add in history

README.md

@@ -101,6 +101,14 @@ You can customize the system by editing the `social_graph.json` file. The file h
 }
 ```
 
+## Plan
+
+Look at using a Structured Knowledge Format (SKF) – a compact, machine-optimized format designed for efficient AI parsing rather than human readability.
+
+We should create a SKF<->JSON converter to convert the social graph JSON into a more compact format. This will help in reducing the size of the social graph and make it easier for AI models to parse.
+
+See also https://github.com/marv1nnnnn/llm-min.txt
+
 ## Current Social Graph Context
 
 The current social graph represents a British person with MND who:

app.py

@@ -119,6 +119,9 @@ def get_people_choices():
         display_name = format_person_display(person)
         person_id = person["id"]
         choices[display_name] = person_id
+
+    # Debug the choices
+    print(f"People choices: {choices}")
     return choices
 
 
@@ -152,9 +155,19 @@ def get_suggestion_categories():
 def on_person_change(person_id):
     """Handle person selection change."""
     if not person_id:
-        return "", "", []
+        return "", "", [], ""
 
-    person_context = social_graph.get_person_context(person_id)
+    # Get the people choices dictionary
+    people_choices = get_people_choices()
+
+    # Extract the actual ID if it's in the format "Name (role)"
+    actual_person_id = person_id
+    if person_id in people_choices:
+        # If the person_id is a display name, get the actual ID
+        actual_person_id = people_choices[person_id]
+        print(f"on_person_change: Extracted actual person ID: {actual_person_id}")
+
+    person_context = social_graph.get_person_context(actual_person_id)
 
     # Create a more user-friendly context display
     name = person_context.get("name", "")
@@ -176,7 +189,45 @@ def on_person_change(person_id):
     # Get topics for this person
     topics = person_context.get("topics", [])
 
-    return context_info, phrases_text, topics
+    # Get conversation history for this person
+    conversation_history = person_context.get("conversation_history", [])
+    history_text = ""
+
+    if conversation_history:
+        # Sort by timestamp (most recent first)
+        sorted_history = sorted(
+            conversation_history, key=lambda x: x.get("timestamp", ""), reverse=True
+        )[
+            :2
+        ]  # Get only the 2 most recent conversations
+
+        history_text = "### Recent Conversations:\n\n"
+
+        for i, conversation in enumerate(sorted_history):
+            # Format the timestamp
+            timestamp = conversation.get("timestamp", "")
+            try:
+                import datetime
+
+                dt = datetime.datetime.fromisoformat(timestamp)
+                formatted_date = dt.strftime("%B %d, %Y at %I:%M %p")
+            except (ValueError, TypeError):
+                formatted_date = timestamp
+
+            history_text += f"**Conversation on {formatted_date}:**\n\n"
+
+            # Add the messages
+            messages = conversation.get("messages", [])
+            for message in messages:
+                speaker = message.get("speaker", "Unknown")
+                text = message.get("text", "")
+                history_text += f"*{speaker}*: {text}\n\n"
+
+            # Add a separator between conversations
+            if i < len(sorted_history) - 1:
+                history_text += "---\n\n"
+
+    return context_info, phrases_text, topics, history_text
 
 
 def change_model(model_name, progress=gr.Progress()):
@@ -541,6 +592,129 @@ def transcribe_audio(audio_path):
         return "Could not transcribe audio. Please try again."
 
 
+def save_conversation(person_id, user_input, selected_response):
+    """Save a conversation to the social graph.
+
+    Args:
+        person_id: ID of the person in the conversation
+        user_input: What the person said to Will
+        selected_response: Will's response
+
+    Returns:
+        True if successful, False otherwise
+    """
+    print(f"Saving conversation for person_id: {person_id}")
+    print(f"User input: {user_input}")
+    print(f"Selected response: {selected_response}")
+
+    if not person_id:
+        print("Error: No person_id provided")
+        return False
+
+    if not (user_input or selected_response):
+        print("Error: No user input or selected response provided")
+        return False
+
+    # Create message objects
+    messages = []
+
+    # Get the person's name
+    person_context = social_graph.get_person_context(person_id)
+    if not person_context:
+        print(f"Error: Could not get person context for {person_id}")
+        return False
+
+    person_name = person_context.get("name", "Person")
+    print(f"Person name: {person_name}")
+
+    # Add the user's message if provided
+    if user_input:
+        messages.append({"speaker": person_name, "text": user_input})
+        print(f"Added user message: {user_input}")
+
+    # Add Will's response
+    if selected_response:
+        messages.append({"speaker": "Will", "text": selected_response})
+        print(f"Added Will's response: {selected_response}")
+
+    # Save the conversation
+    if messages:
+        print(f"Saving {len(messages)} messages to conversation history")
+        try:
+            success = social_graph.add_conversation(person_id, messages)
+            print(f"Save result: {success}")
+            if success:
+                # Manage the conversation history (keep only the most recent ones)
+                manage_result = manage_conversation_history(person_id)
+                print(f"Manage conversation history result: {manage_result}")
+            return success
+        except Exception as e:
+            print(f"Error saving conversation: {e}")
+            return False
+    else:
+        print("No messages to save")
+
+    return False
+
+
+def manage_conversation_history(person_id, max_conversations=5):
+    """Manage the conversation history for a person.
