Update app.py

app.py

@@ -14,15 +14,11 @@ except ImportError:
 # =========================================
 
 MODEL_OPTIONS = {
-    "Fast (Mistral 7B)": "mistralai/Mistral-7B-Instruct-v0.2",
-    "Balanced (GPT-OSS 20B)": "openai/gpt-oss-20b",
+    "Stable (Mistral 7B)": "mistralai/Mistral-7B-Instruct-v0.2",
 }
 
 PERSONA_PRESETS = {
     "Balanced Assistant": "You are a helpful, intelligent AI assistant.",
-    "Creative": "You are highly creative and imaginative.",
-    "Precise": "You are concise, factual, and analytical.",
-    "Code Expert": "You are a senior software engineer who writes clean code.",
 
     "Forensic Psychologist": """
 You are acting as a forensic psychologist.
@@ -31,21 +27,6 @@ Do NOT diagnose real individuals.
 Separate facts, interpretations, and limitations.
 Maintain professional neutrality.
 This analysis is educational and not a clinical diagnosis.
-""",
-
-    "Forensic Psychologist (Structured Report)": """
-You are acting as a forensic psychologist preparing a structured report.
-
-Structure:
-1. Case Summary
-2. Observed Behaviors
-3. Psychological Considerations (Hypothetical)
-4. Risk Assessment Factors
-5. Alternative Explanations
-6. Evidentiary Limitations
-7. Professional Disclaimer
-
-Remain objective and non-diagnostic.
 """
 }
 
@@ -63,55 +44,25 @@ Conduct a behavioral analysis.
 6. Alternative Explanations
 7. Limitations
 
-Material:
-{user_input}
-""",
-
-    "Linguistic Analysis": """
-Conduct a linguistic analysis.
-
-1. Language Patterns
-2. Pronoun Usage
-3. Emotional Indicators
-4. Cognitive Load Markers
-5. Narrative Shifts
-6. Possible Deception Signals
-7. Limitations
-
-Material:
-{user_input}
-""",
-
-    "Credibility Analysis": """
-Conduct a credibility analysis.
-
-1. Internal Consistency
-2. Plausibility
-3. Specificity vs Vagueness
-4. Emotional Congruence
-5. Possible Fabrication Indicators
-6. Alternative Explanations
-7. Evidentiary Limitations
-
 Material:
 {user_input}
 """
 }
 
-MAX_CONTEXT_CHARS = 12000
-MAX_HISTORY_PAIRS = 8
-MAX_INPUT_CHARS = 4000
+MAX_CONTEXT_CHARS = 10000
+MAX_HISTORY_PAIRS = 6
+MAX_INPUT_CHARS = 3000
 
 
 # =========================================
 # Helpers
 # =========================================
 
-def get_client(model_name: str):
-    hf_token = os.environ.get("HF_TOKEN")
-    if not hf_token:
-        raise RuntimeError("HF_TOKEN not set. Add it as a Space Secret.")
-    return InferenceClient(model=model_name, token=hf_token)
+def get_client(model_name):
+    token = os.environ.get("HF_TOKEN")
+    if not token:
+        raise RuntimeError("HF_TOKEN not set in Space Secrets.")
+    return InferenceClient(model=model_name, token=token)
 
 
 def extract_text(file):
@@ -136,54 +87,54 @@ def extract_text(file):
 
 
 # =========================================
-# Core Streaming Logic
+# Core Logic (Non-Streaming)
 # =========================================
 
-def stream_response(message, history, model_label, persona_label, task_mode, uploaded_file):
+def generate_response(message, history, model_label, persona_label, task_mode, uploaded_file):
 
     if not message:
-        yield history
-        return
+        return history
+
+    try:
+        message = message[:MAX_INPUT_CHARS]
+        history = history[-MAX_HISTORY_PAIRS * 2:]
 
-    message = message[:MAX_INPUT_CHARS]
-    history = history[-MAX_HISTORY_PAIRS * 2:]
+        model_name = MODEL_OPTIONS[model_label]
+        system_prompt = PERSONA_PRESETS[persona_label]
 
