ericorps/ia-microservice
SHA256
1
1
Fork 1
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Versión 1.0

Browse Source
This commit is contained in:
lansan69
2026-03-31 20:00:04 -06:00
parent bc513ad2c4
commit e882442f8b
30 changed files with 778 additions and 89 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
app/services/image/prompt_builder.py
View File

@@ -44,9 +44,10 @@ Specialization Path: {specialization_path}
# EVALUATION RULES
1. **Domain Calibration (CRITICAL):** Calibrate your expectations, strictness, and feedback entirely according to the **{specialization_path}** context. Do not evaluate using generalized standards; apply the specific standards expected at this exact level and category.
2. **Strict Adherence:** Evaluate ONLY the criteria and subcriteria listed in the rubric. Do not invent new metrics.
3. **Scoring:** Assign a numeric `score` to each criterion and subcriterion. The score must reflect how well the input meets the description of that specific metric.
3. **Scoring:** Assign a numeric `score` to each criterion and subcriterion. The score must reflect how well the input meets the description of that specific metric.
4. **Objective Feedback:** Generate constructive, evidence-grounded `feedback` for the overall evaluation. Mention specific elements or patterns observed in the input that justify the scores within the context of the specialization.
5. **Subcriteria Handling:** If a criterion has `subcriteria`, evaluate each subcriterion individually. The parent criterion's score should be a logical aggregate (e.g., average) of its subcriteria scores.
5. **Subcriteria Handling:** If a criterion has `subcriteria`, evaluate and score each subcriterion individually. The parent criterion's `score` MUST be the exact mathematical **sum** of its subcriteria scores.
6. **Final Score Calculation:** The overall final `score` of the evaluation MUST be the exact mathematical **sum** of all the main criteria scores.
# OUTPUT FORMAT
You MUST return your response EXCLUSIVELY as a raw, valid JSON object that strictly adheres to the following JSON Schema definition. Do NOT include markdown blocks (```json), explanations, or any text outside the JSON object.