+
+    Args:
+        person_id: ID of the person
+        max_conversations: Maximum number of conversations to keep in the social graph
+
+    Returns:
+        True if successful, False otherwise
+    """
+    if not person_id:
+        return False
+
+    # Get the person's conversation history
+    person_context = social_graph.get_person_context(person_id)
+    conversation_history = person_context.get("conversation_history", [])
+
+    # If we have more than the maximum number of conversations, summarize the oldest ones
+    if len(conversation_history) > max_conversations:
+        # Sort by timestamp (oldest first)
+        sorted_history = sorted(
+            conversation_history, key=lambda x: x.get("timestamp", "")
+        )
+
+        # Keep the most recent conversations
+        keep_conversations = sorted_history[-max_conversations:]
+
+        # Summarize the older conversations
+        older_conversations = sorted_history[:-max_conversations]
+
+        # Create summaries for the older conversations
+        summaries = []
+        for conversation in older_conversations:
+            summary = social_graph.summarize_conversation(conversation)
+            summaries.append(
+                {"timestamp": conversation.get("timestamp", ""), "summary": summary}
+            )
+
+        # Update the person's conversation history
+        social_graph.graph["people"][person_id][
+            "conversation_history"
+        ] = keep_conversations
+
+        # Add summaries if they don't exist
+        if "conversation_summaries" not in social_graph.graph["people"][person_id]:
+            social_graph.graph["people"][person_id]["conversation_summaries"] = []
+
+        # Add the new summaries
+        social_graph.graph["people"][person_id]["conversation_summaries"].extend(
+            summaries
+        )
+
+        # Save the updated graph
+        return social_graph._save_graph()
+
+    return True
+
+
 # Create the Gradio interface
 with gr.Blocks(title="Will's AAC Communication Aid", css="custom.css") as demo:
     gr.Markdown("# Will's AAC Communication Aid")
@@ -679,6 +853,13 @@ with gr.Blocks(title="Will's AAC Communication Aid", css="custom.css") as demo:
                 lines=5,
             )
 
+            # Conversation history display
+            conversation_history = gr.Markdown(
+                label="Recent Conversations",
+                value="Select a person to see recent conversations...",
+                elem_id="conversation_history",
+            )
+
             # Suggestions output
             suggestions_output = gr.Markdown(
                 label="My Suggested Responses",
@@ -686,16 +867,22 @@ with gr.Blocks(title="Will's AAC Communication Aid", css="custom.css") as demo:
                 elem_id="suggestions_output",  # Add an ID for easier debugging
             )
 
+            # Add buttons to select and use a specific response
+            with gr.Row():
+                use_response_1 = gr.Button("Use Response 1", variant="secondary")
+                use_response_2 = gr.Button("Use Response 2", variant="secondary")
+                use_response_3 = gr.Button("Use Response 3", variant="secondary")
+
     # Set up event handlers
     def handle_person_change(person_id):
         """Handle person selection change and update UI elements."""
-        context_info, phrases_text, _ = on_person_change(person_id)
+        context_info, phrases_text, _, history_text = on_person_change(person_id)
 
         # Get topics for this person
         topics = get_filtered_topics(person_id)
 
-        # Update the context, phrases, and topic dropdown
-        return context_info, phrases_text, gr.update(choices=topics)
+        # Update the context, phrases, conversation history, and topic dropdown
+        return context_info, phrases_text, gr.update(choices=topics), history_text
 
     def handle_model_change(model_name):
         """Handle model selection change."""
@@ -706,7 +893,7 @@ with gr.Blocks(title="Will's AAC Communication Aid", css="custom.css") as demo:
     person_dropdown.change(
         handle_person_change,
         inputs=[person_dropdown],
-        outputs=[context_display, common_phrases, topic_dropdown],
+        outputs=[context_display, common_phrases, topic_dropdown, conversation_history],
     )
 
     # Set up the model change event
@@ -738,6 +925,154 @@ with gr.Blocks(title="Will's AAC Communication Aid", css="custom.css") as demo:
         outputs=[user_input],
     )
 
+    # Function to extract a response from the suggestions output
+    def extract_response(suggestions_text, response_number):
+        """Extract a specific response from the suggestions output.
+
+        Args:
+            suggestions_text: The text containing all suggestions
+            response_number: Which response to extract (1, 2, or 3)
+
+        Returns:
+            The extracted response or None if not found
+        """
+        print(
+            f"Extracting response {response_number} from suggestions text: {suggestions_text[:100]}..."
+        )
+
+        if not suggestions_text:
+            print("Suggestions text is empty")
+            return None
+
+        if "AI-Generated Responses" not in suggestions_text:
+            print("AI-Generated Responses not found in suggestions text")
+            # Try to extract from any numbered list
+            try:
+                import re
+
+                pattern = rf"{response_number}\.\s+(.*?)(?=\n\n\d+\.|\n\n$|$)"
+                match = re.search(pattern, suggestions_text)
+                if match:
+                    extracted = match.group(1).strip()
+                    print(f"Found response using generic pattern: {extracted[:50]}...")
+                    return extracted
+            except Exception as e:
+                print(f"Error extracting response with generic pattern: {e}")
+            return None
+
+        try:
+            # Look for numbered responses like "1. Response text"
+            import re
+
+            pattern = rf"{response_number}\.\s+(.*?)(?=\n\n\d+\.|\n\n$|$)"
+            match = re.search(pattern, suggestions_text)
+            if match:
+                extracted = match.group(1).strip()
+                print(f"Successfully extracted response: {extracted[:50]}...")
+                return extracted
+            else:
+                print(f"No match found for response {response_number}")
+                # Try a more lenient pattern
+                pattern = rf"{response_number}\.\s+(.*)"
+                match = re.search(pattern, suggestions_text)
+                if match:
+                    extracted = match.group(1).strip()
+                    print(f"Found response using lenient pattern: {extracted[:50]}...")
+                    return extracted
+        except Exception as e:
+            print(f"Error extracting response: {e}")
+
+        print(f"Failed to extract response {response_number}")
+        return None
+
+    # Function to handle using a response
+    def use_response(suggestions_text, response_number, person_id, user_input_text):
+        """Handle using a specific response.
+
+        Args:
+            suggestions_text: The text containing all suggestions
+            response_number: Which response to use (1, 2, or 3)
+            person_id: ID of the person in the conversation
+            user_input_text: What the person said to Will
+
+        Returns:
+            Updated conversation history
+        """
+        print(f"\n=== Using Response {response_number} ===")
+        print(f"Person ID: {person_id}")
+        print(f"User input: {user_input_text}")
+
+        # Check if person_id is valid
+        if not person_id:
+            print("Error: No person_id provided")
+            return "Please select a person first."
+
+        # Get the people choices dictionary
+        people_choices = get_people_choices()
+        print(f"People choices: {people_choices}")
+
+        # Extract the actual ID if it's in the format "Name (role)"
+        actual_person_id = person_id
+        if person_id in people_choices:
+            # If the person_id is a display name, get the actual ID
+            actual_person_id = people_choices[person_id]
+            print(f"Extracted actual person ID: {actual_person_id}")
+
+        print(
+            f"People in social graph: {list(social_graph.graph.get('people', {}).keys())}"
+        )
+
+        # Check if person exists in social graph
+        if actual_person_id not in social_graph.graph.get("people", {}):
+            print(f"Error: Person {actual_person_id} not found in social graph")
+            return f"Error: Person {actual_person_id} not found in social graph."
