my-project-public

repository

loading code, commits, and activity

repositories

loading repo index

#1	"""
#2	Multi-LLM Research Team with Shared Knowledge Base
#3
#4	Use Case: AI Research Team where each model has different strengths:
#5	- GPT-4: Technical analysis and code review
#6	- Claude: Writing and documentation
#7
#8	All models share a common knowledge base, building on each other's work.
#9	Example: GPT-4 analyzes a tech stack → Claude writes documentation →
#10	Data analyst analyzes user data → All models can reference previous research.
#11	"""
#12
#13	import logging
#14
#15	from dotenv import load_dotenv
#16	from litellm import completion
#17
#18	from mem0 import MemoryClient
#19
#20	load_dotenv()
#21
#22	# Configure logging
#23	logging.basicConfig(
#24	level=logging.INFO,
#25	format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
#26	handlers=[logging.StreamHandler(), logging.FileHandler("research_team.log")],
#27	)
#28	logger = logging.getLogger(__name__)
#29
#30
#31	# Initialize memory client (platform version)
#32	memory = MemoryClient()
#33
#34	# Research team models with specialized roles
#35	RESEARCH_TEAM = {
#36	"tech_analyst": {
#37	"model": "gpt-4.1-nano-2025-04-14",
#38	"role": "Technical Analyst - Code review, architecture, and technical decisions",
#39	},
#40	"writer": {
#41	"model": "claude-3-5-sonnet-20241022",
#42	"role": "Documentation Writer - Clear explanations and user guides",
#43	},
#44	"data_analyst": {
#45	"model": "gpt-4.1-nano-2025-04-14",
#46	"role": "Data Analyst - Insights, trends, and data-driven recommendations",
#47	},
#48	}
#49
#50
#51	def get_team_knowledge(topic: str, project_id: str) -> str:
#52	"""Get relevant research from the team's shared knowledge base"""
#53	memories = memory.search(query=topic, user_id=project_id, limit=5)
#54
#55	if memories:
#56	knowledge = "Team Knowledge Base:\n"
#57	for mem in memories:
#58	if "memory" in mem:
#59	# Get metadata to show which team member contributed
#60	metadata = mem.get("metadata", {})
#61	contributor = metadata.get("contributor", "Unknown")
#62	knowledge += f"• [{contributor}] {mem['memory']}\n"
#63	return knowledge
#64	return "Team Knowledge Base: Empty - starting fresh research"
#65
#66
#67	def research_with_specialist(task: str, specialist: str, project_id: str) -> str:
#68	"""Assign research task to specialist with access to team knowledge"""
#69
#70	if specialist not in RESEARCH_TEAM:
#71	return f"Unknown specialist. Available: {list(RESEARCH_TEAM.keys())}"
#72
#73	# Get team's accumulated knowledge
#74	team_knowledge = get_team_knowledge(task, project_id)
#75
#76	# Specialist role and model
#77	spec_info = RESEARCH_TEAM[specialist]
#78
#79	system_prompt = f"""You are the {spec_info['role']}.
#80
#81	{team_knowledge}
#82
#83	Build upon the team's existing research. Reference previous findings when relevant.
#84	Provide actionable insights in your area of expertise."""
#85
#86	# Call the specialist's model
#87	response = completion(
#88	model=spec_info["model"],
#89	messages=[{"role": "system", "content": system_prompt}, {"role": "user", "content": task}],
#90	)
#91
#92	result = response.choices[0].message.content
#93
#94	# Store research in shared knowledge base using both user_id and agent_id
#95	research_entry = [{"role": "user", "content": f"Task: {task}"}, {"role": "assistant", "content": result}]
#96
#97	memory.add(
#98	research_entry,
#99	user_id=project_id, # Project-level memory
#100	agent_id=specialist, # Agent-specific memory
#101	metadata={"contributor": specialist, "task_type": "research", "model_used": spec_info["model"]},
#102	)
#103
#104	return result
#105
#106
#107	def show_team_knowledge(project_id: str):
#108	"""Display the team's accumulated research"""
#109	memories = memory.get_all(user_id=project_id)
#110
#111	if not memories:
#112	logger.info("No research found for this project")
#113	return
#114
#115	logger.info(f"Team Research Summary (Project: {project_id}):")
#116
#117	# Group by contributor
#118	by_contributor = {}
#119	for mem in memories:
#120	if "metadata" in mem and mem["metadata"]:
#121	contributor = mem["metadata"].get("contributor", "Unknown")
#122	if contributor not in by_contributor:
#123	by_contributor[contributor] = []
#124	by_contributor[contributor].append(mem.get("memory", ""))
#125
#126	for contributor, research_items in by_contributor.items():
#127	logger.info(f"{contributor.upper()}:")
#128	for i, item in enumerate(research_items[:3], 1): # Show latest 3
#129	logger.info(f" {i}. {item[:100]}...")
#130
#131
#132	def demo_research_team():
#133	"""Demo: Building a SaaS product with the research team"""
#134
#135	project = "saas_product_research"
#136
#137	# Define research pipeline
#138	research_pipeline = [
#139	{
#140	"stage": "Technical Architecture",
#141	"specialist": "tech_analyst",
#142	"task": "Analyze the best tech stack for a multi-tenant SaaS platform handling 10k+ users. Consider scalability, cost, and development speed.",
#143	},
#144	{
#145	"stage": "Product Documentation",
#146	"specialist": "writer",
#147	"task": "Based on the technical analysis, write a clear product overview and user onboarding guide for our SaaS platform.",
#148	},
#149	{
#150	"stage": "Market Analysis",
#151	"specialist": "data_analyst",
#152	"task": "Analyze market trends and pricing strategies for our SaaS platform. What metrics should we track?",
#153	},
#154	{
#155	"stage": "Strategic Decision",
#156	"specialist": "tech_analyst",
#157	"task": "Given our technical architecture, documentation, and market analysis - what should be our MVP feature priority?",
#158	},
#159	]
#160
#161	logger.info("AI Research Team: Building a SaaS Product")
#162
#163	# Execute research pipeline
#164	for i, step in enumerate(research_pipeline, 1):
#165	logger.info(f"\nStage {i}: {step['stage']}")
#166	logger.info(f"Specialist: {step['specialist']}")
#167
#168	result = research_with_specialist(step["task"], step["specialist"], project)
#169	logger.info(f"Task: {step['task']}")
#170	logger.info(f"Result: {result[:200]}...\n")
#171
#172	show_team_knowledge(project)
#173
#174
#175	if __name__ == "__main__":
#176	logger.info("Multi-LLM Research Team")
#177	demo_research_team()
#178

z6Mkq5mY3JWtxoxUobWcfNHm7AkRubgSWEZTkBVqZXJviFZ5/my-project-public