Advanced Settings
Extended AI agent settings for fine-tuning behavior, memory, and working mode.
Where to Find
Open the agent editor
Go to the "Advanced" tab
AI Model
Select the language model for the agent.
Google Gemini 2.5 Flash
Fast, economical (default)
OpenAI GPT-4o
High response quality
OpenAI GPT-4.1
Advanced capabilities
Claude Sonnet 4
Excellent logic and analysis
Response Language
Language Auto-detection
The agent detects the customer's language and responds in the same language.
Example:
Customer writes in English → Agent responds in English
Customer writes in Spanish → Agent responds in Spanish
Fixed Language
If auto-detection is disabled, select a language for responses from the list.
Working Schedule
Configure hours when the agent will respond to customers.
Enabling Schedule
Enable the "Working Schedule" toggle
For each day of the week:
Enable/disable the day
Set start and end time
Configuration Example
Monday
09:00 - 18:00
Tuesday
09:00 - 18:00
Wednesday
09:00 - 18:00
Thursday
09:00 - 18:00
Friday
09:00 - 18:00
Saturday
Disabled
Sunday
Disabled
What happens outside schedule? The agent doesn't respond to messages. They will wait for a manager's response or the next working period.
Time Zone
The schedule operates according to the time zone configured in your Constanta profile.
To change the time zone:
Click on your name in the upper right corner
Select "Profile"
Change "Time Zone"
Click "Save"
Response Delay
Set a pause before sending the response to simulate typing.
0 seconds
Instant response
30-60 seconds
Realistic delay (recommended)
Up to 120 sec
Maximum delay
Recommendation: 45-60 seconds creates a feeling that a real person is responding.
Memory and Context
Long-term memory settings for the agent.
Long-term Memory
The agent remembers facts about the customer between sessions.
Examples of remembered facts:
Customer name
Company
Products of interest
Previous agreements
Special requests
How does it work? AI extracts important facts from the dialogue and saves them. On the next contact, the agent "remembers" the customer.
Relationship Graph
The agent understands connections between entities (people, companies, products).
Example:
"John works at Acme Inc, interested in product X"
The agent will connect John with the company and product.
Context Window
How many facts from memory to use when generating the response.
5-10
Minimum, fast responses
20-30
Optimal (recommended)
40-50
Maximum context
Important! Large context window slows down responses and increases token consumption.
Semantic Search
Use AI to search for relevant information in the knowledge base.
Disabled
Keyword search
Enabled
AI understands query meaning (recommended)
Internal AI Models
Models for agent's internal tasks.
By default, all tasks use one model (recommended GPT-4o-mini for cost savings).
If needed, you can configure separate models for each task:
Fact Extraction
Extracting important information
Trigger Evaluation
Determining customer intent
Message Generation
Creating responses
Email Generation
Creating emails
Instruction Parsing
Understanding system prompt
KB Analysis
Working with knowledge base
Tip: Use fast models (GPT-4o-mini) for internal tasks, and powerful ones (GPT-4o) for main responses.
Need Help?
Telegram: @constanta_support_bot
Email: [email protected]
Last updated
Was this helpful?