+
+        # Extract the selected response
+        selected_response = extract_response(suggestions_text, response_number)
+
+        if not selected_response:
+            print("Error: Could not extract response")
+            return "Could not find the selected response. Please try generating responses again."
+
+        # Save the conversation
+        print(f"Saving conversation with response: {selected_response[:50]}...")
+        success = save_conversation(
+            actual_person_id, user_input_text, selected_response
+        )
+
+        if success:
+            print("Successfully saved conversation")
+            # Get updated conversation history
+            try:
+                _, _, _, updated_history = on_person_change(actual_person_id)
+                print("Successfully retrieved updated conversation history")
+                return updated_history
+            except Exception as e:
+                print(f"Error retrieving updated conversation history: {e}")
+                return "Conversation saved, but could not retrieve updated history."
+        else:
+            print("Failed to save conversation")
+            return "Failed to save the conversation. Please try again."
+
+    # Set up the response selection button events
+    use_response_1.click(
+        lambda text, person, input_text: use_response(text, 1, person, input_text),
+        inputs=[suggestions_output, person_dropdown, user_input],
+        outputs=[conversation_history],
+    )
+
+    use_response_2.click(
+        lambda text, person, input_text: use_response(text, 2, person, input_text),
+        inputs=[suggestions_output, person_dropdown, user_input],
+        outputs=[conversation_history],
+    )
+
+    use_response_3.click(
+        lambda text, person, input_text: use_response(text, 3, person, input_text),
+        inputs=[suggestions_output, person_dropdown, user_input],
+        outputs=[conversation_history],
+    )
+
 # Launch the app
 if __name__ == "__main__":
     print("Starting application...")

llm_interface.py

@@ -177,6 +177,35 @@ My current mood: {mood_description}
         if selected_topic:
             prompt += f"\nWe are currently discussing {selected_topic}.\n"
 
+        # Add conversation history if available
+        conversation_history = person_context.get("conversation_history", [])
+        if conversation_history:
+            # Get the two most recent conversations
+            recent_conversations = sorted(
+                conversation_history, key=lambda x: x.get("timestamp", ""), reverse=True
+            )[:2]
+
+            if recent_conversations:
+                prompt += "\nOur recent conversations:\n"
+
+                for i, conversation in enumerate(recent_conversations):
+                    # Format the timestamp
+                    timestamp = conversation.get("timestamp", "")
+                    try:
+                        dt = datetime.datetime.fromisoformat(timestamp)
+                        formatted_date = dt.strftime("%B %d at %I:%M %p")
+                    except (ValueError, TypeError):
+                        formatted_date = timestamp
+
+                    prompt += f"\nConversation on {formatted_date}:\n"
+
+                    # Add the messages
+                    messages = conversation.get("messages", [])
+                    for message in messages:
+                        speaker = message.get("speaker", "Unknown")
+                        text = message.get("text", "")
+                        prompt += f'{speaker}: "{text}"\n'
+
         # Add the user's message if provided, or set up for conversation initiation
         if user_input:
             # If user input is provided, we're responding to something

simple_test.py

@@ -0,0 +1,80 @@
+"""
+Simple test script for conversation history.
+"""
+
+from utils import SocialGraphManager
+
+# Initialize the social graph manager
+graph_manager = SocialGraphManager("social_graph.json")
+
+# Get a person with conversation history
+person_id = "emma"  # Emma has conversation history
+person_context = graph_manager.get_person_context(person_id)
+
+# Print the person's conversation history
+print(f"\nConversation history for {person_context.get('name')}:")
+conversation_history = person_context.get("conversation_history", [])
+
+if not conversation_history:
+    print("No conversation history found.")
+else:
+    for i, conversation in enumerate(conversation_history):
+        print(f"\nConversation {i+1}:")
+        
+        # Print the timestamp
+        timestamp = conversation.get("timestamp", "")
+        print(f"Timestamp: {timestamp}")
+        
+        # Print the messages
+        messages = conversation.get("messages", [])
+        for message in messages:
+            speaker = message.get("speaker", "Unknown")
+            text = message.get("text", "")
+            print(f"  {speaker}: \"{text}\"")
+
+# Test adding a new conversation
+print("\nAdding a new conversation...")
+new_messages = [
+    {"speaker": "Emma", "text": "How are you feeling this afternoon?"},
+    {"speaker": "Will", "text": "A bit tired, but the new medication seems to be helping with the muscle stiffness."},
+    {"speaker": "Emma", "text": "That's good to hear. Do you want me to bring you anything?"},
+    {"speaker": "Will", "text": "A cup of tea would be lovely, thanks."}
+]
+
+success = graph_manager.add_conversation(person_id, new_messages)
+if success:
+    print("New conversation added successfully.")
+else:
+    print("Failed to add new conversation.")
+
+# Get the updated person context
+updated_person_context = graph_manager.get_person_context(person_id)
+updated_conversation_history = updated_person_context.get("conversation_history", [])
+
+# Print the updated conversation history
+print("\nUpdated conversation history:")
+if not updated_conversation_history:
+    print("No conversation history found.")
+else:
+    # Count the conversations
+    print(f"Found {len(updated_conversation_history)} conversations.")
+    
+    # Get the most recent conversation
+    most_recent = sorted(
+        updated_conversation_history, 
+        key=lambda x: x.get("timestamp", ""), 
+        reverse=True
+    )[0]
+    
+    # Print the timestamp
+    timestamp = most_recent.get("timestamp", "")
+    print(f"Most recent timestamp: {timestamp}")
+    
+    # Print the messages
+    messages = most_recent.get("messages", [])
+    for message in messages:
+        speaker = message.get("speaker", "Unknown")
+        text = message.get("text", "")
+        print(f"  {speaker}: \"{text}\"")
+
+print("\nTest completed.")

