updated innovation classifier (prompt)

src/pipeline.py

@@ -132,16 +132,19 @@ def process_data(uploaded_file, sens_level, azure_client, azure_deployment):
             )
         
         elif model_name == 'INNOVATION_CLASSIFICATION':
-            df['innovation_classification'] = df.apply(
-                lambda x: classify_innovation(
+            results = []
+            total_rows = len(df)
+            for idx, row in df.iterrows():
+                result = classify_innovation(
                     azure_client,
                     azure_deployment,
-                    x['id'],
-                    x['tech_txt'],
-                    x['tech_rationale_txt']
-                ),
-                axis=1
-            )
+                    row['id'],
+                    row['tech_txt'],
+                    row['tech_rationale_txt']
+                )
+                results.append(result)
+                progress_bar.progress((len(results)) / total_rows)
+            df['innovation_classification'] = results
 
 
         logger.info(f"Completed: {model_name}")

src/prompts.py

@@ -14,13 +14,41 @@ def prompt_innovation(concept: str) -> str:
     locally manufactured lithium-ion batteries for custom-built e-mobility solutions, or reactive power
     compensation systems to enhance grid stability.
 
-    Classification rubric:
-    - NOT INNOVATIVE: already common/deployed in the local context, or standard practice with no clear novelty.
-    - MODERATELY INNOVATIVE: established elsewhere but new to the local context, or a meaningful adaptation.
-    - VERY INNOVATIVE: novel approach globally or locally with clear differentiation from standard solutions.
-    - INSUFFICIENT INFO: not enough detail to classify innovation level.
+    Important clarifications for classification:
+
+    TECHNOLOGY AVAILABILITY: If the text indicates that the core technologies are already available,
+    deployed, or proven in the country (even for different sectors or applications) this strongly
+    signals an established market. Applying existing in-country technology to a new sector (e.g.,
+    solar from residential to mining) is deployment, not innovation.
+
+    SCALING vs. INNOVATION: Scaling up existing pilots, increasing unit counts, or expanding
+    geographic reach of proven technologies does not constitute innovation. Similarly, changing
+    ownership models (e.g., community-owned vs. privately-owned) or delivery mechanisms for
+    established technologies is not technical innovation.
 
-    If the concept is too vague to judge novelty, choose the lowest classification that is defensible and reflect uncertainty in confidence.
+    COMBINING TECHNOLOGIES: Bundling or integrating multiple established technologies (e.g., solar +
+    e-mobility + enterprise systems) does not create innovation if each component technology is
+    already available and proven in-country. The combination must involve genuinely novel technical
+    integration or adaptation to qualify.
+
+    CLAIMS vs. EVIDENCE: Discount self-described "innovation" claims. Focus on whether the underlying
+    technology/approach itself is novel, not whether the project framing or business model is described as new.
+
+    Classification rubric:
+    - NOT INNOVATIVE: Technologies already available, deployed, or proven in-country (even if in
+    different sectors); scaling or replicating existing pilots; combining established technologies
+    without novel technical integration; standard practice with no clear technical novelty.
+    - MODERATELY INNOVATIVE: Technology genuinely new to the country (not just the sector) with no
+    prior in-country availability or deployment; OR a technically meaningful adaptation that goes
+    beyond repackaging or recombining existing solutions.
+    - VERY INNOVATIVE: Novel approach globally or locally with clear technical differentiation from
+    standard solutions; represents a genuine technological advancement, not just deployment scale
+    or business model change.
+    - INSUFFICIENT INFO: Not enough detail to classify innovation level.
+
+    Default to the lowest defensible classification. If technologies are described as "available"
+    or "proven" in-country, default to NOT INNOVATIVE unless there is clear evidence of genuine
+    technical novelty beyond deployment context.
 
     Concept for review:
     {concept}