Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+assets/background.jpg filter=lfs diff=lfs merge=lfs -text
+transformers-4.57.0.dev0-py3-none-any.whl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,14 +1,12 @@
 ---
 title: Hollow Knight Helper
-emoji: 🐠
+emoji: 🕸️
 colorFrom: red
 colorTo: gray
 sdk: gradio
 sdk_version: 5.44.1
 app_file: app.py
 pinned: false
-license: cc-by-sa-3.0
-short_description: Hollow Knight Helper
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+An example chatbot using [Gradio](https://gradio.app) and Gemma.

app.py

@@ -0,0 +1,274 @@
+import gradio as gr
+import requests
+import os
+import pickle
+import spaces
+from bs4 import BeautifulSoup
+from html_to_markdown import convert_to_markdown
+from huggingface_hub import login
+from sentence_transformers import SentenceTransformer
+from transformers import pipeline, TextIteratorStreamer
+from threading import Thread
+from tqdm import tqdm
+
+# --- 1. CONFIGURATION ---
+# Centralized place for all settings and constants.
+
+# Hugging Face & Model Configuration
+HF_TOKEN = os.getenv('HF_TOKEN')
+EMBEDDING_MODEL_ID = "google/embeddinggemma-300M"
+LLM_MODEL_ID = "google/gemma-3-12B-it"
+
+# Data Source Configuration
+BASE_URL = "https://hollowknight.wiki"
+
+# Hollow Knight Boss Data
+ENTRY_POINT_HOLLOW_KNIGHT = "/w/Category:Bosses_(Hollow_Knight)"
+CACHE_FILE_HOLLOW_KNIGHT = "hollow_knight_boss.pkl"
+
+# Silksong Boss Data
+ENTRY_POINT_SILKSONG = "/w/Category:Bosses_(Silksong)"
+CACHE_FILE_SILKSONG = "silksong_boss.pkl"
+
+# Gradio App Configuration
+DEFAULT_SIMILARITY_THRESHOLD = 0.5
+DEFAULT_MESSAGE_NO_MATCH = "I'm sorry, I can't find a relevant document to answer that question. Try asking about a specific boss in Hollow Knight."
+
+
+# --- 2. HELPER FUNCTIONS ---
+# Reusable functions for web scraping and data processing.
+
+def get_html(url: str) -> str:
+    """Fetches HTML content from a URL."""
+    try:
+        response = requests.get(url)
+        response.raise_for_status()  # Raises an HTTPError for bad responses (4xx or 5xx)
+        return response.text
+    except requests.exceptions.RequestException as e:
+        print(f"Error fetching {url}: {e}")
+        return ""
+
+def find_wiki_links(html_content: str) -> list[str]:
+    """Parses HTML to find all boss links within the 'mw-pages' div."""
+    soup = BeautifulSoup(html_content, 'html.parser')
+    mw_pages_div = soup.find('div', id='mw-pages')
+    if not mw_pages_div:
+        return []
+    return [a['href'] for a in mw_pages_div.find_all('a', href=True)]
+
+def get_markdown_from_url(url: str) -> str:
+    """Fetches and converts a webpage's content to Markdown."""
+    html = get_html(url)
+    if not html:
+        return ""
+    soup = BeautifulSoup(html, 'html.parser')
+    # Assuming convert_to_markdown correctly processes the soup object
+    return convert_to_markdown(soup)
+
+
+# --- 3. DATA PROCESSING & CACHING ---
+# Scrapes data and generates embeddings, using a cache to avoid re-running.
+
+def load_or_process_source(entry_point: str, cache_file: str, label: str, embedding_model):
+    """
+    Loads processed data from a cache file if it exists. Otherwise, scrapes,
+    processes, generates embeddings, and saves to the cache.
+    """
+    if os.path.exists(cache_file):
+        print(f"✅ Found cache for {label}. Loading data from '{cache_file}'...")
+        with open(cache_file, 'rb') as f:
+            return pickle.load(f)
+
+    print(f"ℹ️ No cache for {label}. Starting data scraping and processing...")
+    main_page_html = get_html(BASE_URL + entry_point)
+    extracted_links = find_wiki_links(main_page_html)
+
+    contents = {"titles": [], "texts": [], "embeddings": []}
+
+    for doc_path in tqdm(extracted_links, desc=f"Processing {label} Pages"):
+        full_url = BASE_URL + doc_path
