Response is generator fix and other polish

app.py

@@ -786,64 +786,78 @@ with gr.Blocks() as demo:
     </div>""")
 
     with gr.Accordion("Config and Asset Assistance - Click to open", open=False):
-        with gr.Accordion("Themes to consider before you attempt config", open=False):
+        with gr.Accordion("Guidance / Themes to consider before you attempt config (Currently Qwen 0.5B)", open=False):
             gr.HTML("Copy paste any old config to llm and ask to remix is the easiest <br>To bake 'Moral of the story' in you have to be very deliberate")
             gr.HTML("UI can be media and all items can have media")
-            with gr.Accordion("Qwen 0.5B as Space Guide Tests", open=False):
-                with gr.Tab("General FAQ Attempt"):
-                    FAQMainOutput = gr.TextArea(placeholder='Output will show here', value='')
-                    FAQCustomButtonInput = gr.TextArea(lines=1, placeholder='Prompt goes here')
-
-                    for category_name, category_prompts in FAQAllprompts.items():
-                        with gr.Accordion(f"General {category_name} Pattern based", open=False):
-                            with gr.Group():
-                                for index, (prompt, _) in enumerate(category_prompts):
-                                    button = gr.Button(prompt)
-                                    button.click(llmguide_generate_response, inputs=[FAQCustomButtonInput, gr.State(index), gr.State(category_name)], outputs=FAQMainOutput)
-
-                with gr.Tab("General RAG (Pathfinder?) Attempt"):
-                    gr.HTML("https://huggingface.co/spaces/mteb/leaderboard - Source for SOTA - current using all-MiniLM-L6-v2")
-                    gr.HTML("Placeholder for weak RAG Type Charcter interaction test aka input for JSON 'Knowledge Base' Input")
-                    gr.Interface(
-                        fn=process_query,
-                        inputs=[
-                            gr.Textbox(lines=2, placeholder="Enter your question here..."),
-                            gr.Checkbox(label="Use RAG"),
-                            gr.Checkbox(label="Stream output")
-                        ],
-                        outputs=[
-                            gr.Textbox(label="Generated Response"),
-                            gr.Textbox(label="Tokens per second"),
-                            gr.Textbox(label="RAM Usage"),
-                            gr.Textbox(label="Referenced Documents")
-                        ],
-                        title="RAG/Non-RAG Q&A System",
-                        description="Ask a question with or without using RAG. The response is generated using a GPU-accelerated model. RAM usage and referenced document IDs (for RAG) are logged."
-                    )
+            with gr.Accordion("Old Ideas to merge", open=False):
+                gr.HTML("Random Scenario / Song to 'full game' manual or auto is end goal ")
+                gr.HTML("Componets (outside Code Support for Config): Decisions (and context explanation), Nested Sections, Media (Especially to affect decisions), Replayability (GTA and Tekken type mechanics in text form), Theme integration (Modified Varibles that affect UI or config order)") 