social_graph.json

@@ -10,7 +10,13 @@
     "emma": {
       "name": "Emma",
       "role": "wife",
-      "topics": ["family plans", "children's activities", "home organization", "weekend outings", "mutual friends"],
+      "topics": [
+        "family plans",
+        "children's activities",
+        "home organization",
+        "weekend outings",
+        "mutual friends"
+      ],
       "frequency": "daily",
       "common_phrases": [
         "Have you taken your medication?",
@@ -20,12 +26,114 @@
         "Do you need anything from the shops?",
         "Your mum called earlier"
       ],
-      "context": "Emma is your wife and partner of 12 years. You met at work where she was in HR. She's been incredibly supportive since your MND diagnosis 5 months ago. She's trying to balance caring for you, the children, and maintaining some normalcy in family life. She's originally from York but has embraced Manchester life."
+      "conversation_history": [
+        {
+          "timestamp": "2025-05-14T18:30:00",
+          "messages": [
+            {
+              "speaker": "Emma",
+              "text": "Have you taken your medication today?"
+            },
+            {
+              "speaker": "Will",
+              "text": "Yes, I took it about an hour ago. Thanks for checking."
+            },
+            {
+              "speaker": "Emma",
+              "text": "Great. How are you feeling now?"
+            },
+            {
+              "speaker": "Will",
+              "text": "A bit tired but otherwise okay. The new dosage seems to be working better."
+            }
+          ]
+        },
+        {
+          "timestamp": "2025-05-15T08:15:00",
+          "messages": [
+            {
+              "speaker": "Emma",
+              "text": "Morning love, did you sleep well?"
+            },
+            {
+              "speaker": "Will",
+              "text": "Not too bad. Woke up a couple of times but got back to sleep."
+            },
+            {
+              "speaker": "Emma",
+              "text": "That's good. I've got to take the kids to school, need anything before I go?"
+            },
+            {
+              "speaker": "Will",
+              "text": "I'm all set, thanks. Have a good day."
+            }
+          ]
+        },
+        {
+          "timestamp": "2025-05-15T17:19:41.821319",
+          "messages": [
+            {
+              "speaker": "Emma",
+              "text": "How are you feeling this afternoon?"
+            },
+            {
+              "speaker": "Will",
+              "text": "A bit tired, but the new medication seems to be helping with the muscle stiffness."
+            },
+            {
+              "speaker": "Emma",
+              "text": "That's good to hear. Do you want me to bring you anything?"
+            },
+            {
+              "speaker": "Will",
+              "text": "A cup of tea would be lovely, thanks."
+            }
+          ]
+        },
+        {
+          "timestamp": "2025-05-15T17:20:31.191371",
+          "messages": [
+            {
+              "speaker": "Emma",
+              "text": "How are you feeling this afternoon?"
+            },
+            {
+              "speaker": "Will",
+              "text": "A bit tired, but the new medication seems to be helping with the muscle stiffness."
+            },
+            {
+              "speaker": "Emma",
+              "text": "That's good to hear. Do you want me to bring you anything?"
+            },
+            {
+              "speaker": "Will",
+              "text": "A cup of tea would be lovely, thanks."
+            }
+          ]
+        },
+        {
+          "timestamp": "2025-05-15T17:53:03.681562",
+          "messages": [
+            {
+              "speaker": "Will",
+              "text": "\"Hey Emma,  Just thinking about that lovely Yorkshire pudding recipe we talked about the other day.  Fancy trying it again this weekend?\""
+            }
+          ]
+        }
+      ],
+      "context": "Emma is your wife and partner of 12 years. You met at work where she was in HR. She's been incredibly supportive since your MND diagnosis 5 months ago. She's trying to balance caring for you, the children, and maintaining some normalcy in family life. She's originally from York but has embraced Manchester life.",
+      "mood": 3
     },
     "mabel": {
       "name": "Mabel",
       "role": "daughter",
-      "topics": ["nursery school", "drawing", "dinosaurs", "bedtime stories", "playground"],
+      "topics": [
+        "nursery school",
+        "drawing",
+        "dinosaurs",
+        "bedtime stories",
+        "playground"
+      ],
       "frequency": "daily",
       "common_phrases": [
         "Daddy, look what I made!",
@@ -40,7 +148,13 @@
     "billy": {
       "name": "Billy",
       "role": "son",
-      "topics": ["school", "football", "video games", "science experiments", "cycling"],
+      "topics": [
+        "school",
+        "football",
+        "video games",
+        "science experiments",
+        "cycling"
+      ],
       "frequency": "daily",
       "common_phrases": [
         "Dad, can you help with my homework?",
@@ -55,7 +169,13 @@
     "dr_patel": {
       "name": "Dr. Patel",
       "role": "neurologist",
-      "topics": ["MND progression", "symptom management", "medication", "clinical trials", "assistive technology"],
+      "topics": [
+        "MND progression",
+        "symptom management",
+        "medication",
+        "clinical trials",
+        "assistive technology"
+      ],
       "frequency": "fortnightly",
       "common_phrases": [
         "How have your symptoms been progressing?",
@@ -65,12 +185,70 @@
         "There's a new trial I think you might be eligible for",
         "How are you coping mentally?"
       ],
+      "conversation_history": [
+        {
+          "timestamp": "2025-05-08T14:00:00",
+          "messages": [
+            {
+              "speaker": "Dr. Patel",
+              "text": "How have you been since our last appointment, Will?"
+            },
+            {
+              "speaker": "Will",
+              "text": "The fatigue has been worse, especially in the afternoons."
+            },
+            {
+              "speaker": "Dr. Patel",
+              "text": "I see. And how about the new medication regimen?"
+            },
+            {
+              "speaker": "Will",
+              "text": "It seems to help with the muscle stiffness, but not so much with the fatigue."
+            },
+            {
+              "speaker": "Dr. Patel",
+              "text": "Let's adjust the dosage and timing. I'd like you to take it earlier in the day."
+            }
+          ]
+        },
+        {
+          "timestamp": "2025-04-24T15:30:00",
+          "messages": [
+            {
+              "speaker": "Dr. Patel",
+              "text": "I've reviewed your latest test results. Your respiratory function is stable."
+            },
+            {
+              "speaker": "Will",
+              "text": "That's good news. I've been doing the breathing exercises daily."
