How It Works
1
Command Understanding
Your natural language input is processed by OpenAI’s GPT-4 to understand intent and extract
parameters.
2
Tool Selection
The system automatically selects the appropriate tool(s) from the 10 available LangChain tools.
3
Execution
The selected tools execute the blockchain operations with proper error handling.
4
Response
Results are formatted in natural language for easy understanding.
Command Patterns
Identity Commands
Token Commands
Canister Commands
Cycles Commands
Multi-Step Operations
The NLP processor can handle complex, multi-step operations:Context-Aware Conversations
The system maintains context within a session:Using Specialized Agents
Different agents provide focused expertise:General Agent
Developer Agent
DeFi Agent
Advanced Patterns
Conditional Operations
Batch Operations
Queries and Filters
Error Handling
Natural language errors are user-friendly:Best Practices
Be Specific with Amounts
Be Specific with Amounts
Always include units: “5 ICP”, “2T cycles”, “1000 tokens”
Use Full Identifiers
Use Full Identifiers
For principals and canister IDs, use the complete identifier for accuracy.
Confirm Large Operations
Confirm Large Operations
The system will ask for confirmation on large transfers or irreversible operations.
Check Before Acting
Check Before Acting
Use query commands first: “Check balance” before “Transfer all my ICP”
Leverage Agent Expertise
Leverage Agent Expertise
Use specialized agents for domain-specific help and recommendations.
Conversation Examples
Example 1: Token Management Flow
Example 2: Canister Development Flow
Debugging Natural Language
Enable verbose mode to see tool selection:Language Tips
Common Phrases Reference
Intent | Example Phrases |
---|---|
Check Balance | ”What’s my balance?”, “Show ICP”, “How much do I have?” |
Transfer | ”Send X to Y”, “Transfer X ICP to Y”, “Pay Y with X ICP” |
Create | ”Make a new…”, “Create a…”, “Generate a…” |
List | ”Show all…”, “List my…”, “What are my…” |
Help | ”How do I…”, “What’s the best way to…”, “Help me…” |
Next Steps
Ready to build something more complex? Check out:- LangChain Examples for real-world scenarios
- Specialized Agents for agent-specific features
- Tools Reference for direct tool usage