| # Instructions Directory | |
| This directory contains all agent instructions used by the Data Extractor application in JSON format. | |
| ## Structure | |
| ``` | |
| instructions/ | |
| βββ README.md (this file) | |
| βββ agents/ | |
| βββ data_extractor.json # Data extraction agent instructions | |
| βββ data_arranger.json # Data organization agent instructions | |
| βββ code_generator.json # Excel code generation agent instructions | |
| ``` | |
| ## JSON Format | |
| Each instruction file follows this structure: | |
| ```json | |
| { | |
| "instructions": [ | |
| "First instruction line", | |
| "Second instruction line", | |
| "..." | |
| ], | |
| "agent_type": "data_extractor|data_arranger|code_generator", | |
| "description": "Brief description of the agent's role", | |
| "category": "agents or other category" | |
| } | |
| ``` | |
| ## Benefits of JSON Format | |
| 1. **Structure**: Clean separation of instructions as array elements | |
| 2. **Metadata**: Includes agent type and description for context | |
| 3. **No Conversion**: Direct use as lists - no need to split strings | |
| 4. **Maintainability**: Easy to add, remove, or reorder instructions | |
| 5. **Validation**: JSON schema validation possible | |
| ## Usage | |
| ```python | |
| from utils.prompt_loader import prompt_loader | |
| # Load as list for agent initialization | |
| instructions_list = prompt_loader.load_instructions_as_list("agents/data_extractor") | |
| # Load as string for other uses | |
| instructions_text = prompt_loader.load_instruction("agents/data_extractor") | |
| ``` |