Merge pull request #829 from jmorganca/mattw/example-summarize-news

added python rag news summary

examples/python-rag-newssummary/README.md

@@ -0,0 +1,22 @@
+# News Summarizer
+
+This example goes through a series of steps:
+
+  1. You choose a topic area (e.g., "news", "NVidia", "music", etc.).
+  2. Gets the most recent articles on that topic from various sources.
+  3. Uses Ollama to summarize each article.
+  4. Creates chunks of sentences from each article.
+  5. Uses Sentence Transformers to generate embeddings for each of those chunks.
+  6. You enter a question regarding the summaries shown.
+  7. Uses Sentence Transformers to generate an embedding for that question.
+  8. Uses the embedded question to find the most similar chunks.
+  9. Feeds all that to Ollama to generate a good answer to your question based on these news articles.
+
+This example lets you pick from a few different topic areas, then summarize the most recent x articles for that topic. It then creates chunks of sentences from each article and then generates embeddings for each of those chunks.
+
+You can run the example like this:
+
+```bash
+pip install -r requirements.txt
+python summ.py
+```

examples/python-rag-newssummary/requirements.txt

@@ -0,0 +1,9 @@
+beautifulsoup4==4.12.2
+feedparser==6.0.10
+mattsollamatools==0.0.8
+newspaper3k==0.2.8
+nltk==3.8.1
+numpy==1.24.3
+Requests==2.31.0
+scikit_learn==1.3.0
+sentence_transformers==2.2.2

examples/python-rag-newssummary/summ.py

@@ -0,0 +1,86 @@
+import curses
+import json
+from utils import get_url_for_topic, topic_urls, menu, getUrls, get_summary, getArticleText, knn_search
+import requests
+from sentence_transformers import SentenceTransformer
+from mattsollamatools import chunker
+
+if __name__ == "__main__":
+    chosen_topic = curses.wrapper(menu)
+    print("Here is your news summary:\n")
+    urls = getUrls(chosen_topic, n=5)
+    model = SentenceTransformer('all-MiniLM-L6-v2')
+    allEmbeddings = []
+
+    for url in urls:
+      article={}
+      article['embeddings'] = []
+      article['url'] = url
+      text = getArticleText(url)
+      summary = get_summary(text)
+      chunks = chunker(text)  # Use the chunk_text function from web_utils
+      embeddings = model.encode(chunks)
+      for (chunk, embedding) in zip(chunks, embeddings):
+        item = {}
+        item['source'] = chunk
+        item['embedding'] = embedding.tolist()  # Convert NumPy array to list
+        item['sourcelength'] = len(chunk)
+        article['embeddings'].append(item)
+    
+      allEmbeddings.append(article)
+
+      print(f"{summary}\n")
+
+    
+    while True:
+      context = []
+      # Input a question from the user
+      question = input("Enter your question about the news, or type quit: ")
+
+      if question.lower() == 'quit':
+        break
+
+      # Embed the user's question
+      question_embedding = model.encode([question])
+
+      # Perform KNN search to find the best matches (indices and source text)
+      best_matches = knn_search(question_embedding, allEmbeddings, k=10)
+
+
+      sourcetext=""
+      for i, (index, source_text) in enumerate(best_matches, start=1):
+          sourcetext += f"{i}. Index: {index}, Source Text: {source_text}"
+
+      systemPrompt = f"Only use the following information to answer the question. Do not use anything else: {sourcetext}"
+
+      url = "http://localhost:11434/api/generate"
+
+      payload = {
+      "model": "mistral-openorca",
+      "prompt": question, 
+      "system": systemPrompt,
+      "stream": False, 
+      "context": context
+      }
+
+      # Convert the payload to a JSON string
+      payload_json = json.dumps(payload)
+
+      # Set the headers to specify JSON content
+      headers = {
+          "Content-Type": "application/json"
+      }
+
+      # Send the POST request
+      response = requests.post(url, data=payload_json, headers=headers)
+
+      # Check the response
+      if response.status_code == 200:
+          output = json.loads(response.text)
+          context = output['context']
+          print(output['response']+ "\n")
+          
+
+      else:
+          print(f"Request failed with status code {response.status_code}")
+