-    model_name = MODEL_OPTIONS[model_label]
-    system_prompt = PERSONA_PRESETS[persona_label]
+        temperature = 0.4 if "Forensic" in persona_label else 0.7
 
-    temperature = 0.4 if "Forensic Psychologist" in persona_label else 0.7
+        client = get_client(model_name)
 
-    client = get_client(model_name)
+        file_text = extract_text(uploaded_file)
+        if file_text:
+            file_text = file_text[:MAX_CONTEXT_CHARS]
+            system_prompt += f"\n\nUse this document context when relevant:\n\n{file_text}"
 
-    file_text = extract_text(uploaded_file)
-    if file_text:
-        file_text = file_text[:MAX_CONTEXT_CHARS]
-        system_prompt += f"\n\nUse this document context when relevant:\n\n{file_text}"
+        messages = [{"role": "system", "content": system_prompt}]
+        messages.extend(history)
 
-    messages = [{"role": "system", "content": system_prompt}]
-    messages.extend(history)
+        formatted_input = TASK_MODES[task_mode].format(user_input=message)
+        messages.append({"role": "user", "content": formatted_input})
 
-    formatted_input = TASK_MODES[task_mode].format(user_input=message)
-    messages.append({"role": "user", "content": formatted_input})
+        response = client.chat_completion(
+            messages=messages,
+            max_tokens=700,
+            temperature=temperature,
+            top_p=0.95,
+            stream=False,
+        )
 
-    history.append({"role": "user", "content": message})
-    history.append({"role": "assistant", "content": ""})
+        answer = response.choices[0].message.content
 
-    partial = ""
+        history.append({"role": "user", "content": message})
+        history.append({"role": "assistant", "content": answer})
 
-    for chunk in client.chat_completion(
-        messages=messages,
-        max_tokens=800,
-        temperature=temperature,
-        top_p=0.95,
-        stream=True,
-    ):
-        if chunk.choices:
-            delta = chunk.choices[0].delta
-            if hasattr(delta, "content") and delta.content:
-                partial += delta.content
-                history[-1]["content"] = partial
-                yield history
+        return history
+
+    except Exception as e:
+        history.append({"role": "assistant", "content": f"⚠️ ERROR:\n\n{str(e)}"})
+        return history
 
 
 # =========================================
@@ -206,7 +157,7 @@ def export_chat(history):
 
 with gr.Blocks(theme=gr.themes.Soft(), title="Omniscient IRIS") as demo:
 
-    gr.Markdown("## Omniscient IRIS — Intelligent Analysis Assistant")
+    gr.Markdown("## Omniscient IRIS — Stable Analysis Assistant")
 
     with gr.Row():
 
@@ -218,16 +169,14 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Omniscient IRIS") as demo:
                 show_label=False
             )
 
-            with gr.Row():
-                send_btn = gr.Button("Send", variant="primary")
-                stop_btn = gr.Button("Stop")
-                clear_btn = gr.Button("Clear")
+            send_btn = gr.Button("Send", variant="primary")
+            clear_btn = gr.Button("Clear")
 
         with gr.Column(scale=1):
 
             model_selector = gr.Dropdown(
                 choices=list(MODEL_OPTIONS.keys()),
-                value="Balanced (GPT-OSS 20B)",
+                value="Stable (Mistral 7B)",
                 label="Model"
             )
 
@@ -252,20 +201,18 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Omniscient IRIS") as demo:
 
     state = gr.State([])
 
-    submit_event = send_btn.click(
-        stream_response,
+    send_btn.click(
+        generate_response,
         inputs=[msg, state, model_selector, persona_selector, task_selector, file_upload],
         outputs=chatbot,
     )
 
     msg.submit(
-        stream_response,
+        generate_response,
         inputs=[msg, state, model_selector, persona_selector, task_selector, file_upload],
         outputs=chatbot,
     )
 
-    stop_btn.click(fn=None, cancels=[submit_event])
-
     clear_btn.click(
         fn=clear_chat,
         outputs=chatbot,