AI Agents
Advanced Agent Settings
Fine-tune your AI agent with advanced configuration options for optimal performance, accuracy, and cost efficiency.
Temperature
Range: 0.0 - 1.0 • Default: 0.7
Controls randomness and creativity in responses
- 0.0-0.3: Very consistent, factual responses (recommended for support)
- 0.4-0.7: Balanced creativity and consistency (default)
- 0.8-1.0: More creative and varied responses (use with caution)
- Lower values = more predictable, Higher values = more creative
- Adjust based on testing and user feedback
Max Tokens
Range: 50 - 2000 • Default: 500
Maximum length of AI responses in tokens (~4 chars per token)
- 50-200: Short, concise answers
- 200-500: Medium-length explanations (recommended)
- 500-1000: Detailed responses with examples
- 1000+: Very comprehensive answers
- Higher limits increase cost per response
Context Window
Range: 1 - 10 • Default: 5
Number of previous messages the agent remembers
- 1-2: Only remembers immediate previous message
- 3-5: Good for multi-turn conversations (recommended)
- 6-10: Extended conversation memory
- More context = better continuity but higher cost
- Reset context when topic changes significantly
Confidence Threshold
Range: 0.0 - 1.0 • Default: 0.7
Minimum confidence score before answering
- 0.5-0.6: Answer most questions (more responses, less accurate)
- 0.7-0.8: Balanced confidence (recommended)
- 0.9-1.0: Only highly confident answers (fewer responses, more accurate)
- Below threshold: Triggers escalation or "I don't know"
- Monitor analytics to find optimal value
Optimization Strategies
Advanced techniques to improve performance and reduce costs
Response Caching
Cache frequent questions for faster responses and lower costs
- Enable: Cache responses for identical questions
- Duration: 1 hour to 7 days
- Invalidate: Manual or automatic on knowledge base updates
Prompt Engineering
Craft effective system prompts for better performance
- Be specific about agent role and boundaries
- Include output format instructions
- Add examples of good responses
- Test and iterate based on results
Fallback Strategies
Handle edge cases when agent is unsure
- Custom "I don't know" messages
- Suggest related articles or FAQs
- Offer human escalation
- Collect question for knowledge base improvement
Rate Limiting
Control API usage and costs
- Max requests per user per minute
- Max tokens per conversation
- Daily budget limits
- Alert when approaching limits
Multi-tenant platform
Users create, manage, and deploy bots
Sign up, upload documents, customize widgets, monitor analytics, and handle escalations.
Have questions? Our community is here to help. Learn more about our community
100+
Active tenants