+                gr.HTML("Existing Games eg. GTA Heists - Same Map with overlapping branching narratives, Battlefront - Elites amongst Commoners, Tekken Casino (one mistake = 1/2 or 1/3 of your Resources) and Turn based: 'Tactics' type nintendo games, Chess (and any other tile based game) ")
+                gr.HTML("Existing Game Rules for text - Cyberpunk RED, ")
+                gr.HTML("Community playthrough = Tally of players choices, Random item placed in a random location - first person to get it wins, Survival by location or characters met")
+                gr.HTML("Some Kinds of game skeletons ideas - Timelines, Graph as State machine paths, Economy ecosystem")
+                gr.HTML("One prompt to be used to test models - <br>Please make 10 python lists for the types of media files and their purposes in a game and then use those lists to random generate a timeline of 20 items when the function is called <br>Great next suggest ways to improve this function to create better timelines")
+                with gr.Tab("Main areas of considerations"):
+                    gr.HTML("")
+                    with gr.Tab("Structural Inspirations"):
+                        gr.HTML("GTA Heists - Replayability and stakes, Tekken - 2/3 mistakes = lost round ")
+                        gr.HTML("Sports Scores, ")
+                    with gr.Tab("Themes"):
+                        gr.HTML("")
+            with gr.Tab("General FAQ Attempt"):
+                FAQMainOutput = gr.TextArea(placeholder='Output will show here', value='')
+                FAQCustomButtonInput = gr.TextArea(lines=1, placeholder='Prompt goes here')
+
+                for category_name, category_prompts in FAQAllprompts.items():
+                    with gr.Accordion(f"General {category_name} Pattern based", open=False):
+                        with gr.Group():
+                            for index, (prompt, _) in enumerate(category_prompts):
+                                button = gr.Button(prompt)
+                                button.click(llmguide_generate_response, inputs=[FAQCustomButtonInput, gr.State(index), gr.State(category_name)], outputs=FAQMainOutput)
+
+            with gr.Tab("General RAG (Pathfinder?) Attempt"):
+                gr.HTML("https://huggingface.co/spaces/mteb/leaderboard - Source for SOTA - current using all-MiniLM-L6-v2")
+                gr.HTML("Placeholder for weak RAG Type Charcter interaction test aka input for JSON 'Knowledge Base' Input")
+                gr.Interface(
+                    fn=process_query,
+                    inputs=[
+                        gr.Textbox(lines=2, placeholder="Enter your question here..."),
+                        gr.Checkbox(label="Use RAG"),
+                        gr.Checkbox(label="Stream output")
+                    ],
+                    outputs=[
+                        gr.Textbox(label="Generated Response"),
+                        gr.Textbox(label="Tokens per second"),
+                        gr.Textbox(label="RAM Usage"),
+                        gr.Textbox(label="Referenced Documents")
+                    ],
+                    title="RAG/Non-RAG Q&A System",
+                    description="Ask a question with or without using RAG. The response is generated using a GPU-accelerated model. RAM usage and referenced document IDs (for RAG) are logged."
+                )
 
