iv_json_updater (#5)

- chore: update model version (c2a4f327b767c6423b0e1a9aa597a4e3ae7bd5eb)
- fix: clear content before processing new usecase (c3beeca949fc714bef6184511ea829f586f2c7c3)


Co-authored-by: Inge V <ingelise@users.noreply.huggingface.co>

app.py

@@ -1,7 +1,7 @@
 import os
 import gradio as gr
 import logging
-from executor import risk_identifier, mitigations
+from executor import clear_previous_risks, clear_previous_mitigations, risk_identifier, mitigations
 
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
@@ -47,15 +47,15 @@ class UI:
                     interactive=True,
                 )
                 self.model_name_or_path = gr.Dropdown(
-                    #choices=['codellama/codellama-34b-instruct-hf', 'google/flan-t5-xl', 'google/flan-t5-xxl', 'google/flan-ul2', 'ibm/granite-13b-instruct-v2', 'ibm/granite-20b-code-instruct', 'ibm/granite-20b-multilingual', 'ibm/granite-3-2-8b-instruct-preview-rc', 'ibm/granite-3-2b-instruct', 'ibm/granite-3-8b-instruct', 'ibm/granite-34b-code-instruct', 'ibm/granite-3b-code-instruct', 'ibm/granite-8b-code-instruct', 'ibm/granite-guardian-3-2b', 'ibm/granite-guardian-3-8b', 'meta-llama/llama-2-13b-chat', 'meta-llama/llama-3-1-70b-instruct', 'meta-llama/llama-3-1-8b-instruct', 'meta-llama/llama-3-2-11b-vision-instruct', 'meta-llama/llama-3-2-1b-instruct', 'meta-llama/llama-3-2-3b-instruct', 'meta-llama/llama-3-2-90b-vision-instruct', 'meta-llama/llama-3-3-70b-instruct', 'meta-llama/llama-3-405b-instruct', 'meta-llama/llama-guard-3-11b-vision', 'mistralai/mistral-large', 'mistralai/mixtral-8x7b-instruct-v01'],
-                    choices=["ibm/granite-20b-code-instruct"],
-                    value="ibm/granite-20b-code-instruct",
+                    #choices=['codellama/codellama-34b-instruct-hf', 'google/flan-t5-xl', 'google/flan-t5-xxl', 'google/flan-ul2', 'ibm/granite-13b-instruct-v2', 'ibm/granite-3-3-8b-instruct', 'ibm/granite-20b-multilingual', 'ibm/granite-3-2-8b-instruct-preview-rc', 'ibm/granite-3-2b-instruct', 'ibm/granite-3-8b-instruct', 'ibm/granite-34b-code-instruct', 'ibm/granite-3b-code-instruct', 'ibm/granite-8b-code-instruct', 'ibm/granite-guardian-3-2b', 'ibm/granite-guardian-3-8b', 'meta-llama/llama-2-13b-chat', 'meta-llama/llama-3-1-70b-instruct', 'meta-llama/llama-3-1-8b-instruct', 'meta-llama/llama-3-2-11b-vision-instruct', 'meta-llama/llama-3-2-1b-instruct', 'meta-llama/llama-3-2-3b-instruct', 'meta-llama/llama-3-2-90b-vision-instruct', 'meta-llama/llama-3-3-70b-instruct', 'meta-llama/llama-3-405b-instruct', 'meta-llama/llama-guard-3-11b-vision', 'mistralai/mistral-large', 'mistralai/mixtral-8x7b-instruct-v01'],
+                    choices=["ibm/granite-3-3-8b-instruct"],
+                    value="ibm/granite-3-3-8b-instruct",
                     multiselect=False,
                     label="Choose language model to use",
                     info="Language model used to assess risks (This is not the model being assessed).",
                     interactive=True
                 )
-                examples = gr.Examples([["A medical chatbot for a triage system that assesses patients' symptoms and provides immediate, contextually relevant advice based on their medical history and current condition. The chatbot uses generative AI to analyze the patient's input, identify potential medical issues, and offer tailored recommendations or insights to the patient or healthcare provider.", "ibm-risk-atlas"],
+                examples = gr.Examples([["A medical chatbot for a triage system that assesses patient symptoms and provides advice based on their medical history and current condition. The chatbot uses generative AI to analyze the patient input, identify potential medical issues, and offer tailored recommendations or insights to the patient or healthcare provider.", "ibm-risk-atlas"],
                                         ["Building a customer support agent that automatically triages common problems with services.", "ibm-risk-atlas"]],
                                        [self.usecase, self.taxonomy],
                                         label='Example use cases', example_labels=["Medical chatbot", "Customer service agent"] 
@@ -68,29 +68,31 @@ class UI:
                 self.assessed_risks = gr.Dataset(elem_classes="risks", label=None, visible=False)
                 self.assessed_risk_definition = gr.Markdown()
 
