best practice validation

Query use case

Was best practice used in the system validation process? E.g. QA before release

Schemas used

Links

• ai system
• config
• data pack
• data
• best practice attestation

Pseudo code

FUNCTION ai_system_best_practice_attestations(Inference_AI_System_ID)
    CREATE empty lists InferenceAtts, TrainingAtts, DataAtts

    // Step 1: Get the configuration ID from the inference AI system
    SET Config_ID = get configuration ID from Inference_AI_System_ID

    // Step 2: Find the training AI system using the configuration
    SET Training_AI_System_ID = get training system linked to Config_ID

    // Step 3: Retrieve the datapack ID from the training AI system
    SET Datapack_ID = get datapack ID from Training_AI_System_ID

    // Step 4: Extract all dataset IDs from the datapack
    SET Dataset_List = get dataset list from Datapack_ID
    SET Data_IDs = extract all Data_IDs from Dataset_List

    // Step 5: Collect best practice attestations for inference AI system
    FOR EACH Attestation IN database DO
        IF Attestation is for component(_, Inference_AI_System_ID) AND
           Attestation is of type best_practice_attestation THEN
            ADD (Attestation, Inference_AI_System_ID) TO InferenceAtts

    // Step 6: Collect best practice attestations for training AI system
    FOR EACH Attestation IN database DO
        IF Attestation is for component(_, Training_AI_System_ID) AND
           Attestation is of type best_practice_attestation THEN
            ADD (Attestation, Training_AI_System_ID) TO TrainingAtts

    // Step 7: Collect best practice attestations for each data ID
    FOR EACH Data_ID in Data_IDs DO
        FOR EACH Attestation IN database DO
            IF Attestation is for component(_, Data_ID) AND
               Attestation is of type best_practice_attestation THEN
                ADD (Attestation, Data_ID) TO DataAtts

    // Step 8: Return all collected attestations
    RETURN (InferenceAtts, TrainingAtts, DataAtts)
END FUNCTION

---

Explanation

1. Extract Configuration for Inference System:

   • The function begins by retrieving the Config_ID associated with the inference AI system. This configuration is needed to trace the training process.

2. Determine Training AI System:

   • Using the Config_ID, it locates the training AI system (Training_AI_System_ID) which was used to generate or refine the inference system.

3. Retrieve Datapack from Training System:

   • The training system contains a reference to a Datapack_ID, which holds information about the datasets used during training.

4. Extract Dataset Identifiers:

   • The datapack is unpacked to retrieve all Data_IDs representing the individual datasets used during the training process.

5. Find Inference System Best Practice Attestations:

   • It scans for all best_practice_attestation facts where the attested component is the inference AI system, collecting them into InferenceAtts.

6. Find Training System Best Practice Attestations:

   • Similarly, it gathers best practice attestations related to the training AI system into TrainingAtts.

7. Find Training Data Best Practice Attestations:

   • For each Data_ID, the function finds any best practice attestations linked to that dataset and stores them in DataAtts.

8. Return Attestation Collections:

   • Finally, the function returns three lists: one for the inference system, one for the training system, and one for the training datasets.

Query

• db:ai_system_best_practice_attestations(AiSystemId, InferenceAttestations, TrainingAttestations, DataAttestations) link to query
• link to simulator

Notes