+        original_text = get_markdown_from_url(full_url)
+
+        # Trim text from the "References" section onwards for cleaner context
+        text = original_text.split("References\n----------\n", 1)[0].strip()
+
+        if text:
+            contents["titles"].append(doc_path.split('/')[-1])
+            contents["texts"].append(text)
+            # Generate and add embedding
+            embedding = embedding_model.encode(text, prompt=f"title: {doc_path.split('/')[-1]} | text: ")
+            contents["embeddings"].append(embedding)
+
+    print(f"✅ {label} processing complete. Saving data to '{cache_file}'...")
+    with open(cache_file, 'wb') as f:
+        pickle.dump(contents, f)
+
+    return contents
+
+
+# --- 4. CORE AI LOGIC ---
+# Functions for finding context and generating a response.
+
+def find_best_context(model, query: str, contents: dict, similarity_threshold: float):
+    """Finds the most relevant document text based on semantic similarity."""
+    if not query or not contents["embeddings"]:
+        return None
+        
+    query_embedding = model.encode(query, prompt_name="query")
+    similarities = model.similarity(query_embedding, contents["embeddings"])
+    
+    best_index = similarities.argmax().item()
+    best_score = similarities[0, best_index].item()
+
+    print(best_score)
+    if best_score >= similarity_threshold:
+        return contents["texts"][best_index]
+    return None
+
+context = None
+
+@spaces.GPU
+def respond(message: str, history: list, similarity_threshold: float):
+    """Generates a streaming response from the LLM based on the best context found."""
+    global context
+    if (context := find_best_context(embedding_model, message, combined_contents, similarity_threshold) or context):
+        # SUCCESS: A valid context was found and has been saved.
+        pass
+    else:
+        # FAILURE: No context is available.
+        yield DEFAULT_MESSAGE_NO_MATCH
+        return
+
+    system_prompt = f"Answer the following QUESTION based only on the CONTEXT provided. If the answer cannot be found in the CONTEXT, write \"I don't know.\"\n---\nCONTEXT:\n{context}\n"
+    user_prompt = f"QUESTION:\n{message}"
+
+    messages = [{"role": "system", "content": system_prompt}]
+    messages.extend(history)
+    messages.append({"role": "user", "content": user_prompt})
+
+    for item in messages:
+        print(item['role'])
+        print(item['content'])
+
+    streamer = TextIteratorStreamer(llm_pipeline.tokenizer, skip_prompt=True, skip_special_tokens=True)
+
+    thread = Thread(
+        target=llm_pipeline,
+        kwargs=dict(
+            text_inputs=messages,
+            streamer=streamer,
+            max_new_tokens=512,
+            do_sample=True,
+            top_p=0.95,
+        )
+    )
+    thread.start()
+
+    response = ""
+    for new_text in streamer:
+        response += new_text
+        yield response
+
+
+# --- 5. INITIALIZATION ---
+# Login, load models, and process data.
+
+print("Logging into Hugging Face Hub...")
+login(token=HF_TOKEN)
+
+print("Initializing embedding model...")
+embedding_model = SentenceTransformer(EMBEDDING_MODEL_ID)
+
+print("Initializing language model...")
+llm_pipeline = pipeline(
+    "text-generation",
+    model=LLM_MODEL_ID,
+    device_map="auto",
+    dtype="auto",
+)
+
+print("\n--- Processing Game Data ---")
+hk_contents = load_or_process_source(
+    ENTRY_POINT_HOLLOW_KNIGHT, CACHE_FILE_HOLLOW_KNIGHT, "Hollow Knight", embedding_model
+)
+silksong_contents = load_or_process_source(
+    ENTRY_POINT_SILKSONG, CACHE_FILE_SILKSONG, "Silksong", embedding_model
+)
+
+print("\nCombining data sources...")
+combined_contents = {
+    "titles": hk_contents["titles"] + silksong_contents["titles"],
+    "texts": hk_contents["texts"] + silksong_contents["texts"],