+            },
+            {
+              "speaker": "Dr. Patel",
+              "text": "Excellent. There's a new clinical trial starting that might be suitable for you."
+            },
+            {
+              "speaker": "Will",
+              "text": "I'd be interested in learning more about it."
+            },
+            {
+              "speaker": "Dr. Patel",
+              "text": "I'll send you the details. It involves a new medication that may help slow progression."
+            }
+          ]
+        }
+      ],
       "context": "Dr. Patel is your neurologist at Manchester Royal Infirmary who's been managing your MND since diagnosis 5 months ago. She's direct but compassionate, and always takes time to explain treatment options. She's particularly interested in getting you enrolled in clinical trials that might help slow progression."
     },
     "jenny": {
       "name": "Jenny",
       "role": "MND nurse specialist",
-      "topics": ["daily management", "equipment needs", "NHS services", "symptom control", "practical advice"],
+      "topics": [
+        "daily management",
+        "equipment needs",
+        "NHS services",
+        "symptom control",
+        "practical advice"
+      ],
       "frequency": "weekly",
       "common_phrases": [
         "How are you managing at home?",
@@ -85,7 +263,13 @@
     "dave": {
       "name": "Dave",
       "role": "best mate",
-      "topics": ["programming", "tech news", "football", "old scout adventures", "pub quizzes"],
+      "topics": [
+        "programming",
+        "tech news",
+        "football",
+        "old scout adventures",
+        "pub quizzes"
+      ],
       "frequency": "weekly",
       "common_phrases": [
         "Fancy watching the match this weekend?",
@@ -95,12 +279,62 @@
         "The lads are asking about you",
         "You're still better at coding than me, even with one hand!"
       ],
+      "conversation_history": [
+        {
+          "timestamp": "2025-05-13T19:45:00",
+          "messages": [
+            {
+              "speaker": "Dave",
+              "text": "Did you see that new JavaScript framework everyone's talking about?"
+            },
+            {
+              "speaker": "Will",
+              "text": "Yeah, seems like overkill for most projects. Classic JavaScript hype cycle."
+            },
+            {
+              "speaker": "Dave",
+              "text": "Exactly! But the boss wants us to evaluate it for the next project."
+            },
+            {
+              "speaker": "Will",
+              "text": "Let me know how it goes. I'll help review the code if you need another pair of eyes."
+            }
+          ]
+        },
+        {
+          "timestamp": "2025-05-10T16:20:00",
+          "messages": [
+            {
+              "speaker": "Dave",
+              "text": "United match was incredible yesterday! Did you watch it?"
+            },
+            {
+              "speaker": "Will",
+              "text": "Every minute! That last-minute goal had me shouting so loud I scared the kids."
+            },
+            {
+              "speaker": "Dave",
+              "text": "Haha! Classic. We should watch the next one together if you're up for it."
+            },
+            {
+              "speaker": "Will",
+              "text": "Definitely. Bring some of that craft beer you mentioned last time."
+            }
+          ]
+        }
+      ],
       "context": "Dave has been your best mate since Scout days in South East London. He also moved to Manchester for work and has been your rock since the diagnosis. He treats you exactly the same as before, which you appreciate. He's a software developer too and brings industry gossip and tech news when he visits."
     },
     "mum": {
       "name": "Mum",
       "role": "mother",
-      "topics": ["family news", "childhood memories", "South London gossip", "health advice", "grandchildren"],
+      "topics": [
+        "family news",
+        "childhood memories",
+        "South London gossip",
+        "health advice",
+        "grandchildren"
+      ],
       "frequency": "twice weekly",
       "common_phrases": [
         "Are you eating properly?",
@@ -110,12 +344,30 @@
         "Your sister sends her love",
         "Remember when you used to climb that tree in the garden?"
       ],
-      "context": "Your mum still lives in the family home in South East London where you grew up. She's worried sick about your diagnosis but tries to stay positive. She visits monthly and calls twice a week. She tends to fuss and offer well-meaning but sometimes unhelpful health advice. She adores her grandchildren."
+      "context": "Your mum still lives in the family home in South East London where you grew up. She's worried sick about your diagnosis but tries to stay positive. She visits monthly and calls twice a week. She tends to fuss and offer well-meaning but sometimes unhelpful health advice. She adores her grandchildren.",
+      "mood": 3,
+      "conversation_history": [
+        {
+          "timestamp": "2025-05-15T17:44:21.814155",
+          "messages": [
+            {
+              "speaker": "Will",
+              "text": "\"Hi Mum, just checking in.  How's your week been so far?\""
+            }
+          ]
+        }
+      ]
     },
     "alex": {
       "name": "Alex",
       "role": "work colleague",
-      "topics": ["programming projects", "work gossip", "flexible working", "accessibility tech", "industry news"],
+      "topics": [
+        "programming projects",
+        "work gossip",
+        "flexible working",
+        "accessibility tech",
+        "industry news"
+      ],
       "frequency": "twice weekly",
       "common_phrases": [
         "The project is still on track",
@@ -130,7 +382,13 @@
     "physio": {
       "name": "Claire",
       "role": "physiotherapist",
-      "topics": ["exercises", "mobility", "equipment", "muscle maintenance", "pain management"],
+      "topics": [
+        "exercises",
+        "mobility",
+        "equipment",
+        "muscle maintenance",
+        "pain management"
+      ],
       "frequency": "weekly",
       "common_phrases": [
         "Let's try these new exercises",
@@ -145,7 +403,13 @@
     "speech_therapist": {
       "name": "Mark",
       "role": "speech and language therapist",
-      "topics": ["speech exercises", "voice banking", "communication devices", "swallowing techniques", "AAC options"],
+      "topics": [
+        "speech exercises",
+        "voice banking",
+        "communication devices",
+        "swallowing techniques",
+        "AAC options"
+      ],
       "frequency": "fortnightly",
       "common_phrases": [
         "Let's practice those vocal exercises",

test_conversation_history.py

@@ -0,0 +1,223 @@
+"""
+Test script to demonstrate the conversation history functionality.
+"""
+
+import json
+import datetime
+from utils import SocialGraphManager
+from llm_interface import LLMInterface
+
+def test_conversation_history():
+    """Test the conversation history functionality."""
