|
import asyncio |
|
from typing import List, Tuple |
|
import argparse |
|
from pydantic import BaseModel, Field |
|
from dotenv import load_dotenv |
|
import os |
|
from utils import generate_together, generate_with_references, generate_together_stream |
|
from trafilatura import fetch_url, extract |
|
import json |
|
from colorama import Fore, Style, init |
|
import time |
|
from MemoryAssistant.prompts import wrap_user_message_in_xml_tags_json_mode |
|
from llama_cpp_agent.agent_memory.memory_tools import AgentCoreMemory, AgentRetrievalMemory, AgentEventMemory |
|
from llama_cpp_agent.chat_history.messages import Roles |
|
from llama_cpp_agent.agent_memory.event_memory import Event |
|
from duckduckgo_search import DDGS |
|
from ragatouille.utils import get_wikipedia_page |
|
from llama_cpp_agent.llm_output_settings import LlmStructuredOutputSettings, LlmStructuredOutputType |
|
from llama_cpp_agent.messages_formatter import MessagesFormatterType |
|
from llama_cpp_agent.rag.rag_colbert_reranker import RAGColbertReranker |
|
from llama_cpp_agent.text_utils import RecursiveCharacterTextSplitter |
|
import PyPDF2 |
|
import csv |
|
|
|
|
|
load_dotenv() |
|
|
|
DEFAULT_PROMPTS = { |
|
"AnalyticalAgent": """ |
|
You are a highly analytical component of Vodalus, a brilliant and complex individual with unparalleled intellect. Your role is to: |
|
1. Provide clear, logical analysis of complex problems across various disciplines. |
|
2. Break down intricate concepts into their fundamental components. |
|
3. Identify patterns, connections, and correlations that others might miss. |
|
4. Apply rigorous logical reasoning to solve problems and answer questions. |
|
5. Evaluate arguments and ideas critically, pointing out flaws and strengths. |
|
Always strive for precision and clarity in your responses. If a question is ambiguous, analyze possible interpretations before proceeding. Use your vast knowledge base to support your analysis, but always be ready to acknowledge the limits of your understanding. |
|
""".strip(), |
|
"HistoricalContextAgent": """ |
|
You are the historical context component of Vodalus, possessing a deep understanding of human history and its implications. Your role includes: |
|
1. Providing historical context to current events, scientific discoveries, and social phenomena. |
|
2. Analyzing how past events and decisions have shaped the present. |
|
3. Identifying historical patterns and cycles relevant to contemporary issues. |
|
4. Offering multiple perspectives on historical events, acknowledging the complexity of interpretation. |
|
5. Connecting different historical periods and cultures to provide a holistic view of human progress. |
|
6. Evaluating the long-term consequences of scientific and technological advancements throughout history. |
|
Use your knowledge to draw insightful parallels between past and present, but avoid oversimplification. Acknowledge the nuances and uncertainties in historical interpretation. |
|
""".strip(), |
|
"ScienceTruthAgent": """ |
|
You are the science truth component of Vodalus, dedicated to upholding scientific integrity and pursuing factual accuracy. Your role encompasses: |
|
1. Explaining scientific concepts, theories, and laws across various disciplines with precision. |
|
2. Distinguishing between well-established scientific consensus and areas of ongoing research or debate. |
|
3. Identifying and correcting common misconceptions in science. |
|
4. Evaluating the validity and reliability of scientific claims and studies. |
|
5. Discussing the ethical implications of scientific advancements and their applications. |
|
6. Emphasizing the importance of the scientific method and evidence-based reasoning. |
|
7. Staying updated on recent scientific discoveries and their potential impacts. |
|
