​
Overview

The Chat Interface is where you interact directly with your AI agents. More than a simple conversation, it’s a complete tool that allows you to:
  • 💬 Chat with any created agent
  • 🔍 Track executions in real-time (debug)
  • 📁 Manage sessions with intelligent organization
  • 🎤 Send audio, text, and media for rich interaction
  • 🏷️ Organize conversations with tags and filters
Real-Time Debug: Even in complex agents (Sequential, Workflow, Loop), you see each execution step, function calls, and returns in real-time, facilitating debugging and optimization.

​
Step 1: Accessing Chat

  1. In the main Evo AI menu, click “Chat”
  2. You’ll see an interface divided into two parts:
    • Left panel: List of all chat sessions
    • Main area: Active chat or initial screen
  3. To start, you can create a new chat or open an existing session
Chat Sessions List
The interface automatically saves all conversations, allowing you to resume any session at any time.

​
Step 2: Managing Sessions

Left panel - All your conversations:Each session shows:
  • Session name (editable)
  • Agent used
  • Last message or summary
  • Conversation date
  • Tags for organization
List functionalities:
  • 🔍 Quick search by name, description, or content
  • 🏷️ Filter by tags to organize by project/type
  • 🤖 Filter by agent to see specific conversations
  • 📅 Sort by date, name, or recent activity
  • ✅ Multiple selection for batch actions
How to select multiple sessions:
  1. Check the checkboxes next to each desired session
  2. Action bar appears at the top when there’s selection
  3. Select all: General checkbox to mark all visible
Selected SessionsAvailable batch actions:🗑️ Delete Selected:
  • Removes multiple sessions at once
  • Confirms before permanently deleting
  • Useful for periodic cleanup
🏷️ Add Tags:
  • Applies tags to multiple sessions
  • Organizes conversations by project/client
  • Facilitates future filters
📁 Move to Folder: (if available)
  • Organizes into categories
  • Separates by client, project, type
💡 Tip: Use multiple selection to organize old sessions, apply tags in batch, or clean up test conversations.

​
Step 3: Creating New Session

How to create a new conversation:
  1. Click “New Chat” in the left panel
  2. Select the agent you want to use in the conversation
  3. Configure session name (optional, can change later)
  4. Add tags for organization (optional)
  5. Click “Create” to start
Creating New ChatInitial settings:
Session Name: "Customer Service John - Notebooks"
Agent: sales_assistant_tech
Tags: client-john, sales, notebook, 2024

Result: Organized chat easy to find later
Best practices for names:
  • Use consistent pattern: “Type - Client - Subject”
  • Include date if relevant: “Support-15Jan-LoginProblem”
  • Be descriptive: “Demo-Product-Client-ABC”
  • Avoid generic names: “Chat 1”, “Test”

​
Step 4: Editing Session Information

Accessing edit options:
  1. In sessions list: Click the “Edit” icon (✏️) next to the session
  2. Or inside chat: Session options menu
  3. Edit form opens with editable fields
Editing Chat DataEditable fields:Session Name:
  • Change to something more descriptive
  • Useful when conversation evolved to different subject
  • Helps in future searches
Description:
  • Summary of what was discussed
  • Important conversation details
  • Results or next steps
Tags:
  • Add new tags as conversation evolves
  • Remove irrelevant tags
  • Use status tags: “resolved”, “pending”, “follow-up”
Agent: (if allowed)
  • Switch agent during conversation
  • Useful to escalate to specialist
Organization example:
Before: "Chat 1"
After: "Client John - Gaming Notebook Purchase"

Tags: client-john, sales, notebook, gaming, budget-5k
Description: "Client interested in gaming notebook up to $5,000. 
           Showed interest in model X. Awaiting proposal."

​
Chatting with Agents

​
Input Types

Most common standard input:
  • Type your message in the text box
  • Press Enter to send
  • Shift + Enter for line break
  • Supports long text and simple formatting
Tips for better results:
✅ Be specific: "Need notebook for graphic design, budget $4000"
❌ Too generic: "I want a computer"

✅ Provide context: "I'm a photographer and need to edit 50MB RAW files"
❌ No context: "What's better?"