+    "embeddings": hk_contents["embeddings"] + silksong_contents["embeddings"],
+}
+print(f"✅ Total documents processed: {len(combined_contents['texts'])}")
+
+
+# --- 6. GRADIO UI ---
+# Defines the web interface for the chatbot.
+gr.set_static_paths(paths=["assets/"])
+
+# Theme and CSS for the Silksong aesthetic
+silksong_theme = gr.themes.Default(
+    primary_hue=gr.themes.colors.red,
+    secondary_hue=gr.themes.colors.zinc,
+    neutral_hue=gr.themes.colors.zinc,
+    font=[gr.themes.GoogleFont("IM Fell English"), "ui-sans-serif", "system-ui", "sans-serif"],
+)
+
+silksong_css="""
+.gradio-container {
+    background-image: linear-gradient(rgba(255,255,255, 0.5), rgba(255, 255, 255, 1.0)), url("/gradio_api/file=assets/background.jpg");
+    background-size: cover;
+    background-repeat: no-repeat;
+    background-position: center;
+}
+body.dark .gradio-container {
+    background-image: linear-gradient(rgba(0, 0, 0, 0.5), rgba(0, 0, 0, 1.0)), url("/gradio_api/file=assets/background.jpg");
+}
+.header-text { text-align: center; text-shadow: 2px 2px 5px #000; }
+.header-text h1 { font-size: 2.5em; color: #dc2626; }
+.dark .header-text { text-shadow: 2px 2px 5px #FFF; }
+.disclaimer { text-align: center; color: var(--body-text-color-subdued); font-size: 0.9em; padding: 20px; }
+.disclaimer ul { list-style: none; padding: 0; }
+.disclaimer a { color: #dc2626; }
+"""
+
+with gr.Blocks(theme=silksong_theme, css=silksong_css) as demo:
+    gr.HTML("""
+        <div class="header-text">
+            <h1>A Weaver's Counsel</h1>
+            <p>Speak, little traveler. What secrets of Pharloom do you seek?</p>
+            <p style="font-style: italic;">(Note: This bot currently only has knowledge about bosses)</p>
+        </div>
+    """)
+
+    gr.ChatInterface(
+        respond,
+        type="messages",
+        chatbot=gr.Chatbot(type="messages", label=LLM_MODEL_ID),
+        textbox=gr.Textbox(placeholder="Ask about the haunted kingdom...", container=False, submit_btn=True, scale=7),
+        additional_inputs=[
+            gr.Slider(minimum=0.1, maximum=1.0, value=DEFAULT_SIMILARITY_THRESHOLD, step=0.1, label="Similarity Threshold"),
+        ],
+        examples=[
+            ["Where can I find the Moorwing?", DEFAULT_SIMILARITY_THRESHOLD],
+            ["Who is the voice of Lace?", DEFAULT_SIMILARITY_THRESHOLD],
+            ["How can I beat the False Knight?", DEFAULT_SIMILARITY_THRESHOLD],
+            ["What achievement for Hornet Protector?", DEFAULT_SIMILARITY_THRESHOLD],
+        ],
+    )
+
+    gr.HTML("""
+        <div class="disclaimer">
+            <p><strong>Disclaimer:</strong></p>
+            <ul style="list-style: none; padding: 0;">
+                <li>This is a fan-made personal demonstration and not affiliated with any organization.<br>The bot is for entertainment purposes only.</li>
+                <li>Factual information is sourced from the <a href="https://hollowknight.wiki" target="_blank">Hollow Knight Wiki</a>.<br>Content is available under <a href="https://creativecommons.org/licenses/by-sa/3.0/" target="_blank">Commons Attribution-ShareAlike</a> unless otherwise noted.</li>
+                <li>Built by <a href="https://huggingface.co/bebechien" target="_blank">bebechien</a> with a 💖 for the world of Hollow Knight.</li>
+            </ul>
+        </div>
+    """)
+
+if __name__ == "__main__":
+    demo.launch()

hollow_knight_boss.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:608417424fb5f9670689cb318868bc19ddba6a524fa6df8f3d43c47393e65a13
+size 976095

requirements.txt

@@ -0,0 +1,5 @@
+accelerate
+beautifulsoup4
+html_to_markdown
+sentence-transformers
+git+https://github.com/huggingface/[email protected]

silksong_boss.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4e7d57a2234046677a429cae5143d55812de509660a12a5225afcf576e539b7
+size 66657