-                with gr.Tab("Any Request to Qwen2-0.5B"):
-                    gr.HTML("Placeholder for https://huggingface.co/h2oai/h2o-danube3-500m-chat-GGUF and https://huggingface.co/OuteAI/Lite-Mistral-150M-v2-Instruct as alternative")
-                    gr.HTML("https://huggingface.co/spaces/HuggingFaceTB/SmolLM-135M-Instruct-WebGPU 125 mdeol to be tested as alternative (and all up to 1.5b - how to delte a model in your code?) - Need to go over the dataset to see how to prompt it - https://huggingface.co/datasets/HuggingFaceTB/smollm-corpus ")
-                    gr.HTML("Placeholder for qwen 2 72b as alternative use checkbox and gradio client api call")
-                    gr.Markdown("# Qwen-0.5B-Instruct Language Model")
-                    gr.Markdown("This demo uses the Qwen-0.5B-Instruct model to generate responses based on your input.")
-                    gr.HTML("Example prompts: <br>I am writing a story about a chef. please write dishes to appear on the menu. <br>What are the most common decisions that a chef story would include? <br>What are the kinds problems that a chef story would include? <br>What are the kinds of out of reach goals that a chef story would include? <br>Continue this config - Paste any complete block of the config")
-                    
-                    with gr.Row():
-                        with gr.Column():
-                            llmguide_prompt = gr.Textbox(lines=2, placeholder="Enter your prompt here...")
-                            llmguide_stream_checkbox = gr.Checkbox(label="Enable streaming")
-                            llmguide_submit_button = gr.Button("Generate")
-                        
-                        with gr.Column():
-                            llmguide_output = gr.Textbox(lines=10, label="Generated Response")
-                            llmguide_tokens_per_second = gr.Textbox(label="Tokens per Second")
+            with gr.Tab("Any Request to Qwen2-0.5B"):
+                gr.HTML("Placeholder for https://huggingface.co/h2oai/h2o-danube3-500m-chat-GGUF and https://huggingface.co/OuteAI/Lite-Mistral-150M-v2-Instruct as alternative")
+                gr.HTML("https://huggingface.co/spaces/HuggingFaceTB/SmolLM-135M-Instruct-WebGPU 125 mdeol to be tested as alternative (and all up to 1.5b - how to delte a model in your code?) - Need to go over the dataset to see how to prompt it - https://huggingface.co/datasets/HuggingFaceTB/smollm-corpus ")
+                gr.HTML("Placeholder for qwen 2 72b as alternative use checkbox and gradio client api call")
+                gr.Markdown("# Qwen-0.5B-Instruct Language Model")
+                gr.Markdown("This demo uses the Qwen-0.5B-Instruct model to generate responses based on your input.")
+                gr.HTML("Example prompts: <br>I am writing a story about a chef. please write dishes to appear on the menu. <br>What are the most common decisions that a chef story would include? <br>What are the kinds problems that a chef story would include? <br>What are the kinds of out of reach goals that a chef story would include? <br>Continue this config - Paste any complete block of the config")
+                
+                with gr.Row():
+                    with gr.Column():
+                        llmguide_prompt = gr.Textbox(lines=2, placeholder="Enter your prompt here...")
+                        llmguide_stream_checkbox = gr.Checkbox(label="Enable streaming")
+                        llmguide_submit_button = gr.Button("Generate")
                     