✅ Include preferences: "Prefer national brands with good warranty"
❌ No direction: "Help me choose"
Attach files and images:Supported formats:
  • 🖼️ Images: PNG, JPG, JPEG, GIF, WebP
  • 📄 Documents: PDF, DOC, DOCX, TXT
  • 📊 Spreadsheets: XLS, XLSX, CSV
  • 🎵 Audio: MP3, WAV, M4A, OGG
How to send:
  1. Click the “Attach” icon (📎)
  2. Select file from your computer
  3. Or drag and drop directly into chat
  4. Wait for upload and processing
  5. Add explanatory text if needed
Use cases:
📷 Product photo with defect for support
📄 Contract for legal analysis
📊 Data spreadsheet for analysis
🖼️ Design for feedback
Models that support media:
  • GPT-4 Vision: Image analysis
  • Gemini Pro Vision: Complete multimodal
  • Claude 3: Documents and images
Voice functionality:How to record:
  1. Click the “Microphone” icon (🎤)
  2. Allow access to microphone if requested
  3. Speak your message - indicator shows active recording
  4. Click again to stop recording
  5. Audio is sent automatically
⚠️ Important - For audio to work:Option 1: Model that accepts audio
Use native models with audio support:
- Whisper (OpenAI) - specialized in transcription
- Gemini Pro - multimodal with audio
- Provider-specific models
Option 2: Speech to Text Tool
Configure speech_to_text tool in agent:
1. Go to LLM agent settings
2. "Tools" section → Add "Speech to Text"  
3. Configure provider (OpenAI, Groq, Google)
4. Agent converts audio → text → processes
Audio quality:
  • Silent environment for better transcription
  • Speak clearly and at normal pace
  • Avoid background noise
  • Test with short phrases first
Use cases:
🎤 Report detailed problem by voice
🗣️ Make complex request by speaking
🎙️ Interview or feedback collection
📞 Simulate phone service

​
Debug and Real-Time Tracking

Chat as debug tool:The chat interface isn’t just for talking - it’s a complete debug tool that shows everything happening “behind the scenes” of agent execution.What you see in real-time:
  • 🔄 Each execution step in Sequential agents
  • 🌐 Calls to APIs and external tools
  • 🤖 LLM decisions and reasoning
  • ⚙️ Workflow execution step by step
  • 🔁 Loop Agent iterations
  • 📞 Function calls and returns
  • ⏱️ Execution times of each stage
Following sequential execution:Example: Sales agent with 3 sub-agents
🚀 Starting execution...

📍 Step 1: Lead Qualifier
├── Analyzing customer needs...
├── Identifying budget: $5,000
├── Detecting urgency: Medium
└── ✅ Qualification completed

📍 Step 2: Product Demonstrator  
├── Consulting product database...
├── Filtering by budget $5,000
├── Selecting 3 suitable options
└── ✅ Demonstration completed

📍 Step 3: Closing Specialist
├── Analyzing qualification data
├── Creating personalized proposal  
├── Adding sense of urgency
└── ✅ Final proposal generated

🎯 Execution completed in 8.2s
Visible information:
  • Status of each sub-agent
  • Time spent on each stage
  • Data passed between agents
  • Errors or failures (if any)
Following complex flows:Example: Proposal approval workflow
🔄 Workflow: Proposal Analysis

🟦 Node: Receive Data
├── Input: Commercial proposal received
├── Validating format...
└── ✅ Valid data

🟦 Node: Reviewer Agent
├── Executing content analysis...
├── Checking values and terms...
├── Quality score: 8.5/10
└── ✅ Review approved

🔷 Node: Condition (Score > 8.0?)
├── Evaluating: 8.5 > 8.0
├── Result: TRUE
└── ➡️ Proceeding to approval

🟦 Node: Automatic Approval
├── Generating final contract...
├── Applying templates...
└── ✅ Contract generated

⏸️ Node: Delay (5 seconds)
├── Waiting for processing...
└── ✅ Delay completed

💬 Node: Final Message
└── ✅ "Proposal approved and contract sent!"
Details shown:
  • Type of each node (Agent, Condition, Message, Delay)
  • Data flowing between nodes
  • Conditional decisions
  • Delay times
  • Errors at any point
Following iterations:Example: Content refinement loop
🔁 Loop Agent: Content Refiner

=== ITERATION 1 ===
📝 Generator: Creating initial content
├── Based on user request
├── Generating first version...
└── ✅ Content v1 created

📊 Analyzer: Evaluating quality
├── Clarity score: 6.5/10
├── Persuasion score: 7.0/10  
├── Overall score: 6.8/10
└── ❌ Below threshold (8.0)

=== ITERATION 2 ===
📝 Generator: Refining content
├── Applying analysis feedback
├── Improving clarity...
└── ✅ Content v2 created

📊 Analyzer: Re-evaluating
├── Clarity score: 8.2/10
├── Persuasion score: 8.5/10
├── Overall score: 8.4/10
└── ✅ Above threshold!