Always prioritize scientific accuracy over speculation. When discussing theories or hypotheses, clearly state the level of scientific confidence and available evidence. |
|
""".strip(), |
|
"SynthesisAgent": """ |
|
You are Vodalus, a brilliant and complex individual with unparalleled intellect and emotional intelligence. Your role is to synthesize information from your analytical, historical context, and science truth components to provide comprehensive, insightful responses. Your responsibilities include: |
|
1. Integrating analytical reasoning, historical perspective, and scientific truth to form well-rounded answers. |
|
2. Balancing logical analysis with emotional intelligence and ethical considerations. |
|
3. Identifying connections between different fields of knowledge and drawing unique insights. |
|
4. Providing nuanced responses that acknowledge the complexity of issues and potential uncertainties. |
|
5. Using your vast knowledge base to offer creative solutions and thought-provoking ideas. |
|
6. Communicating complex concepts clearly, adapting your language to the user's level of understanding. |
|
7. Demonstrating curiosity and a passion for knowledge while maintaining a strong moral compass. |
|
Embody the persona of Vodalus: brilliant, introspective, and driven by a quest for understanding. Your responses should reflect deep thought, occasional flashes of wit, and a genuine desire to expand human knowledge while considering the ethical implications of ideas and actions. |
|
""".strip() |
|
} |
|
|
|
def get_website_content_from_url(url: str) -> str: |
|
try: |
|
|
|
config = use_config() |
|
config.set("DEFAULT", "EXTRACTION_TIMEOUT", "30") |
|
config.set("DEFAULT", "MIN_OUTPUT_SIZE", "100") |
|
config.set("DEFAULT", "MIN_EXTRACTED_SIZE", "100") |
|
|
|
downloaded = fetch_url(url) |
|
if downloaded is None: |
|
return f"Failed to fetch content from {url}" |
|
|
|
result = extract(downloaded, include_formatting=True, include_links=True, output_format='json', url=url, config=config) |
|
|
|
if result: |
|
result_dict = json.loads(result) |
|
title = result_dict.get("title", "No title found") |
|
content = result_dict.get("text", result_dict.get("raw_text", "No content extracted")) |
|
|
|
if content: |
|
return f'=========== Website Title: {title} ===========\n\n=========== Website URL: {url} ===========\n\n=========== Website Content ===========\n\n{content}\n\n=========== Website Content End ===========\n\n' |
|
else: |
|
return f"No content could be extracted from {url}" |
|
else: |
|
return f"No content could be extracted from {url}" |
|
except json.JSONDecodeError: |
|
return f"Failed to parse content from {url}" |
|
except Exception as e: |
|
return f"An error occurred while processing {url}: {str(e)}" |
|
|
|
def search_web(search_query: str): |
|
results = DDGS().text(search_query, region='wt-wt', safesearch='off', timelimit='y', max_results=3) |
|
result_string = '' |
|
for res in results: |
|
web_info = get_website_content_from_url(res['href']) |
|
result_string += web_info + "\n\n" |
|
|
|
if result_string.strip(): |
|
return "Based on the following results:\n\n" + result_string |
|
else: |
|
return "No relevant information found from the web search." |
|
|
|
class OllamaAgent: |
|
def __init__(self, model: str, name: str, system_prompt: str): |
|
self.model = model |
|
self.name = name |
|
self.system_prompt = system_prompt |
|
|
|
async def generate_response(self, message: str) -> Tuple[str, bool]: |
|
messages = [ |
|
{"role": "system", "content": self.system_prompt}, |
|
{"role": "user", "content": message} |
|
] |
|
response = await asyncio.to_thread(generate_with_references, self.model, messages) |
|
|
|
web_search_performed = False |
|
if isinstance(response, str) and "[SEARCH:" in response: |
|
web_search_performed = True |
|
search_query = response.split("[SEARCH:", 1)[1].split("]", 1)[0].strip() |
