auto-generated title from first message
Messages User questions and AI responses
Metadata Model used, timestamps, settings
Annotations Feedback and quality ratings
Citations Source references (if RAG enabled)

Conversation vs Prompt

Concept Purpose
Prompt The AI's instructions and configuration
Conversation An instance of chatting with that prompt

One prompt can have many conversations. Think of prompts as templates and conversations as sessions.


Conversation Sidebar

The sidebar shows your conversation history for the current prompt.

Accessing the Sidebar

  1. Open any prompt in Use mode
  2. Sidebar appears on the left
  3. Toggle with the hamburger menu if collapsed
Element Description
New Chat Start a fresh conversation
Search Bar Filter conversations by topic
Date Groups Today, Yesterday, Last Week, Last Month
Conversation List Click to switch conversations
Active Highlight Current conversation is highlighted

Date Grouping

Conversations are organized by recency:

  • Today: Conversations from today
  • Yesterday: From yesterday
  • Last Week: Within the past 7 days
  • Last Month: Within the past 30 days

Search Conversations

  1. Type in the search bar
  2. Results filter in real-time
  3. Matches against conversation topics
  4. Clear search to show all

Starting New Conversations

Creating a New Chat

  1. Click New Chat button in sidebar
  2. Or navigate to the prompt's use page fresh
  3. Type your first message
  4. Conversation is created automatically

Automatic Topic Generation

When you send your first message:

  1. AI generates a 3-word topic summary
  2. Topic appears in sidebar
  3. You can rename it later (hover and click edit)

Examples:

  • "Help with Python code" → "Python Code Help"
  • "What's the weather today?" → "Weather Query Today"

What Gets Saved

When a conversation starts, PromptOwl records:

  • The prompt being used
  • Your user account
  • Model and settings at the time
  • All messages exchanged
  • Timestamps for each message

Continuing Conversations

Resuming a Conversation

  1. Find the conversation in sidebar
  2. Click on it
  3. Full history loads in chat area
  4. Type new message to continue

Conversation Context

The AI remembers previous messages in the conversation:

What AI Remembers What AI Doesn't Remember
All messages in current chat Other conversations
Your questions and its responses Conversations with other prompts
Context from earlier in thread Information from before chat started

Long Conversations

For lengthy conversations:

  • Scroll up to see history
  • AI maintains context throughout
  • Very long threads may truncate context (model limits)
  • Consider starting fresh for new topics

Memory and Context

What is Memory?

Memories are persistent context snippets that get injected into your conversations automatically. Unlike conversation history (which is per-chat), memories persist across all conversations.

Memory Categories

Category Example Content
Preferences "I prefer concise answers"
Personal "My name is John"
Professional "I work in marketing"
Technical "I use Python 3.11"
Schedule "I'm in EST timezone"
Prompt Specific Context for specific prompt

Enabling Memory

  1. Open a prompt's settings
  2. Find Memory toggle
  3. Enable to include memories in conversations

Managing Memories

  1. Click the Memory tab in the chat sidebar
  2. View all your memories
  3. Add new memories
  4. Edit or delete existing ones
  5. Tag memories for organization

Memory Questions

Some memories include questions that prompt you for information:

  1. First time using a memory-enabled prompt
  2. Modal appears with questions
  3. Answer the questions
  4. Answers are stored and used in conversations

How Memory Affects Conversations

When memory is enabled:

User sends message →
System retrieves relevant memories →
Memories added to AI context →
AI responds with memory context

Sharing Conversations

Share valuable conversations with colleagues.

Sharing a Conversation

  1. Find conversation in sidebar
  2. Right-click or click menu icon
  3. Select Share
  4. Enter email addresses or select teams
  5. Click Share

Share Requirements

Requirement Description
Prompt Access Recipients need access to the prompt
Auto-Sharing Prompt is shared if recipient doesn't have it
Permission Only conversation owner can share

Shared Conversation Indicators

Indicator Meaning
"Shared by: email@example.com" Someone shared this with you
Share count badge Number of people with access

Viewing Shared Conversations

Shared conversations appear in your sidebar like your own, but show who shared them.


Annotations and Feedback

Provide feedback on AI responses to improve quality.