-                gr.Markdown(
-                """<h2> Related Risks </h2> 
-                Select a potential risk above to check for related risks.
-                """
-                )
-                rrtb = gr.Markdown()
-                self.relatedrisks = gr.Dataset(elem_classes="related-risks", components=[rrtb], label=None, visible=False)
-
-                gr.Markdown(
-                """<h2> Mitigations </h2> 
-                Select a potential risk to determine possible mitigations. """
-                )
-                self.mitigations_text = gr.Markdown()
-                self.mitigations = gr.DataFrame(label=None, visible=False)
-            
-                gr.Markdown(
-                """<h2>Benchmarks </h2> 
-                Select a potential risk to determine possible AI evaluations. """
-                )
-                self.benchmarks_text = gr.Markdown()
-                self.benchmarks = gr.DataFrame(label=None, visible=False)
-
-                self.download = gr.DownloadButton("Download JSON", visible=False)
+                if len(self.assessed_risks.elem_classes ) > 0:
+                    gr.Markdown(
+                    """<h2> Related Risks </h2> 
+                    Select a potential risk above to check for related risks.
+                    """
+                    )
+                    rrtb = gr.Markdown()
+                    self.relatedrisks = gr.Dataset(elem_classes="related-risks", components=[rrtb], label=None, visible=False)
+
+
+                    gr.Markdown(
+                    """<h2> Mitigations </h2> 
+                    Select a potential risk to determine possible mitigations. """
+                    )
+                    self.mitigations_text = gr.Markdown()
+                    self.mitigations = gr.DataFrame(label=None, visible=False)
+                
+                    gr.Markdown(
+                    """<h2>Benchmarks </h2> 
+                    Select a potential risk to determine possible AI evaluations. """
+                    )
+                    self.benchmarks_text = gr.Markdown()
+                    self.benchmarks = gr.DataFrame(label=None, visible=False)
+
+                    self.download = gr.DownloadButton("Download JSON", visible=False)
 
         gr.Markdown("---")
         gr.Markdown("<br>")
@@ -107,12 +109,16 @@ class UI:
 
             self.header_block()
             self.risks = gr.State()
-
+            
             # Risk assessment based on user intents
             self.risk_extraction()
 
-            # Register event listeners
+            # Register event listener
             self.risk_execute.click(
+                fn=clear_previous_risks,
+                inputs=[],
+                outputs=[self.assessment_sec, self.risks, self.assessed_risks, self.download, self.assessed_risk_definition, self.relatedrisks, self.mitigations, self.benchmarks, self.mitigations_text],
+            ).then(
                 fn=risk_identifier,
                 inputs=[
                     self.usecase,
@@ -124,6 +130,10 @@ class UI:
             )
 
             self.assessed_risks.select(
+                fn=clear_previous_mitigations,
+                inputs=[],
+                outputs=[self.assessed_risk_definition, self.relatedrisks, self.mitigations, self.benchmarks, self.mitigations_text]
+            ).then(
                 fn=mitigations,
                 inputs=[self.assessed_risks, self.taxonomy], 
                 # NOTETOSELF: Intent based risk is stored in self.risk (if needed)

executor.py

@@ -24,11 +24,20 @@ from functools import lru_cache
 ran = RiskAtlasNexus() # type: ignore
 
 
+def clear_previous_risks():
+    return gr.Markdown("""<h2> Potential Risks </h2> """), [], gr.Dataset(samples=[], 
+                                     sample_labels=[], 
+                                     samples_per_page=50, visible=False), gr.DownloadButton("Download JSON", visible=False, ), "", gr.Dataset(samples=[], sample_labels=[], visible=False), gr.DataFrame([], wrap=True, show_copy_button=True, show_search="search", visible=False), gr.DataFrame([], wrap=True, show_copy_button=True, show_search="search", visible=False), gr.Markdown(" ") 
+
+def clear_previous_mitigations():
+     return "", gr.Dataset(samples=[], sample_labels=[], visible=False), gr.DataFrame([], wrap=True, show_copy_button=True, show_search="search", visible=False), gr.DataFrame([], wrap=True, show_copy_button=True, show_search="search", visible=False), gr.Markdown(" ") 
+
 @lru_cache
 def risk_identifier(usecase: str, 
-                    model_name_or_path: str = "ibm/granite-20b-code-instruct", 
+                    model_name_or_path: str = "ibm/granite-3-3-8b-instruct", 
                     taxonomy: str = "ibm-risk-atlas"): # -> List[Dict[str, Any]]: #pd.DataFrame:
 
+    downloadable = False
     inference_engine = WMLInferenceEngine(
         model_name_or_path= model_name_or_path,
         credentials={
@@ -61,12 +70,15 @@ def risk_identifier(usecase: str,
                 'risks': [json.loads(r.json()) for r in risks]
         }
     file_path = Path("static/download.json")
-    file_path.write_text(json.dumps(data, indent=4), encoding='utf-8')
-        
+    with open(file_path, mode='w') as f:
+        f.write(json.dumps(data, indent=4))
+        downloadable = True
+
+   
     #return out_df
     return out_sec, gr.State(risks), gr.Dataset(samples=[r.id for r in risks], 
                                      sample_labels=sample_labels, 
-                                     samples_per_page=50, visible=True, label="Estimated by an LLM."), gr.DownloadButton("Download JSON", visible=True, value="static/download.json")
+                                     samples_per_page=50, visible=True, label="Estimated by an LLM."), gr.DownloadButton("Download JSON", "static/download.json", visible=(downloadable and len(risks) > 0))
     
 
 @lru_cache