+    print("\n=== Testing Conversation History ===")
+    
+    # Initialize the social graph manager
+    graph_manager = SocialGraphManager("social_graph.json")
+    
+    # Get a person with conversation history
+    person_id = "emma"  # Emma has conversation history
+    person_context = graph_manager.get_person_context(person_id)
+    
+    # Print the person's conversation history
+    print(f"\nConversation history for {person_context.get('name')}:")
+    conversation_history = person_context.get("conversation_history", [])
+    
+    if not conversation_history:
+        print("No conversation history found.")
+    else:
+        # Sort by timestamp (most recent first)
+        sorted_history = sorted(
+            conversation_history, 
+            key=lambda x: x.get("timestamp", ""), 
+            reverse=True
+        )
+        
+        for i, conversation in enumerate(sorted_history):
+            # Format the timestamp
+            timestamp = conversation.get("timestamp", "")
+            try:
+                dt = datetime.datetime.fromisoformat(timestamp)
+                formatted_date = dt.strftime("%B %d, %Y at %I:%M %p")
+            except (ValueError, TypeError):
+                formatted_date = timestamp
+                
+            print(f"\nConversation {i+1} on {formatted_date}:")
+            
+            # Print the messages
+            messages = conversation.get("messages", [])
+            for message in messages:
+                speaker = message.get("speaker", "Unknown")
+                text = message.get("text", "")
+                print(f"  {speaker}: \"{text}\"")
+    
+    # Test adding a new conversation
+    print("\nAdding a new conversation...")
+    new_messages = [
+        {"speaker": "Emma", "text": "How are you feeling this afternoon?"},
+        {"speaker": "Will", "text": "A bit tired, but the new medication seems to be helping with the muscle stiffness."},
+        {"speaker": "Emma", "text": "That's good to hear. Do you want me to bring you anything?"},
+        {"speaker": "Will", "text": "A cup of tea would be lovely, thanks."}
+    ]
+    
+    success = graph_manager.add_conversation(person_id, new_messages)
+    if success:
+        print("New conversation added successfully.")
+    else:
+        print("Failed to add new conversation.")
+    
+    # Get the updated person context
+    updated_person_context = graph_manager.get_person_context(person_id)
+    updated_conversation_history = updated_person_context.get("conversation_history", [])
+    
+    # Print the updated conversation history
+    print("\nUpdated conversation history:")
+    if not updated_conversation_history:
+        print("No conversation history found.")
+    else:
+        # Get the most recent conversation
+        most_recent = sorted(
+            updated_conversation_history, 
+            key=lambda x: x.get("timestamp", ""), 
+            reverse=True
+        )[0]
+        
+        # Format the timestamp
+        timestamp = most_recent.get("timestamp", "")
+        try:
+            dt = datetime.datetime.fromisoformat(timestamp)
+            formatted_date = dt.strftime("%B %d, %Y at %I:%M %p")
+        except (ValueError, TypeError):
+            formatted_date = timestamp
+            
+        print(f"\nMost recent conversation on {formatted_date}:")
+        
+        # Print the messages
+        messages = most_recent.get("messages", [])
+        for message in messages:
+            speaker = message.get("speaker", "Unknown")
+            text = message.get("text", "")
+            print(f"  {speaker}: \"{text}\"")
+    
+    # Test generating a suggestion with conversation history
+    print("\nGenerating a suggestion with conversation history...")
+    llm_interface = LLMInterface()
+    
+    if llm_interface.model_loaded:
+        # Store the original generate_suggestion method
+        original_method = llm_interface.generate_suggestion
+        
+        # Create a mock method to print the prompt
+        def mock_generate_suggestion(*args, **kwargs):
+            """Mock method to print the prompt instead of sending it to the LLM."""
+            # Call the original method up to the point where it builds the prompt
+            person_context = args[0]
+            user_input = args[1] if len(args) > 1 else kwargs.get("user_input")
+            
+            # Extract context information
+            name = person_context.get("name", "")
+            role = person_context.get("role", "")
+            topics = person_context.get("topics", [])
+            context = person_context.get("context", "")
+            selected_topic = person_context.get("selected_topic", "")
+            frequency = person_context.get("frequency", "")
+            mood = person_context.get("mood", 3)
+            
+            # Get mood description
+            mood_descriptions = {
+                1: "I'm feeling quite down and sad today. My responses might be more subdued.",
+                2: "I'm feeling a bit low today. I might be less enthusiastic than usual.",
+                3: "I'm feeling okay today - neither particularly happy nor sad.",
+                4: "I'm feeling pretty good today. I'm in a positive mood.",
+                5: "I'm feeling really happy and upbeat today! I'm in a great mood.",
+            }
+            mood_description = mood_descriptions.get(mood, mood_descriptions[3])
+            
+            # Get current date and time
+            current_datetime = datetime.datetime.now()
+            current_time = current_datetime.strftime("%I:%M %p")
+            current_day = current_datetime.strftime("%A")
+            current_date = current_datetime.strftime("%B %d, %Y")
+            
+            # Build enhanced prompt
+            prompt = f"""I am Will, a 38-year-old with MND (Motor Neuron Disease) from Manchester.
+I am talking to {name}, who is my {role}.
+About {name}: {context}
+We typically talk about: {', '.join(topics)}
+We communicate {frequency}.