examples/python-rag-newssummary/utils.py

@@ -0,0 +1,108 @@
+import curses
+import feedparser
+import requests
+import unicodedata
+import json
+from newspaper import Article
+from bs4 import BeautifulSoup
+from nltk.tokenize import sent_tokenize, word_tokenize
+import numpy as np
+from sklearn.neighbors import NearestNeighbors
+from mattsollamatools import chunker
+
+# Create a dictionary to store topics and their URLs
+topic_urls = {
+    "Mac": "https://9to5mac.com/guides/mac/feed",
+    "News": "http://www.npr.org/rss/rss.php?id=1001",
+    "Nvidia": "https://nvidianews.nvidia.com/releases.xml",
+    "Raspberry Pi": "https://www.raspberrypi.com/news/feed/", 
+    "Music": "https://www.billboard.com/c/music/music-news/feed/"
+}
+
+# Use curses to create a menu of topics
+def menu(stdscr):
+    chosen_topic = get_url_for_topic(stdscr)  
+    url = topic_urls[chosen_topic] if chosen_topic in topic_urls else "Topic not found"
+    
+    stdscr.addstr(len(topic_urls) + 3, 0, f"Selected URL for {chosen_topic}: {url}")
+    stdscr.refresh()
+    
+    return chosen_topic
+
+# You have chosen a topic. Now return the url for that topic
+def get_url_for_topic(stdscr):
+    curses.curs_set(0)  # Hide the cursor
+    stdscr.clear()
+
+    stdscr.addstr(0, 0, "Choose a topic using the arrow keys (Press Enter to select):")
+
+    # Create a list of topics
+    topics = list(topic_urls.keys())
+    current_topic = 0
+
+    while True:
+        for i, topic in enumerate(topics):
+            if i == current_topic:
+                stdscr.addstr(i + 2, 2, f"> {topic}")
+            else:
+                stdscr.addstr(i + 2, 2, f"  {topic}")
+
+        stdscr.refresh()
+
+        key = stdscr.getch()
+
+        if key == curses.KEY_DOWN and current_topic < len(topics) - 1:
+            current_topic += 1
+        elif key == curses.KEY_UP and current_topic > 0:
+            current_topic -= 1
+        elif key == 10:  # Enter key
+            return topic_urls[topics[current_topic]]
+
+# Get the last N URLs from an RSS feed
+def getUrls(feed_url, n=20):
+    feed = feedparser.parse(feed_url)
+    entries = feed.entries[-n:]
+    urls = [entry.link for entry in entries]
+    return urls
+
+# Often there are a bunch of ads and menus on pages for a news article. This uses newspaper3k to get just the text of just the article.
+def getArticleText(url):
+  article = Article(url)
+  article.download()
+  article.parse()
+  return article.text
+
+def get_summary(text):
+  systemPrompt = "Write a concise summary of the text, return your responses with 5 lines that cover the key points of the text given."
+  prompt = text
+  
+  url = "http://localhost:11434/api/generate"
+
+  payload = {
+    "model": "mistral-openorca",
+    "prompt": prompt, 
+    "system": systemPrompt,
+    "stream": False
+  }
+  payload_json = json.dumps(payload)
+  headers = {"Content-Type": "application/json"}
+  response = requests.post(url, data=payload_json, headers=headers)
+
+  return json.loads(response.text)["response"]
+
+# Perform K-nearest neighbors (KNN) search
+def knn_search(question_embedding, embeddings, k=5):
+    X = np.array([item['embedding'] for article in embeddings for item in article['embeddings']])
+    source_texts = [item['source'] for article in embeddings for item in article['embeddings']]
+    
+    # Fit a KNN model on the embeddings
+    knn = NearestNeighbors(n_neighbors=k, metric='cosine')
+    knn.fit(X)
+    
+    # Find the indices and distances of the k-nearest neighbors
+    distances, indices = knn.kneighbors(question_embedding, n_neighbors=k)
+    
+    # Get the indices and source texts of the best matches
+    best_matches = [(indices[0][i], source_texts[indices[0][i]]) for i in range(k)]
+    
+    return best_matches