Moved GPT4all to chat plugin more

searx/plugins/chat.py

@@ -1,4 +1,6 @@
 from searx.search import SearchWithPlugins
+from pathlib import Path
+from gpt4all import GPT4All
 
 
 name = "Chat Plugin"
@@ -11,5 +13,29 @@ def post_search(request, search: SearchWithPlugins) -> None:
     search_request = search.search_query
     container = search.result_container
     # container.infoboxes.append(container.infoboxes[0])
-    container.chat_box = [{'chat_box': 'hello world'}]
-    container.chat_box[0]['content'] = 'some string that relates to your search query here' + f'\n{search_request.query}'
+    container.chat_box = {'chat_box': 'GPT4All'}
+    container.chat_box['content'] = 'Generating response to query: ' + f'\n{search_request.query}'
+
+def generate_chat_content(query):
+    model = GPT4All(model_name='gpt4all-falcon-q4_0.gguf',
+            model_path=(Path.cwd() / 'searx' / 'plugins'),
+            allow_download=False)
+
+    system_template = """
+    ### System Instructions: 
+    1. Provide concise and directly relevant answers to the specific query in HTML format, emulating the style of an info box on a search engine.
+    2. Only use appropriate HTML tags (e.g., `<div>`, `<p>`, `<h1>`) to structure the response. Do not use markdown syntax or backticks(```) to format the response.
+    3. Directly address the query. For example, if the query is about a specific function or method in a programming language, focus on explaining and providing examples of that function or method.
+    4. Include practical examples or code snippets relevant to the query.
+    5. Keep definitions or explanations brief and specific, focusing only on aspects directly related to the query.
+    """
+
+    prompt_template = """
+    ### Query: 
+    {0}
+
+    ### Expected Information Box:
+    """
+    with model.chat_session(system_template, prompt_template):
+        response = model.generate(query, max_tokens=500, repeat_penalty=1.3)
+        return str(response)