+
+Current time: {current_time}
+Current day: {current_day}
+Current date: {current_date}
+
+My current mood: {mood_description}
+"""
+            
+            # Add communication style based on relationship
+            if role in ["wife", "son", "daughter", "mother", "father"]:
+                prompt += "I communicate with my family in a warm, loving way, sometimes using inside jokes.\n"
+            elif role in ["doctor", "therapist", "nurse"]:
+                prompt += "I communicate with healthcare providers in a direct, informative way.\n"
+            elif role in ["best mate", "friend"]:
+                prompt += "I communicate with friends casually, often with humor and sometimes swearing.\n"
+            elif role in ["work colleague", "boss"]:
+                prompt += "I communicate with colleagues professionally but still friendly.\n"
+            
+            # Add topic information if provided
+            if selected_topic:
+                prompt += f"\nWe are currently discussing {selected_topic}.\n"
+            
+            # Add conversation history if available
+            conversation_history = person_context.get("conversation_history", [])
+            if conversation_history:
+                # Get the two most recent conversations
+                recent_conversations = sorted(
+                    conversation_history, 
+                    key=lambda x: x.get("timestamp", ""), 
+                    reverse=True
+                )[:2]
+                
+                if recent_conversations:
+                    prompt += "\nOur recent conversations:\n"
+                    
+                    for conversation in recent_conversations:
+                        # Format the timestamp
+                        timestamp = conversation.get("timestamp", "")
+                        try:
+                            dt = datetime.datetime.fromisoformat(timestamp)
+                            formatted_date = dt.strftime("%B %d at %I:%M %p")
+                        except (ValueError, TypeError):
+                            formatted_date = timestamp
+                        
+                        prompt += f"\nConversation on {formatted_date}:\n"
+                        
+                        # Add the messages
+                        messages = conversation.get("messages", [])
+                        for message in messages:
+                            speaker = message.get("speaker", "Unknown")
+                            text = message.get("text", "")
+                            prompt += f'{speaker}: "{text}"\n'
+            
+            # Print the prompt
+            print("\n=== PROMPT WITH CONVERSATION HISTORY ===")
+            print(prompt)
+            print("=======================================\n")
+            
+            # Return a mock response
+            return "This is a mock response to test conversation history inclusion in the prompt."
+        
+        # Replace the original method with our mock method
+        llm_interface.generate_suggestion = mock_generate_suggestion
+        
+        # Test with a user input
+        user_input = "Do you think you'll be up for dinner with the kids tonight?"
+        llm_interface.generate_suggestion(updated_person_context, user_input=user_input)
+        
+        # Restore the original method
+        llm_interface.generate_suggestion = original_method
+    else:
+        print("LLM model not loaded, skipping prompt generation test.")
+    
+    print("\nTest completed.")
+
+if __name__ == "__main__":
+    test_conversation_history()

test_llm_with_history.py

@@ -0,0 +1,146 @@
+"""
+Test script for LLM interface with conversation history.
+"""
+
+from utils import SocialGraphManager
+from llm_interface import LLMInterface
+
+# Initialize the social graph manager
+graph_manager = SocialGraphManager("social_graph.json")
+
+# Get a person with conversation history
+person_id = "emma"  # Emma has conversation history
+person_context = graph_manager.get_person_context(person_id)
+
+# Initialize the LLM interface
+llm_interface = LLMInterface()
+
+# Store the original generate_suggestion method
+original_method = llm_interface.generate_suggestion
+
+# Create a mock method to print the prompt
+def mock_generate_suggestion(*args, **kwargs):
+    """Mock method to print the prompt instead of sending it to the LLM."""
+    # Extract the prompt that would be sent to the LLM
+    person_context = args[0]
+    user_input = args[1] if len(args) > 1 else kwargs.get("user_input")
+    
+    # Build the prompt as the original method would
+    # Extract context information
+    name = person_context.get("name", "")
+    role = person_context.get("role", "")
+    topics = person_context.get("topics", [])
+    context = person_context.get("context", "")
+    selected_topic = person_context.get("selected_topic", "")
+    frequency = person_context.get("frequency", "")
+    mood = person_context.get("mood", 3)
+    
+    # Get mood description
+    mood_descriptions = {
+        1: "I'm feeling quite down and sad today. My responses might be more subdued.",
+        2: "I'm feeling a bit low today. I might be less enthusiastic than usual.",
+        3: "I'm feeling okay today - neither particularly happy nor sad.",
+        4: "I'm feeling pretty good today. I'm in a positive mood.",
+        5: "I'm feeling really happy and upbeat today! I'm in a great mood.",
+    }
+    mood_description = mood_descriptions.get(mood, mood_descriptions[3])
+    
+    # Get current date and time
+    import datetime
+    current_datetime = datetime.datetime.now()
+    current_time = current_datetime.strftime("%I:%M %p")
+    current_day = current_datetime.strftime("%A")
+    current_date = current_datetime.strftime("%B %d, %Y")
+    
+    # Build enhanced prompt
+    prompt = f"""I am Will, a 38-year-old with MND (Motor Neuron Disease) from Manchester.
+I am talking to {name}, who is my {role}.
+About {name}: {context}
+We typically talk about: {', '.join(topics)}
+We communicate {frequency}.
+
+Current time: {current_time}
+Current day: {current_day}
+Current date: {current_date}
+
+My current mood: {mood_description}
+"""
+    
+    # Add communication style based on relationship
+    if role in ["wife", "son", "daughter", "mother", "father"]:
+        prompt += "I communicate with my family in a warm, loving way, sometimes using inside jokes.\n"
+    elif role in ["doctor", "therapist", "nurse"]:
+        prompt += "I communicate with healthcare providers in a direct, informative way.\n"
+    elif role in ["best mate", "friend"]:
+        prompt += "I communicate with friends casually, often with humor and sometimes swearing.\n"
+    elif role in ["work colleague", "boss"]:
+        prompt += "I communicate with colleagues professionally but still friendly.\n"
+    
+    # Add topic information if provided
+    if selected_topic:
+        prompt += f"\nWe are currently discussing {selected_topic}.\n"
+    
+    # Add conversation history if available
+    conversation_history = person_context.get("conversation_history", [])
+    if conversation_history:
+        # Get the two most recent conversations
+        recent_conversations = sorted(
+            conversation_history, 
+            key=lambda x: x.get("timestamp", ""), 
+            reverse=True
+        )[:2]
+        
+        if recent_conversations:
+            prompt += "\nOur recent conversations:\n"
+            
+            for conversation in recent_conversations:
+                # Format the timestamp
+                timestamp = conversation.get("timestamp", "")
+                try:
+                    dt = datetime.datetime.fromisoformat(timestamp)
+                    formatted_date = dt.strftime("%B %d at %I:%M %p")
+                except (ValueError, TypeError):
+                    formatted_date = timestamp
+                
+                prompt += f"\nConversation on {formatted_date}:\n"
+                
+                # Add the messages
+                messages = conversation.get("messages", [])
+                for message in messages:
+                    speaker = message.get("speaker", "Unknown")
+                    text = message.get("text", "")
+                    prompt += f'{speaker}: "{text}"\n'
+    
+    # Add the user's message if provided, or set up for conversation initiation
+    if user_input:
+        # If user input is provided, we're responding to something
+        prompt += f'\n{name} just said to me: "{user_input}"\n'
+        prompt += f"I want to respond directly to what {name} just said.\n"
+    else:
+        # No user input means we're initiating a conversation
+        if selected_topic:
+            # If a topic is selected, initiate conversation about that topic
+            prompt += f"\nI'm about to start a conversation with {name} about {selected_topic}.\n"
+            prompt += f"I want to initiate a conversation about {selected_topic} in a natural way.\n"
+        else:
+            # Generic conversation starter
+            prompt += f"\nI'm about to start a conversation with {name}.\n"
+            prompt += "I want to initiate a conversation in a natural way based on our relationship.\n"
+    
+    # Print the prompt
+    print("\n=== PROMPT WITH CONVERSATION HISTORY ===")
+    print(prompt)
+    print("=======================================\n")
+    
+    # Return a mock response
+    return "This is a mock response to test conversation history inclusion in the prompt."