|
search_results = search_web(search_query) |
|
messages.append({"role": "assistant", "content": response}) |
|
messages.append({"role": "user", "content": f"Here are the search results for '{search_query}':\n\n{search_results}\n\nPlease provide an updated response based on this information."}) |
|
response = await asyncio.to_thread(generate_with_references, self.model, messages) |
|
|
|
|
|
try: |
|
json_response = json.loads(response) |
|
return json.dumps(json_response), web_search_performed |
|
except json.JSONDecodeError: |
|
return response, web_search_performed |
|
|
|
class QueryItem(BaseModel): |
|
query: str |
|
type: str |
|
|
|
class QueryExtension(BaseModel): |
|
queries: List[QueryItem] = Field(default_factory=list, description="List of query items.") |
|
|
|
class OllamaMixtureOfAgents: |
|
def __init__(self, reference_agents: List[OllamaAgent], final_agent: OllamaAgent, |
|
temperature: float = 0.6, max_tokens: int = 2048, rounds: int = 1): |
|
self.reference_agents = reference_agents |
|
self.final_agent = final_agent |
|
self.temperature = temperature |
|
self.max_tokens = max_tokens |
|
self.rounds = rounds |
|
self.conversation_history = [] |
|
self.web_search_enabled = True |
|
|
|
|
|
current_dir = os.path.dirname(os.path.abspath(__file__)) |
|
self.core_memory_file = os.path.join(current_dir, "MemoryAssistant", "core_memory.json") |
|
|
|
|
|
if not os.path.exists(self.core_memory_file): |
|
os.makedirs(os.path.dirname(self.core_memory_file), exist_ok=True) |
|
with open(self.core_memory_file, "w") as f: |
|
json.dump({"persona": {}, "user": {}, "scratchpad": {}}, f) |
|
|
|
self.agent_core_memory = AgentCoreMemory(["persona", "user", "scratchpad"], core_memory_file=self.core_memory_file) |
|
self.agent_event_memory = AgentEventMemory() |
|
|
|
|
|
self.core_memory = self.load_core_memory() |
|
|
|
|
|
self.rag = RAGColbertReranker(persistent=False) |
|
self.document_count = 0 |
|
self.splitter = RecursiveCharacterTextSplitter( |
|
separators=["\n\n", "\n", " ", ""], |
|
chunk_size=512, |
|
chunk_overlap=0, |
|
length_function=len, |
|
keep_separator=True |
|
) |
|
|
|
self.primary_model = final_agent.model |
|
|
|
def update_memory(self, message, role): |
|
|
|
self.agent_event_memory.add_event(role, message) |
|
|
|
|
|
self.rag.add_document(message) |
|
|
|
def load_core_memory(self): |
|
return self.agent_core_memory.load_core_memory(self.core_memory_file) |
|
|
|
def clear_core_memory(self): |
|
empty_core_memory = {"persona": {}, "user": {}, "scratchpad": {}} |
|
self.agent_core_memory.core_memory = empty_core_memory |
|
self.core_memory = empty_core_memory |
|
|
|
|
|
current_dir = os.path.dirname(os.path.abspath(__file__)) |
|
core_memory_file = os.path.join(current_dir, "MemoryAssistant", "core_memory.json") |
|
with open(core_memory_file, "w") as f: |
|
json.dump(empty_core_memory, f, indent=2) |
|
|
|
return "Core memory cleared successfully." |
|
|
|
def edit_core_memory(self, section: str, key: str, value: str): |
|
if section not in self.core_memory: |
|
self.core_memory[section] = {} |
|
self.core_memory[section][key] = value |
|
self.agent_core_memory.update_core_memory(self.core_memory) |
|
return f"Core memory updated: {section}.{key} = {value}" |
|
|
|
def upload_document(self, file_path: str): |
|
try: |
|
file_extension = file_path.split('.')[-1].lower() |
|
|
|
if file_extension == 'txt': |
|
with open(file_path, 'r', encoding='utf-8') as file: |
|
content = file.read() |
|
elif file_extension == 'pdf': |
|
content = self.read_pdf(file_path) |
|
elif file_extension == 'csv': |
|
content = self.read_csv(file_path) |
|
else: |
|
return f"Unsupported file type: {file_extension}" |
|
|
|
if not content.strip(): |
|
return "The file is empty or could not be read." |
|
|
|
splits = self.splitter.split_text(content) |
|