Response-Level Feedback

For individual AI responses:

  1. Find the response in chat
  2. Click thumbs up (good) or thumbs down (bad)
  3. Optionally add text feedback
  4. Feedback is saved immediately

Conversation-Level Feedback

For the overall conversation quality:

  1. Click Annotate Conversation button
  2. Select overall sentiment (good/bad)
  3. Add detailed feedback text (max 2000 chars)
  4. Click Save

Sentiment Options

Sentiment Icon When to Use
Good Thumbs up AI was helpful and accurate
Bad Thumbs down AI was unhelpful or wrong
Neutral No icon No strong opinion

Where Feedback Goes

Your annotations are used for:

  • Monitor Dashboard: Review feedback across conversations
  • Evaluation Sets: Create training datasets from feedback
  • Prompt Improvement: Identify patterns in negative feedback
  • Quality Metrics: Track satisfaction over time

For more on using feedback, see Monitoring, Evaluation, and Improvement.


Managing Conversations

Renaming Conversations

  1. Hover over conversation in sidebar
  2. Click the edit icon
  3. Enter new topic name
  4. Press Enter or click away to save

Deleting Conversations

  1. Hover over conversation in sidebar
  2. Click the trash icon
  3. Confirm deletion in dialog
  4. Conversation is removed

Note: Deletion is permanent and cannot be undone.

Conversation Metadata

Each conversation tracks:

Metadata Description
Created At When conversation started
Updated At Last message timestamp
Model AI model used
Provider LLM provider (OpenAI, Anthropic, etc.)
Prompt Version Which version of prompt was active

Conversation Branching

PromptOwl supports conversation branching, allowing different model responses to the same question.

What is Branching?

When you switch models mid-conversation:

  1. Same question can be answered by different models
  2. Each model's response creates a "branch"
  3. You can switch between branches to compare

Viewing Branches

  1. Look for model indicator on responses
  2. If multiple models answered, you can toggle between them
  3. Each branch maintains its own thread

When to Use Branching

  • Comparing model capabilities
  • Getting a second opinion from different AI
  • Testing prompt performance across models

Best Practices

Conversation Management

Do:

  • Use descriptive first messages (better auto-topics)
  • Delete conversations you no longer need
  • Add feedback to help improve the AI
  • Share valuable conversations with your team

Don't:

  • Keep hundreds of old conversations
  • Start new chats for follow-up questions
  • Forget to rename important conversations
  • Skip providing feedback on poor responses

Memory Usage

Do:

  • Add memories for frequently needed context
  • Use categories to organize memories
  • Update memories when information changes
  • Enable memory for prompts that benefit from personalization

Don't:

  • Add sensitive information to memories
  • Create duplicate memories
  • Enable memory if you need stateless interactions
  • Forget that memories affect all conversations

Feedback Best Practices

Do:

  • Provide specific feedback text, not just thumbs
  • Note what was wrong in bad responses
  • Highlight what was helpful in good responses
  • Review feedback patterns to improve prompts

Sharing Guidelines

  • Only share conversations that are helpful to others
  • Ensure recipients have appropriate prompt access
  • Consider privacy when sharing customer-related chats
  • Use team sharing for group access

Troubleshooting

Conversation Not Saving

  1. Check internet connection
  2. Refresh the page
  3. Verify you're logged in
  4. Try sending another message
  5. Check for error messages

Can't Find a Conversation

  1. Check all date groups (scroll down)
  2. Use search to filter by topic
  3. Verify you're in the correct prompt
  4. Check if conversation was shared (look in shared section)
  5. Conversation may have been deleted

Memory Not Working

  1. Verify memory is enabled on the prompt
  2. Check you have memories created
  3. Ensure memories aren't empty
  4. Try removing and re-enabling memory
  5. Memories may need relevant keywords to activate

Feedback Not Appearing

  1. Ensure you clicked save after annotating
  2. Refresh the page
  3. Check in Monitor dashboard for feedback
  4. Verify you have edit permissions
  5. Feedback may take a moment to sync

Shared Conversation Not Visible

  1. Ask sender to verify sharing
  2. Ensure you have prompt access
  3. Refresh your conversation list
  4. Check your email for correct sharing
  5. May take a moment to appear

Model Branching Issues

  1. Branches only work if same prompt used
  2. Ensure multiple models are enabled
  3. Check enterprise settings allow model switching
  4. Branches maintain separate context

Quick Reference

Keyboard Shortcuts

Action Shortcut
Send message Enter
New line Shift + Enter
New chat Click button (no shortcut)

Conversation Limits

Limit Value
Topic length ~50 characters
Annotation text 2,000 characters
Message length Varies by model
History retention Unlimited

Status Indicators

Indicator Meaning
Highlighted row Active conversation
"Shared by" text Someone shared this with you
Thumbs icon Feedback has been provided