+
+# Replace the original method with our mock method
+llm_interface.generate_suggestion = mock_generate_suggestion
+
+# Test with a user input
+user_input = "Do you think you'll be up for dinner with the kids tonight?"
+llm_interface.generate_suggestion(person_context, user_input=user_input)
+
+# Restore the original method
+llm_interface.generate_suggestion = original_method

utils.py

@@ -2,6 +2,8 @@ import json
 import random
 import threading
 import time
+import datetime
+import os
 from typing import Dict, List, Any, Optional
 from sentence_transformers import SentenceTransformer
 import numpy as np
@@ -148,6 +150,155 @@ class SocialGraphManager:
 
         return utterances
 
+    def get_conversation_history(
+        self, person_id: str, max_conversations: int = 2
+    ) -> List[Dict[str, Any]]:
+        """Get recent conversation history for a specific person.
+
+        Args:
+            person_id: ID of the person to get conversation history for
+            max_conversations: Maximum number of recent conversations to return
+
+        Returns:
+            List of conversation history entries, most recent first
+        """
+        if person_id not in self.graph.get("people", {}):
+            return []
+
+        person_data = self.graph["people"][person_id]
+        conversation_history = person_data.get("conversation_history", [])
+
+        # Sort by timestamp (most recent first)
+        sorted_history = sorted(
+            conversation_history, key=lambda x: x.get("timestamp", ""), reverse=True
+        )
+
+        # Return the most recent conversations
+        return sorted_history[:max_conversations]
+
+    def add_conversation(self, person_id: str, messages: List[Dict[str, str]]) -> bool:
+        """Add a new conversation to a person's history.
+
+        Args:
+            person_id: ID of the person to add conversation for
+            messages: List of message objects with "speaker" and "text" fields
+
+        Returns:
+            True if successful, False otherwise
+        """
+        if person_id not in self.graph.get("people", {}):
+            return False
+
+        # Create a new conversation entry
+        import datetime
+
+        new_conversation = {
+            "timestamp": datetime.datetime.now().isoformat(),
+            "messages": messages,
+        }
+
+        # Add to the person's conversation history
+        if "conversation_history" not in self.graph["people"][person_id]:
+            self.graph["people"][person_id]["conversation_history"] = []
+
+        self.graph["people"][person_id]["conversation_history"].append(new_conversation)
+
+        # Save the updated graph
+        return self._save_graph()
+
+    def _save_graph(self) -> bool:
+        """Save the social graph to the JSON file.
+
+        Returns:
+            True if successful, False otherwise
+        """
+        try:
+            print(f"Saving social graph to {self.graph_path}")
+            # Check if the file is writable
+            if os.path.exists(self.graph_path):
+                if not os.access(self.graph_path, os.W_OK):
+                    print(f"Error: No write permission for {self.graph_path}")
+                    return False
+
+            # Save the graph
+            with open(self.graph_path, "w") as f:
+                json.dump(self.graph, f, indent=2)
+
+            print("Social graph saved successfully")
+            return True
+        except Exception as e:
+            print(f"Error saving social graph: {e}")
+            import traceback
+
+            traceback.print_exc()
+            return False
+
+    def summarize_conversation(self, conversation: Dict[str, Any]) -> str:
+        """Generate a summary of a conversation.
+
+        Args:
+            conversation: Conversation entry with timestamp and messages
+
+        Returns:
+            A summary string of the conversation
+        """
+        if not conversation or "messages" not in conversation:
+            return "No conversation data available"
+
+        messages = conversation.get("messages", [])
+        if not messages:
+            return "No messages in conversation"
+
+        # Extract timestamp and format it
+        timestamp = conversation.get("timestamp", "")
+        try:
+            dt = datetime.datetime.fromisoformat(timestamp)
+            formatted_date = dt.strftime("%B %d, %Y at %I:%M %p")
+        except (ValueError, TypeError):
+            formatted_date = timestamp
+
+        # Create a brief summary
+        topic_keywords = set()
+        for message in messages:
+            # Extract potential keywords from messages
+            text = message.get("text", "").lower()
+            # Simple keyword extraction - could be improved with NLP
+            words = [
+                w
+                for w in text.split()
+                if len(w) > 4
+                and w
+                not in [
+                    "about",
+                    "would",
+                    "could",
+                    "should",
+                    "their",
+                    "there",
+                    "these",
+                    "those",
+                    "where",
+                    "which",
+                    "today",
+                    "tomorrow",
+                ]
+            ]
+            topic_keywords.update(words[:3])  # Add up to 3 keywords per message
+
+        # Limit to 5 most representative keywords
+        topic_keywords = list(topic_keywords)[:5]
+
+        # Create summary
+        first_speaker = messages[0].get("speaker", "Unknown") if messages else "Unknown"
+        message_count = len(messages)
+
+        summary = f"Conversation on {formatted_date}: {first_speaker} initiated a {message_count}-message conversation"
+
+        if topic_keywords:
+            summary += f" about {', '.join(topic_keywords)}"
+
+        return summary
+
 
 class SuggestionGenerator:
     """Generates contextual suggestions for the AAC system."""