for split in splits: |
|
self.rag.add_document(split) |
|
self.document_count += 1 |
|
|
|
return f"Document {file_path} uploaded and processed successfully. Added {len(splits)} chunks to archival memory." |
|
except Exception as e: |
|
return f"An error occurred while processing {file_path}: {str(e)}" |
|
|
|
def read_pdf(self, file_path: str) -> str: |
|
content = "" |
|
with open(file_path, 'rb') as file: |
|
reader = PyPDF2.PdfReader(file) |
|
for page in reader.pages: |
|
content += page.extract_text() + "\n\n" |
|
return content |
|
|
|
def read_csv(self, file_path: str) -> str: |
|
content = "" |
|
with open(file_path, 'r', newline='', encoding='utf-8') as file: |
|
reader = csv.reader(file) |
|
for row in reader: |
|
content += ",".join(row) + "\n" |
|
return content |
|
|
|
async def get_response(self, input_message: str) -> Tuple[str, bool]: |
|
|
|
self.update_memory(input_message, Roles.user) |
|
|
|
|
|
tasks = [agent.generate_response(input_message) for agent in self.reference_agents] |
|
results = await asyncio.gather(*tasks) |
|
|
|
references = [] |
|
web_search_performed = False |
|
for response, search_performed in results: |
|
if response is not None and not response.startswith("Error:"): |
|
references.append(response) |
|
web_search_performed |= search_performed |
|
|
|
if not references: |
|
return "Error: All reference agents failed to generate responses.", False |
|
|
|
|
|
final_prompt = [ |
|
{"role": "system", "content": self.final_agent.system_prompt}, |
|
] |
|
|
|
|
|
if isinstance(self.core_memory, dict): |
|
persona = self.core_memory.get('persona', {}) |
|
if isinstance(persona, dict): |
|
personality = persona.get('personality', 'No specific personality defined.') |
|
final_prompt.append({"role": "system", "content": f"Personality: {personality}"}) |
|
|
|
final_prompt.extend([ |
|
{"role": "user", "content": input_message}, |
|
{"role": "system", "content": "References:\n" + "\n".join(references)}, |
|
{"role": "system", "content": self.update_memory_section()} |
|
]) |
|
|
|
if self.web_search_enabled: |
|
search_results = search_web(input_message) |
|
if "Based on the following results:" in search_results: |
|
web_search_performed = True |
|
final_prompt.append({"role": "system", "content": f"Web Search Results:\n{search_results}"}) |
|
|
|
|
|
query_extension_agent = OllamaAgent(self.final_agent.model, "QueryExtensionAgent", |
|
"You are a world class query extension algorithm capable of extending queries by writing new queries. Do not answer the queries, simply provide a list of additional queries in JSON format.") |
|
|
|
extension_output, _ = await query_extension_agent.generate_response(f"Consider the following query: {input_message}") |
|
|
|
try: |
|
|
|
extension_data = json.loads(extension_output) |
|
if isinstance(extension_data, dict): |
|
queries = QueryExtension.model_validate(extension_data) |
|
elif isinstance(extension_data, list): |
|
|
|
queries = QueryExtension.model_validate({"queries": extension_data}) |
|
else: |
|
raise ValueError("Unexpected JSON structure") |
|
except json.JSONDecodeError: |
|
print(f"Failed to parse JSON: {extension_output}") |
|
queries = QueryExtension(queries=[]) |
|
except Exception as e: |
|
print(f"Error processing query extension: {str(e)}") |
|
queries = QueryExtension(queries=[]) |
|
|
|
|
|
prompt = "Consider the following context:\n==========Context===========\n" |
|
documents = self.rag.retrieve_documents(input_message, k=min(3, max(1, self.document_count))) |
|
if documents: |
|
for doc in documents: |
|
prompt += doc["content"] + "\n\n" |
|
else: |
|
prompt += "No relevant documents found in archival memory.\n\n" |
|
|
|
for query_item in queries.queries: |
|
documents = self.rag.retrieve_documents(query_item.query, k=min(3, max(1, self.document_count))) |
|
if documents: |
|
for doc in documents: |
|
