(Tentative)

Function Reference Model Input/Output Data
SubAIMs JSON Metadata Profiles

1. Function

Prompt Creation AIM (PGM-PRC)

  1. Receives
    1. Context
    2. Audio and Visual Scene Descriptors from Audio and Visual Spatial Reasoning AIMs and, indirectly, Domain Access (PGM-DAC).
    3. Prompt Creation Directive from A-User Control.
  2. Synthesises these multimodal and contextual signals into coherent natural language prompts that describe the User’s interaction, clarify intent, and confirm inferred goals.
  3. Provides
    1. Textual PC-Prompt to Basic Knowledge LLM.
    2. Prompt Creation Statusto A-User Control.

The resulting outputs enable A-User Control, Personality Alignment, and Rendering AIMs to operate with full awareness of the User’s communicative intent, supporting expression coherence, goal-driven orchestration, and context-sensitive interaction.

2. Reference Model

Figure 1 gives Reference Model of Prompt Creation (PGM-PRC).

Figure – The Reference Model of Prompt Creation (PGM-PRC)

3. Input/Output Data

Table 1 – Input/Output Data of PGM-PRC

Input Description
Context A structured and time-stamped snapshot representing the initial understanding that the A-User achieves of the environment and of the User posture.
Audio Scene Descriptors A Data Type that conveys spatially grounded semantic audio data from the Spatial Reasoning AIM that enables the Prompt Construction AIM to generate context-aware and referentially precise prompts.
Visual Scene Descriptors A Data Type that conveys spatially grounded semantic visual data from the Spatial Reasoning AIM that enables the Prompt Construction AIM to generate context-aware and referentially precise prompts.
Prompt Creation Directive Trigger to initiates prompt generation or refinement from PGM-AUC.
Output Description
PC-Prompt Prompt to Basic Knowledge
Prompt Creation Status Prompt readiness, alignment status, and semantic goal framing to PGM-AUA.

4. SubAIMs

The identification of SubAIMs is based on the following steps:

  1. Structure the Response
    The initial natural language output needs to be converted into a structured format (DA Input) so downstream modules can process it reliably.
  2. Ensure Multimodal Consistency
    Items referenced in the prompt must match what’s detected in audio and visual scenes—avoiding ambiguity and grounding the interaction.
  3. Validate Domain Rules and Constraints
    Before planning any action, the Prompt Creation checks compliance with operational, safety, and domain-specific rules to prevent invalid behaviours.
  4. Assess Zone Feasibility
    Actions often depend on spatial zones (e.g., reachable areas). Verifying zone feasibility ensures the intended behaviour is physically or logically possible.
  5. Confirm Scene Integrity
    Scene consistency checks prevent mismatches between what the Prompt Creation thinks the environment is and what it actually is.
  6. Select Appropriate Behaviours
    From validated context and user state, candidate behaviours are retrieved to align with the User’s goals and domain logic.
  7. Check Execution Feasibility
    Even valid behaviours must be executable under current constraints to ensure practical feasibility before committing.
  8. Provide Fallback Options
    If execution isn’t feasible, fallback actions or adjusted prompts are generated to keep the interaction flowing without failure.
  9. Prepare a Structured PC-Prompt Plan
    The final output is a PC-Prompt Plan that integrates all validated context, constraints, and behaviours for subsequent natural language generation.
  10. Produce PC-Prompt
    Convert the PC-Prompt into natual language prompt..

Figure 2 identifies a possible partitioning of SubAIMs.

Figure 2 – Reference Model of Prompt Creation Composite (PGM-PRC) AIM

For graphical reasons, Data Types are identified by three-letter acronyms

ASD Audio Scene Descriptors IRS Initial Response
BCL Behaviour Candidates List IZM Item Zone Match
CCR Constraint Check Result PPP PC-Prompt Plan
DAI DA-Input SCF Scene Consistency Flag
DIF DA-Input Flag USS User State
EFS Execution Feasibility Status VSD Visual Scene Descriptors
FPA Fallback Prompt Action ZFT Zone Feasibility Tag

Table 2 defines a possible set of SubAIMs for PGM-PRC

Table 2 – Potential SubAIMs for PGM-PRC

Sub-AIM Function Inputs Outputs To Description
CDI Create DA Input Initial Response DA Input; DA Input Flags CVI, VFC, CSC, LTM Converts natural language response into structured JSON and generates flags for constraint checks.
CVI Cross Validate Items DA Input; ASD; VSD Item Zone Match VFC Confirms item positions and associations across audio and visual descriptors for multimodal consistency.
VFC Verify Constraints DA Input Flags; ASD; VSD; Item Zone Match Constraint Check Result CZC, EEF Checks compliance with domain rules and operational constraints using multimodal context and validation flags.
CZC Check Zone Classification ASD; VSD; Constraint Check Result Zone Feasibility Tag CSC, EEF Determines whether zones are feasible for intended actions based on spatial and audio descriptors.
CSC Check Scene Consistency SceneID (DA Input); SceneID (Spatial Reasoning); Zone Feasibility Tag Scene Consistency Flag LBR Validates that the scene context matches expected spatial and logical structure for safe execution.
LBR Lookup Behaviour Registry User State; Item Zone Match; Scene Consistency Flag Behaviour Candidates List EEF Retrieves candidate behaviors from a registry based on user state and validated scene context.
EEF Evaluate Execution Feasibility Behaviour Candidates List; Constraint Check Result; Zone Feasibility Tag Execution Feasibility Status FBA, EEP Assesses whether candidate behaviours can be executed under current constraints and zone conditions.
FBA Fallback or Adjust Execution Feasibility Status; ASD; VSD Fallback Prompt Action PCP (PC Prompt Planner) Generates alternative actions or fallback prompts when execution feasibility fails.
EEP Emit Execution Payload Execution Feasibility Status; Behaviour Candidates List Execution Payload Object PAL, USR Produces the final structured payload for execution, including selected behaviour and feasibility metadata.
LTM Log Trace Metadata DA Input ID; ASD ID; VSD ID; Execution Feasibility Status Trace Provenance Log AUC Records provenance and traceability data for auditing and lifecycle management.

5. JSON Metadata

https://schemas.mpai.community/PGM1/V1.0/AIMs/PromptCreation.json

6. Profiles

No Profiles.