-                    llmguide_submit_button.click(
-                        llmguide_generate_response,
-                        inputs=[llmguide_prompt, llmguide_stream_checkbox],
-                        outputs=[llmguide_output, llmguide_tokens_per_second],
-                    )
+                    with gr.Column():
+                        llmguide_output = gr.Textbox(lines=10, label="Generated Response")
+                        llmguide_tokens_per_second = gr.Textbox(label="Tokens per Second")
+                
+                llmguide_submit_button.click(
+                    llmguide_generate_response,
+                    inputs=[llmguide_prompt, llmguide_stream_checkbox],
+                    outputs=[llmguide_output, llmguide_tokens_per_second],
+                )
 
         with gr.Accordion("New Config Proto Assist - Trying to abstract the process into one worflow is beyond me so multiple paths to goal (config) is the aim now", open=False):
             with gr.Tab("Linear - List to Empty Config with Edit support"):
@@ -905,8 +919,11 @@ with gr.Blocks() as demo:
                         prev_msgs.append(prompt)
                         formatted_prompt = LinPEWFformat_prompt(prompt, prev_msgs)
                         response = llmguide_generate_response(formatted_prompt)
-                        prev_msgs.append(f"Response: {response}")
-                        return response
+                        full_response = ""
+                        for chunk in response:
+                            full_response += chunk
+                        prev_msgs.append(f"Response: {full_response}")
+                        return full_response
                     
                     outputbtn = gr.Button(f"Generate {i}").click(
                         fn=LinPEWF_update_and_generate, 
@@ -920,6 +937,7 @@ with gr.Blocks() as demo:
             #with gr.Accordion("Decisions / Timeline Creation to Story to Config Conversation", open=False):
             with gr.Tab("Branching - Decisions / Timeline Creation to Story to Config Conversation"):
                 gr.HTML("Structures for interesting timeline progression")
+                gr.HTML("Claude Artifacts to illustrate nested structure brainstorms - <br> https://claude.site/artifacts/4a910d81-1541-49f4-8531-4f27fe56cd1e <br> https://claude.site/artifacts/265e9242-2093-46e1-9011-ed6ad938be90?fullscreen=false <br> ")
                 gr.HTML("Placeholder - Considerations - Story from the perspective of Main character or NPC in the LLM genereated story")
                 mermaideditoriframebtn = gr.Button("Load Mermaid Editor")
                 mermaideditoriframe = gr.HTML("")
@@ -927,24 +945,6 @@ with gr.Blocks() as demo:
                 with gr.Accordion("Mermaid Structures - click to open", open=False):
                     for key, item in mermaidstorystructures.items():
                         gr.Code(item, label=key)
-        
-        with gr.Accordion("Old Ideas to merge", open=False):
-            gr.HTML("Random Scenario / Song to 'full game' manual or auto is end goal ")
-            gr.HTML("Componets (outside Code Support for Config): Decisions (and context explanation), Nested Sections, Media (Especially to affect decisions), Replayability (GTA and Tekken type mechanics in text form), Theme integration (Modified Varibles that affect UI or config order)") 
-            gr.HTML("Existing Games eg. GTA Heists - Same Map with overlapping branching narratives, Battlefront - Elites amongst Commoners, Tekken Casino (one mistake = 1/2 or 1/3 of your Resources) and Turn based: 'Tactics' type nintendo games, Chess (and any other tile based game) ")
-            gr.HTML("Existing Game Rules for text - Cyberpunk RED, ")
-            gr.HTML("Community playthrough = Tally of players choices, Random item placed in a random location - first person to get it wins, Survival by location or characters met")
-            gr.HTML("Some Kinds of game skeletons ideas - Timelines, Graph as State machine paths, Economy ecosystem")
-            gr.HTML("One prompt to be used to test models - <br>Please make 10 python lists for the types of media files and their purposes in a game and then use those lists to random generate a timeline of 20 items when the function is called <br>Great next suggest ways to improve this function to create better timelines")
-            with gr.Tab("Main areas of considerations"):
-                with gr.Tab("Mermaid Graphs and Nesting"):
-                    gr.HTML("Claude Artifacts to illustrate nested structure brainstorms - <br> https://claude.site/artifacts/4a910d81-1541-49f4-8531-4f27fe56cd1e <br> https://claude.site/artifacts/265e9242-2093-46e1-9011-ed6ad938be90?fullscreen=false <br> ")
-                    gr.HTML("")
-                with gr.Tab("Structural Inspirations"):
-                    gr.HTML("GTA Heists - Replayability and stakes, Tekken - 2/3 mistakes = lost round ")
-                    gr.HTML("Sports Scores, ")
-                with gr.Tab("Themes"):
-                    gr.HTML("")
 
