Redesign Session Log with full conversations and clinical highlights

- Session selector dropdown to pick saved sessions
- Full conversation display (complete back-and-forth, not preview)
- Clinical Highlights panel with 'Generate' button per session
- Highlights identify key moments, quote specific language, note clinical observations
- Shows prompt sculpting observations (what's working, what tensions emerged)
- Cross-session reports (Session Report, A/B Comparison) remain available
- Cleaner two-column layout: conversation left, highlights right

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

app.py

@@ -89,27 +89,128 @@ def save_to_session_log(sessions, system_prompt, history, persona, label):
 
 
 def format_session_display(sessions):
-    """Format sessions for display."""
+    """Format sessions for dropdown display."""
     if not sessions:
-        return "No sessions saved yet. Use 'Save to Session Log' in Test & Analyze tab."
+        return "No sessions saved yet."
+    return f"{len(sessions)} session(s) saved. Select one below to view."
 
-    display = ""
-    for i, s in enumerate(sessions):
-        display += f"### {i+1}. {s['label']}\n"
-        display += f"*{s['timestamp']} | Persona: {s['persona']} | {s['exchange_count']} exchanges*\n\n"
-        display += f"**Prompt excerpt:** {s['system_prompt'][:200]}...\n\n"
 
-        # Show first exchange as preview
-        if s['conversation']:
-            user_msg, bot_msg = s['conversation'][0]
-            display += f"**First exchange:**\n> User: {user_msg[:100]}...\n\n"
-            display += f"> Bot: {bot_msg[:150]}...\n\n"
+def get_session_choices(sessions):
+    """Get dropdown choices for sessions."""
+    if not sessions:
+        return []
+    return [(f"{s['label']} ({s['timestamp']}, {s['persona']})", i) for i, s in enumerate(sessions)]
+
 
-        display += "---\n\n"
+def format_full_conversation(sessions, selected_index):
+    """Format the full conversation for a selected session."""
+    if not sessions or selected_index is None:
+        return "Select a session to view the full conversation."
+
+    try:
+        idx = int(selected_index)
+        s = sessions[idx]
+    except (ValueError, IndexError, TypeError):
+        return "Select a session to view."
+
+    display = f"## {s['label']}\n"
+    display += f"*{s['timestamp']} | Persona: {s['persona']}*\n\n"
+    display += f"---\n\n"
+    display += f"**System Prompt:**\n```\n{s['full_prompt']}\n```\n\n"
+    display += f"---\n\n"
+    display += "### Conversation\n\n"
+
+    for i, (user_msg, bot_msg) in enumerate(s['conversation']):
+        display += f"**USER:**\n{user_msg}\n\n"
+        display += f"**BOT:**\n{bot_msg}\n\n"
+        if i < len(s['conversation']) - 1:
+            display += "---\n\n"
 
     return display
 
 
+def generate_session_highlights(api_key_input, sessions, selected_index):
+    """Generate clinical highlights for a specific session."""
+    key_to_use = api_key_input.strip() if api_key_input else ""
+    if not key_to_use:
+        key_to_use, _ = get_api_key_from_env()
+
+    if not key_to_use:
+        return "API key required."
+
+    if not sessions or selected_index is None:
+        return "Select a session first."
+
+    try:
+        idx = int(selected_index)
+        s = sessions[idx]
+    except (ValueError, IndexError, TypeError):
+        return "Select a session first."
+
+    # Format conversation
+    conv_text = ""
+    for user_msg, bot_msg in s['conversation']:
+        conv_text += f"USER: {user_msg}\n\nBOT: {bot_msg}\n\n---\n\n"
+
+    highlights_prompt = f"""You are a clinical consultant reviewing this AI chatbot conversation. Generate concise clinical highlights—observations a trauma-informed prompt engineer would find valuable.
+
+SYSTEM PROMPT:
+{s['full_prompt']}
+
+CONVERSATION:
+{conv_text}
+
+---
+
+Generate clinical highlights in this format:
+
+## Key Moments
+
+For each notable exchange, quote the specific language and note what's happening clinically:
+
+**Moment 1: [Brief title]**
+> [Quote the key phrase]
+
+*Clinical note:* [1-2 sentences on what this reveals about the prompt's effect—good or concerning]
+
+**Moment 2: [Brief title]**
+> [Quote]
+
+*Clinical note:* [Observation]
+
+(Continue for 3-5 key moments)
+
+---
+
+## Prompt Sculpting Observations
+
+What is the system prompt successfully doing here?
+- [Observation with quote]
+- [Observation with quote]
+
+What tensions or risks emerged?
+- [Observation with quote]
+
+## One-Line Summary
+
+[Single sentence capturing the psychodynamic signature of this exchange]
+
+---
+
+Be specific. Quote actual phrases. Focus on the nuances a clinician would notice—boundary maintenance, first-person language choices, bridging to human support, capacity-building vs dependency, the displaced listener."""
+
+    try:
+        client = anthropic.Anthropic(api_key=key_to_use)
+        response = client.messages.create(
+            model="claude-sonnet-4-20250514",
+            max_tokens=1500,
+            messages=[{"role": "user", "content": highlights_prompt}]
+        )
+        return response.content[0].text
+    except Exception as e:
+        return f"Error: {str(e)}"
+
+
 def generate_session_report(api_key_input, sessions):
     """Generate clinical summary across all saved sessions."""
     key_to_use = api_key_input.strip() if api_key_input else ""
@@ -776,32 +877,55 @@ with gr.Blocks(title="PromptWork", theme=gr.themes.Soft()) as app:
         # TAB 3: Session Log
         with gr.Tab("Session Log"):
             gr.Markdown("### Saved Test Sessions")
-            gr.Markdown("*Conversations saved from Test & Analyze appear here. Label sessions for A/B comparison.*")
+            gr.Markdown("*View full conversations with clinical highlights. Label sessions 'A'/'B' for comparison.*")
 