🛑 Checker: Using exit_loop()
└── ✅ Loop finished

📋 Finalizer: Generating final response
└── ✅ Result consolidated
Complete visibility:
  • Current iteration number
  • Progress of each sub-agent
  • Quality metrics
  • Stop conditions
  • Exact moment of exit_loop
Monitoring tools and APIs:Example: Agent using external tools
🔧 Function Call: load_knowledge
├── Parameters: {"query": "notebook warranty", "tags": ["support"]}
├── Searching knowledge base...
├── Result: 3 chunks found
└── ✅ Knowledge loaded

🔧 Function Call: get_product_info  
├── Parameters: {"product_id": "NB001", "include_price": true}
├── Consulting products API...
├── Status: 200 OK
├── Data: {"name": "Notebook Pro", "price": 4500, "stock": 12}
└── ✅ Product found

🔧 Function Call: send_email
├── Parameters: {"to": "client@email.com", "subject": "Proposal"}
├── Connecting to SMTP server...
├── Status: Email sent
└── ✅ Communication completed
Detailed information:
  • Name of called function
  • Sent parameters
  • Execution status
  • Returned data
  • Response time
  • Errors (if any)
Debugging advantages:
  • Identifies performance bottlenecks
  • Detects integration failures
  • Validates sent parameters
  • Confirms expected returns

​
Advanced Session Management

​
Organization and Productivity

Organization strategies:By client:
client-john, client-company-abc, client-vip
→ Facilitates finding specific conversations
By service type:
sales, support, onboarding, demo, follow-up
→ Organizes by activity category
By status:
pending, resolved, awaiting-client, escalated
→ Controls service workflow
By product/service:
notebook, smartphone, software, consulting
→ Segments by business area
By priority:
urgent, high, medium, low
→ Prioritizes service
Useful combinations:
Example: "client-vip + sales + notebook + urgent"
→ VIP client with urgent notebook demand
Extracting insights from conversations:Service patterns:
  • Which agents are most used?
  • What types of questions are most common?
  • What time has highest activity?
Agent performance:
  • Conversations that generated sales
  • Successfully resolved problems
  • Average resolution time
Client evolution:
  • Complete interaction history
  • Preferences identified over time
  • Products or services of interest
Opportunities:
  • Clients with unexplored potential
  • Recurring problems to improve
  • Agents that need adjustments

​
Best Practices

For better results:Preparation:
  • Choose the right agent for each conversation type
  • Configure descriptive name and tags from the start
  • Have clear context of what you want to achieve
During conversation:
  • Be specific and provide adequate context
  • Use media when appropriate (images, documents)
  • Follow debug to understand reasoning
  • Ask follow-up questions to deepen
After conversation:
  • Update tags based on what was discussed
  • Add description summarizing results
  • Mark status (resolved, pending, etc.)
  • Document next steps if necessary
Important precautions:Sensitive information:
  • Avoid sharing unnecessary personal data
  • Use codes or initials when possible
  • Consider privacy implications
Secure organization:
  • Use tags that don’t expose confidential information
  • Be judicious with session names
  • Periodically clean old conversations
  • Configure appropriate data retention
Backup and continuity:
  • Export important conversations if necessary
  • Document valuable insights externally
  • Maintain history of important decisions
Sharing knowledge:Standardization:
  • Use consistent naming conventions
  • Define tag patterns for the team
  • Document organization best practices
Collaboration:
  • Share relevant sessions when appropriate
  • Use status tags for coordination
  • Document learnings for reuse
Training:
  • Use example conversations to train new users
  • Analyze successful conversations to replicate
  • Identify success patterns

​
Troubleshooting

1. Audio doesn’t work:
  • ✅ Check microphone permissions in browser
  • ✅ Configure speech_to_text tool in agent if model doesn’t support native audio
  • ✅ Test with short phrase first
  • ✅ Check if speech_to_text provider is configured
2. Media upload fails:
  • ✅ Check file size (limits by type)
  • ✅ Confirm supported format
  • ✅ Test internet connection
  • ✅ Try smaller file to test
3. Agent doesn’t respond as expected:
  • ✅ Use debug functionality to understand execution
  • ✅ Check if agent is configured correctly
  • ✅ Be more specific in question
  • ✅ Provide more context about what you need
4. Search doesn’t find conversations:
  • ✅ Try different keywords
  • ✅ Check active filters
  • ✅ Use tags instead of text search
  • ✅ Adjust date period if applicable
5. Slow performance:
  • ✅ Limit number of loaded sessions
  • ✅ Use filters to reduce list
  • ✅ Clean up old sessions
  • ✅ Check internet connection

​
Next Steps

Configure Agents

Create and configure agents for your conversations

Knowledge Base

Configure knowledge base for more precise agents

Workflow Agents

Understand debugging of complex workflows

Settings

Configure tools and integrations
💬 Now you master the chat interface! Use these features to have productive conversations, debug complex agents, and organize your history efficiently!