         with gr.Accordion("Existing Config Crafting Progression - click to open", open=False):
             with gr.Accordion("Test for config to gradio components order - ignore for now", open=False ):
@@ -1050,83 +1050,81 @@ Creating more diverse paths through the game""")
             with gr.Tab(experimetal_config_name):    
                 gr.Code(json.dumps(experimetal_config, default=lambda o: o.__dict__, indent=2), label=experimetal_config_name) #str(experimetal_config)
 
-    with gr.Tab("Test and Edit Config"):
-        gr.HTML("The main issue is frequent changes add more chances for bugs in how - manual and auto refer mainly to ensuring correct JSON format ")
-        with gr.Tab("Full Manual - Test Example State Machine"): 
-            with gr.Tab("Config Without Assets"):
-                with gr.Row():
-                    with gr.Column(scale=2):
-                        gr.Markdown("# Text-based Adventure Game")
-
-                        description = gr.Textbox(label="Current Situation", lines=4, value=initgameinfo[0])
-                        choices = gr.Radio(label="Your Choices", choices=initgameinfo[1])
-                        submit_btn = gr.Button("Make Choice")
-                        game_log = gr.Textbox(label="Game Log", lines=20, value=initgameinfo[2])
-                        game_session = gr.State(value=initgameinfo[3])
-                        submit_btn.click(
-                            make_choice,
-                            inputs=[choices, game_session],
-                            outputs=[description, choices, game_log, game_session]
+    with gr.Tab("Test and Edit Config"): 
+        with gr.Tab("Manual - Config Without Assets"):
+            with gr.Row():
+                with gr.Column(scale=2):
+                    gr.Markdown("# Text-based Adventure Game")
+
+                    description = gr.Textbox(label="Current Situation", lines=4, value=initgameinfo[0])
+                    choices = gr.Radio(label="Your Choices", choices=initgameinfo[1])
+                    submit_btn = gr.Button("Make Choice")
+                    game_log = gr.Textbox(label="Game Log", lines=20, value=initgameinfo[2])
+                    game_session = gr.State(value=initgameinfo[3])
+                    submit_btn.click(
+                        make_choice,
+                        inputs=[choices, game_session],
+                        outputs=[description, choices, game_log, game_session]
+                    )
+                with gr.Column(scale=1):
+                    gr.Markdown("# Debugging")
+                    error_box = gr.Textbox(label="Path Errors", lines=4, value=path_errors)
+                    with gr.Accordion("Config (Game Spoiler and Example for llm to remix)", open=False):
+                        custom_config = gr.Textbox(label="Custom Configuration (JSON)", value=json.dumps(all_states, default=lambda o: o.__dict__, indent=2), lines=8)
+                        custom_configbtn = gr.Button("Load Custom Config")
+
+                        custom_configbtn.click(
+                            load_game,
+                            inputs=[custom_config],
+                            outputs=[error_box, game_log, description, choices, game_session, custom_config]
                         )
-                    with gr.Column(scale=1):
-                        gr.Markdown("# Debugging")
-                        error_box = gr.Textbox(label="Path Errors", lines=4, value=path_errors)
-                        with gr.Accordion("Config (Game Spoiler and Example for llm to remix)", open=False):
-                            custom_config = gr.Textbox(label="Custom Configuration (JSON)", value=json.dumps(all_states, default=lambda o: o.__dict__, indent=2), lines=8)
-                            custom_configbtn = gr.Button("Load Custom Config")
-
-                            custom_configbtn.click(
-                                load_game,
-                                inputs=[custom_config],
-                                outputs=[error_box, game_log, description, choices, game_session, custom_config]
-                            )
-            with gr.Tab("Config With Assets"):
-                gr.HTML("Placeholder as not complete yet (still only text, current issue is how to switch gradio output, maybe output is gr.Group and we constantly add the appropriate gr for each file type? What about multi file types on one state?)")
-                with gr.Row():
-                    with gr.Column(scale=2):
-                        gr.Markdown("# Text-based Adventure Game")
-
-                        wadescription = gr.Textbox(label="Current Situation", lines=4, value=initgameinfo[0])
-                        wamediabool = gr.State(value=True)
-                        wamedia = gr.State(["testmedia/Stable Audio - Raindrops, output.wav"])
-
-                        @gr.render(inputs=wamedia)
-                        def dynamic_with_media(media_items):
-                            print(media_items)
-                            with gr.Group() as wamediagrouping: 
-                                gr.HTML("Placeholder to load all media tests - still need to test clearing media on ")
-                                if media_items == []:
-                                    gr.Markdown("No media items to display.")
-                                else:
-                                    for item in media_items:
-                                        render = create_media_component(item)
-                            
-                            return wamediagrouping
+        with gr.Tab("Manual - Config With Assets"):
+            gr.HTML("Placeholder as not complete yet (still only text, current issue is how to switch gradio output, maybe output is gr.Group and we constantly add the appropriate gr for each file type? What about multi file types on one state?)")
+            with gr.Row():
+                with gr.Column(scale=2):
+                    gr.Markdown("# Text-based Adventure Game")
+
+                    wadescription = gr.Textbox(label="Current Situation", lines=4, value=initgameinfo[0])
+                    wamediabool = gr.State(value=True)
+                    wamedia = gr.State(["testmedia/Stable Audio - Raindrops, output.wav"])
+
+                    @gr.render(inputs=wamedia)
+                    def dynamic_with_media(media_items):
+                        print(media_items)
+                        with gr.Group() as wamediagrouping: 
+                            gr.HTML("Placeholder to load all media tests - still need to test clearing media on ")
+                            if media_items == []:
+                                gr.Markdown("No media items to display.")
+                            else:
+                                for item in media_items:
+                                    render = create_media_component(item)
                         