             with gr.Row():
-                with gr.Column(scale=2):
-                    session_display = gr.Markdown("No sessions saved yet. Use 'Save to Session Log' in Test & Analyze tab.")
+                session_display = gr.Markdown("No sessions saved yet.")
+                session_selector = gr.Dropdown(
+                    label="Select Session",
+                    choices=[],
+                    value=None,
+                    interactive=True,
+                    scale=2
+                )
+                generate_highlights_btn = gr.Button("Generate Clinical Highlights", variant="primary", scale=1)
 
+            with gr.Row():
                 with gr.Column(scale=1):
-                    gr.Markdown("### Generate Reports")
+                    gr.Markdown("### Full Conversation")
+                    conversation_display = gr.Markdown(
+                        "Select a session above to view the full conversation.",
+                        elem_id="conversation-display"
+                    )
 
-                    gr.Markdown("**Session Report** - Clinical summary across all saved sessions")
-                    generate_report_btn = gr.Button("Generate Session Report", variant="primary")
+                with gr.Column(scale=1):
+                    gr.Markdown("### Clinical Highlights")
+                    highlights_display = gr.Markdown(
+                        "*Click 'Generate Clinical Highlights' to analyze the selected conversation.*\n\nHighlights will identify key moments, quote specific language, and note what's happening clinically—boundary maintenance, first-person choices, bridging to human support, capacity-building vs dependency."
+                    )
 
-                    gr.Markdown("---")
+            gr.Markdown("---")
 
-                    gr.Markdown("**A/B Comparison** - Compare sessions labeled 'A' vs 'B'")
-                    gr.Markdown("*Tip: Label sessions as 'Prompt A', 'Prompt B' (or just 'A', 'B') when saving*")
-                    generate_ab_btn = gr.Button("Generate A/B Comparison", variant="primary")
+            with gr.Row():
+                with gr.Column(scale=1):
+                    gr.Markdown("### Cross-Session Reports")
+                    gr.Markdown("*Synthesize patterns across all saved sessions*")
 
-                    gr.Markdown("---")
+                with gr.Column(scale=1):
+                    generate_report_btn = gr.Button("Generate Session Report", variant="secondary")
+                    gr.Markdown("*Clinical summary across all sessions*")
+
+                with gr.Column(scale=1):
+                    generate_ab_btn = gr.Button("Generate A/B Comparison", variant="secondary")
+                    gr.Markdown("*Compare sessions labeled A vs B*")
 
+                with gr.Column(scale=1):
                     clear_sessions_btn = gr.Button("Clear All Sessions", variant="stop")
                     clear_status = gr.Textbox(label="", interactive=False, show_label=False)
 
             gr.Markdown("---")
             gr.Markdown("### Report Output")
-            report_output = gr.Markdown("*Reports will appear here after generation*")
+            report_output = gr.Markdown("*Cross-session reports will appear here*")
 
         # TAB 4: Compare Responses (manual paste)
         with gr.Tab("Compare Responses"):
@@ -858,6 +982,10 @@ with gr.Blocks(title="PromptWork", theme=gr.themes.Soft()) as app:
     clear_btn.click(clear_chat, [], [chatbot])
 
     # Session log events
+    def update_session_dropdown(sessions):
+        choices = get_session_choices(sessions)
+        return gr.Dropdown(choices=choices, value=choices[-1][1] if choices else None)
+
     save_session_btn.click(
         save_to_session_log,
         [session_state, prompt_input, chatbot, persona_dropdown, session_label],
@@ -866,6 +994,24 @@ with gr.Blocks(title="PromptWork", theme=gr.themes.Soft()) as app:
         format_session_display,
         [session_state],
         [session_display]
+    ).then(
+        update_session_dropdown,
+        [session_state],
+        [session_selector]
+    )
+
+    # When session is selected, show full conversation
+    session_selector.change(
+        format_full_conversation,
+        [session_state, session_selector],
+        [conversation_display]
+    )
+
+    # Generate highlights for selected session
+    generate_highlights_btn.click(
+        generate_session_highlights,
+        [api_key, session_state, session_selector],
+        [highlights_display]
     )
 
     generate_report_btn.click(
@@ -880,10 +1026,13 @@ with gr.Blocks(title="PromptWork", theme=gr.themes.Soft()) as app:
         [report_output]
     )
 
+    def clear_and_reset():
+        return [], "Sessions cleared.", gr.Dropdown(choices=[], value=None), "Select a session to view.", "*Click 'Generate Clinical Highlights' to analyze.*"
+
     clear_sessions_btn.click(
-        clear_sessions,
+        clear_and_reset,
         [],
-        [session_state, clear_status]
+        [session_state, clear_status, session_selector, conversation_display, highlights_display]
     ).then(
         format_session_display,
         [session_state],