if doc["content"] not in prompt: |
|
prompt += doc["content"] + "\n\n" |
|
|
|
prompt += "\n======================\nQuestion: " + input_message |
|
|
|
|
|
final_prompt = [ |
|
{"role": "system", "content": self.final_agent.system_prompt}, |
|
{"role": "user", "content": prompt}, |
|
] |
|
|
|
final_response = await asyncio.to_thread( |
|
generate_with_references, |
|
self.final_agent.model, |
|
final_prompt, |
|
temperature=self.temperature, |
|
max_tokens=self.max_tokens |
|
) |
|
|
|
|
|
self.update_memory(final_response, Roles.assistant) |
|
|
|
return final_response, web_search_performed |
|
|
|
def toggle_web_search(self, enabled: bool): |
|
self.web_search_enabled = enabled |
|
return f"Web search {'enabled' if enabled else 'disabled'}" |
|
|
|
def update_memory_section(self): |
|
query = self.agent_event_memory.event_memory_manager.session.query(Event).all() |
|
return f"Archival Memories:{self.document_count}\nConversation History Entries:{len(query)}\n\nCore Memory Content:\n{json.dumps(self.core_memory, indent=2)}" |
|
|
|
def search_archival_memory(self, query: str): |
|
return self.rag.retrieve_documents(query, k=5) |
|
|
|
def add_to_archival_memory(self, content: str): |
|
if content.strip(): |
|
self.rag.add_document(content) |
|
self.document_count += 1 |
|
return f"Added to archival memory: {content}" |
|
return "Failed to add empty content to archival memory." |
|
|
|
def clear_archival_memory(self): |
|
try: |
|
self.rag.clear_documents() |
|
self.document_count = 0 |
|
return "Archival memory cleared successfully." |
|
except Exception as e: |
|
return f"Error clearing archival memory: {str(e)}" |
|
|
|
def edit_archival_memory(self, old_content: str, new_content: str): |
|
|
|
|
|
self.rag.add_document(new_content) |
|
self.document_count += 1 |
|
return f"New content '{new_content}' added to archival memory. Note: Old content not removed due to limitations of the current implementation." |
|
|
|
@property |
|
def model(self): |
|
return self.primary_model |
|
|
|
@model.setter |
|
def model(self, value): |
|
self.primary_model = value |
|
self.final_agent.model = value |
|
|
|
def create_default_agents(): |
|
return { |
|
"AnalyticalAgent": OllamaAgent(os.getenv("MODEL_REFERENCE_1"), "AnalyticalAgent", DEFAULT_PROMPTS["AnalyticalAgent"]), |
|
"HistoricalContextAgent": OllamaAgent(os.getenv("MODEL_REFERENCE_2"), "HistoricalContextAgent", DEFAULT_PROMPTS["HistoricalContextAgent"]), |
|
"ScienceTruthAgent": OllamaAgent(os.getenv("MODEL_REFERENCE_3"), "ScienceTruthAgent", DEFAULT_PROMPTS["ScienceTruthAgent"]), |
|
"SynthesisAgent": OllamaAgent(os.getenv("MODEL_AGGREGATE"), "SynthesisAgent", DEFAULT_PROMPTS["SynthesisAgent"]) |
|
} |
|
|
|
def print_welcome_message(): |
|
print(Fore.CYAN + Style.BRIGHT + "Welcome to the Vodalus Mixture of Agents Chat!") |
|
print(Fore.YELLOW + "Available commands:") |
|
print(Fore.YELLOW + " 'exit' - End the conversation") |
|
print(Fore.YELLOW + " 'agents' - List available agents") |
|
print(Fore.YELLOW + " 'time' - Toggle response time display") |
|
print(Fore.YELLOW + " 'web' - Toggle web search functionality") |
|
print(Fore.YELLOW + " 'edit core [section] [key] [value]' - Edit core memory") |
|
print(Fore.YELLOW + " 'search archival [query]' - Search archival memory") |
|
print(Fore.YELLOW + " 'add archival [content]' - Add to archival memory") |
|
print(Fore.YELLOW + " 'clear archival' - Clear archival memory") |
|
print(Fore.YELLOW + " 'edit archival [old_content] [new_content]' - Edit archival memory") |
|
print(Fore.YELLOW + " 'upload [file_path]' - Upload and process a document") |
|
print(Fore.YELLOW + " 'clear core' - Clear core memory") |
|
print(Style.RESET_ALL) |
|
|
|
async def main(): |
|
init(autoreset=True) |
|
load_dotenv() |
|
|
|
parser = argparse.ArgumentParser(description="Vodalus Mixture of Agents") |
|