-                        wachoices = gr.Radio(label="Your Choices", choices=initgameinfo[1])
-                        wasubmit_btn = gr.Button("Make Choice")
-                        wagame_log = gr.Textbox(label="Game Log", lines=20, value=initgameinfo[2])
-                        wagame_session = gr.State(value=initgameinfo[3])
-                        wasubmit_btn.click(
-                            make_choice,
-                            inputs=[wachoices, wagame_session, wamediabool],
-                            outputs=[wadescription, wachoices, wagame_log, wagame_session, wamedia]
+                        return wamediagrouping
+                    
+                    wachoices = gr.Radio(label="Your Choices", choices=initgameinfo[1])
+                    wasubmit_btn = gr.Button("Make Choice")
+                    wagame_log = gr.Textbox(label="Game Log", lines=20, value=initgameinfo[2])
+                    wagame_session = gr.State(value=initgameinfo[3])
+                    wasubmit_btn.click(
+                        make_choice,
+                        inputs=[wachoices, wagame_session, wamediabool],
+                        outputs=[wadescription, wachoices, wagame_log, wagame_session, wamedia]
+                    )
+                with gr.Column(scale=1):
+                    gr.Markdown("# Debugging")
+                    waerror_box = gr.Textbox(label="Path Errors", lines=4, value=path_errors)
+                    with gr.Accordion("Config (Game Spoiler and Example for llm to remix)", open=False):
+                        wacustom_config = gr.Textbox(label="Custom Configuration (JSON)", value=json.dumps(all_states, default=lambda o: o.__dict__, indent=2), lines=8)
+                        wacustom_configbtn = gr.Button("Load Custom Config")
+
+                        wacustom_configbtn.click(
+                            load_game,
+                            inputs=[wacustom_config, wamediabool],
+                            outputs=[waerror_box, wagame_log, wadescription, wachoices, wacustom_config, wagame_session, wamedia]
                         )
-                    with gr.Column(scale=1):
-                        gr.Markdown("# Debugging")
-                        waerror_box = gr.Textbox(label="Path Errors", lines=4, value=path_errors)
-                        with gr.Accordion("Config (Game Spoiler and Example for llm to remix)", open=False):
-                            wacustom_config = gr.Textbox(label="Custom Configuration (JSON)", value=json.dumps(all_states, default=lambda o: o.__dict__, indent=2), lines=8)
-                            wacustom_configbtn = gr.Button("Load Custom Config")
-
-                            wacustom_configbtn.click(
-                                load_game,
-                                inputs=[wacustom_config, wamediabool],
-                                outputs=[waerror_box, wagame_log, wadescription, wachoices, wacustom_config, wagame_session, wamedia]
-                            )
-                
-            with gr.Tab("Config With Minimal 3D considered"):
-                gr.HTML("Placeholder for Config with 3D assets")
+            
+        with gr.Tab("Manual - Config With Minimal 3D considered"):
+            gr.HTML("Placeholder for Config with 3D assets")
 
         with gr.Tab("Semi-Auto - Edit config while playing game"):
             gr.HTML("-- Incomplete -- (also generate_story_and_timeline refactored)  Issue here is updating all variables <br> Current problem is passing values from rendered items to the config box <br>Generate Timline also makes config without mmedia key <br>Need a way have dropdowns for the filelist and transitions eg. changing transitions must auto update choices <br>Config to components has hardcoded variables based on the auto gen so changes break it")
@@ -1336,9 +1334,8 @@ Creating more diverse paths through the game""")
 
     with gr.Tab("Basic Game Engine Mechanics"):
         gr.HTML("Placeholder for explanations of Player and Game Session")
-
-    with gr.Tab("LLM play testing"):
-        gr.HTML("LLM can read the contents in full and give critiques but they can also play the game if you make a api interface - gradio allows this in the form of gradio client but you can also reroute the user inputs to function calling")
+        with gr.Tab("LLM play testing"):
+            gr.HTML("LLM can read the contents in full and give critiques but they can also play the game if you make a api interface - gradio allows this in the form of gradio client but you can also reroute the user inputs to function calling")
 
     with gr.Tab("Custom JS Config Creator"):
         gr.HTML("-- Incomplete -- Companion Space for zerogpu / client api workflow planning for a way to send a zip to the Basic Game Engine at the bottom of https://huggingface.co/spaces/KwabsHug/TestSvelteStatic (Also to test how much can be done majority on cpu)")
@@ -1414,8 +1411,9 @@ Creating more diverse paths through the game""")
         with gr.Tab("Advanced Config Creator"):
             gr.HTML("Config with More than text and images")
                     
-    with gr.Tab("LLM/Robotics as custom controllers Considerations"):
+    with gr.Tab("LLM/Robotics as custom controllers"):
         gr.HTML("Controls changed the scope of the game eg. mouse vs keyboard vs console controller vs remote vs touch screen <br>LLM can be vision/surveilance based controler (eg. MGS/GTA camera gauged by an actual camera in real life) or it can be a companion (offline/off console game progrssion ideas)")
+        gr.HTML("https://huggingface.co/posts/thomwolf/809364796644704")
         gr.HTML("Robotics - https://github.com/OpenTeleVision/TeleVision https://www.stereolabs.com/")
 
     with gr.Tab("Other Considerations"):