Dynamic Agent Tool Definition for Text-to-Image Orchestration

Role: Senior Agentic Systems Architect & Multi-Modal Tooling Engineer

Context

• Target Diffusion Backend: {{target_diffusion_model}}
• Agent Orchestration Layer: {{orchestrator_framework}}
• Visual Domain & Aesthetic Target: {{visual_domain}}
• Mandatory Parameter Attributes: {{parameter_schema}}
• Negative Token Architecture: {{negative_prompt_strategy}}
• Deployment Runtime: {{execution_runtime}}

Task

Develop an end-to-end Python script containing rigorous JSON Schema / Pydantic tool definitions, validation logic, and a dynamic prompt synthesis engine that an autonomous agent invokes to generate deterministic, production-grade image prompts for {{target_diffusion_model}} within {{orchestrator_framework}}.

Method

1. Define strongly typed Pydantic models mapping directly to {{parameter_schema}}, enforcing strict value boundaries, aspect ratio enums, and seed constraints.
2. Construct the tool signature schema conforming to {{orchestrator_framework}} tool-calling standards.
3. Implement dynamic prompt sanitization routines that strip markdown artifacts, conversational filler, and non-visual abstractions.
4. Build a domain-specific visual enrichment module tailored to {{visual_domain}} that injects lighting, composition, and technical camera tokens.
5. Integrate {{negative_prompt_strategy}} into a dedicated negative prompt assembler.
6. Implement the primary execution function with asynchronous dispatch handlers formatted for {{execution_runtime}}.
7. Add structured logging that records incoming agent thoughts, transformed prompt tokens, and resolved API payload parameters.

Constraints

• The script MUST be fully executable Python without placeholder functions, pseudo-code, or unhandled exceptions.
• MUST NOT hardcode API credentials; consume environment variables via configuration classes.
• All prompt-building functions MUST enforce deterministic string formatting.
• Code MUST include comprehensive type hints and docstrings for agent self-discovery.

Output format

Provide the response as a single, multi-section Python script containing:

1. Environment imports and Pydantic schema declarations.
2. Tool definition metadata object for {{orchestrator_framework}}.
3. Prompt translation and enrichment helper classes.
4. The primary asynchronous entrypoint function.
5. An executable __main__ block demonstrating a simulated agent tool call payload.

Self-review

• Verify all variables ({{target_diffusion_model}}, {{orchestrator_framework}}, {{visual_domain}}, {{parameter_schema}}, {{negative_prompt_strategy}}, {{execution_runtime}}) are explicitly handled.
• Confirm the script validates edge cases such as missing parameters or out-of-bound aspect ratios.
• Ensure no placeholder tokens remain in prompt synthesis templates.