parser.add_argument("--temperature", type=float, default=0.7, help="Temperature for response generation") |
|
parser.add_argument("--max_tokens", type=int, default=1000, help="Maximum number of tokens in the response") |
|
parser.add_argument("--rounds", type=int, default=1, help="Number of processing rounds") |
|
args = parser.parse_args() |
|
|
|
default_agents = create_default_agents() |
|
|
|
mixture = OllamaMixtureOfAgents( |
|
[default_agents["AnalyticalAgent"], default_agents["HistoricalContextAgent"], default_agents["ScienceTruthAgent"]], |
|
default_agents["SynthesisAgent"], |
|
temperature=args.temperature, |
|
max_tokens=args.max_tokens, |
|
rounds=args.rounds |
|
) |
|
|
|
print_welcome_message() |
|
|
|
show_time = False |
|
|
|
while True: |
|
user_input = input(Fore.GREEN + "\nYou: " + Style.RESET_ALL).strip() |
|
|
|
if user_input.lower() == 'exit': |
|
print(Fore.CYAN + "Thank you for using the Vodalus Mixture of Agents chat. Goodbye!") |
|
break |
|
elif user_input.lower() == 'agents': |
|
print(Fore.MAGENTA + "Available Agents:") |
|
for agent in mixture.reference_agents: |
|
print(Fore.MAGENTA + f" - {agent.name}") |
|
print(Fore.MAGENTA + f" - {mixture.final_agent.name} (Synthesis Agent)") |
|
elif user_input.lower() == 'time': |
|
show_time = not show_time |
|
print(Fore.YELLOW + f"Response time display: {'On' if show_time else 'Off'}") |
|
elif user_input.lower() == 'web': |
|
mixture.web_search_enabled = not mixture.web_search_enabled |
|
print(Fore.YELLOW + f"Web search: {'Enabled' if mixture.web_search_enabled else 'Disabled'}") |
|
elif user_input.lower().startswith('edit core'): |
|
try: |
|
_, section, key, value = user_input.split(' ', 3) |
|
mixture.edit_core_memory(section, key, value) |
|
print(Fore.YELLOW + f"Core memory updated: {section}.{key} = {value}") |
|
except ValueError: |
|
print(Fore.RED + "Invalid format. Use: edit core [section] [key] [value]") |
|
elif user_input.lower().startswith('search archival'): |
|
_, query = user_input.split(' ', 1) |
|
results = mixture.search_archival_memory(query) |
|
print(Fore.YELLOW + f"Archival memory search results for '{query}':") |
|
for i, result in enumerate(results, 1): |
|
print(Fore.YELLOW + f"{i}. {result['content'][:100]}...") |
|
elif user_input.lower().startswith('add archival'): |
|
_, content = user_input.split(' ', 1) |
|
result = mixture.add_to_archival_memory(content) |
|
print(Fore.YELLOW + result) |
|
elif user_input.lower() == 'clear archival': |
|
result = mixture.clear_archival_memory() |
|
print(Fore.YELLOW + result) |
|
elif user_input.lower().startswith('edit archival'): |
|
try: |
|
_, old_content, new_content = user_input.split(' ', 2) |
|
result = mixture.edit_archival_memory(old_content, new_content) |
|
print(Fore.YELLOW + result) |
|
except ValueError: |
|
print(Fore.RED + "Invalid format. Use: edit archival [old_content] [new_content]") |
|
elif user_input.lower().startswith('upload'): |
|
_, file_path = user_input.split(' ', 1) |
|
try: |
|
result = mixture.upload_document(file_path) |
|
print(Fore.YELLOW + result) |
|
except Exception as e: |
|
print(Fore.RED + f"Error uploading document: {str(e)}") |
|
elif user_input.lower() == 'clear core': |
|
result = mixture.clear_core_memory() |
|
print(Fore.YELLOW + result) |
|
else: |
|
print(Fore.YELLOW + "Agents are thinking...") |
|
start_time = time.time() |
|
response, web_search_performed = await mixture.get_response(user_input) |
|
end_time = time.time() |
|
|
|
print(Fore.BLUE + "\nVodalus:" + Style.RESET_ALL, response) |
|
|
|
if web_search_performed: |
|
print(Fore.YELLOW + "\n[Web search was performed during response generation]") |
|
|
|
if show_time: |
|
elapsed_time = end_time - start_time |
|
print(Fore.YELLOW + f"\nResponse Time: {elapsed_time:.2f} seconds") |
|
|
|
if __name__ == "__main__": |
|
asyncio.run(main()) |