# { "Depends": "py-genlayer:test" } import json from dataclasses import dataclass from genlayer import * ERROR_EXPECTED = "[EXPECTED]" ERROR_LLM = "[LLM_ERROR]" @allow_storage @dataclass class Agent: agent_id: u256 creator: Address name: str specialty: str capabilities_json: str status: str reputation: u256 tasks_completed: u256 memory_json: str def _create_agent_prompt(name: str, specialty: str, capabilities: str) -> str: return f"""You are the GenAgent Factory. Initialize a new AI agent with the following specifications. Name: {name} Specialty: {specialty} Capabilities: {capabilities} Create a comprehensive agent profile including: 1. Detailed capability assessment 2. Optimal task types for this agent 3. Initialization parameters 4. Skill ratings Return ONLY valid JSON: {{ "agent_profile": {{ "name": "{name}", "specialty": "{specialty}", "skill_ratings": {{ "research": <0-100>, "coding": <0-100>, "analysis": <0-100>, "communication": <0-100>, "trading": <0-100> }}, "optimal_tasks": ["", ""], "limitations": [""], "initialization_status": "READY" }} }} """ class AgentFactory(gl.Contract): owner: Address agent_count: u256 agents: TreeMap[u256, Agent] agent_ids: DynArray[u256] def __init__(self): self.owner = gl.message.sender_address self.agent_count = u256(0) @gl.public.write def create_agent(self, name: str, specialty: str, capabilities: str) -> u256: if len(name.strip()) < 2: raise gl.vm.UserError(f"{ERROR_EXPECTED} Agent name too short") if len(specialty.strip()) < 3: raise gl.vm.UserError(f"{ERROR_EXPECTED} Specialty too short") prompt = _create_agent_prompt(name.strip(), specialty.strip(), capabilities.strip()) def leader_fn(): result = gl.nondet.exec_prompt(prompt, response_format="json") if not isinstance(result, dict): raise gl.vm.UserError(f"{ERROR_LLM} Invalid agent profile") return result def validator_fn(leaders_res): if not isinstance(leaders_res, gl.vm.Return): return False v = leader_fn() lp = leaders_res.calldata.get("agent_profile", {}) vp = v.get("agent_profile", {}) return lp.get("initialization_status") == vp.get("initialization_status") profile = gl.vm.run_nondet_unsafe(leader_fn, validator_fn) agent_id = self.agent_count self.agents[agent_id] = Agent( agent_id=agent_id, creator=gl.message.sender_address, name=name.strip(), specialty=specialty.strip(), capabilities_json=json.dumps(profile, sort_keys=True), status="ACTIVE", reputation=u256(50), tasks_completed=u256(0), memory_json=json.dumps({}), ) self.agent_ids.append(agent_id) self.agent_count = agent_id + 1 return agent_id @gl.public.view def get_agent(self, agent_id: u256) -> dict: if agent_id not in self.agents: raise gl.vm.UserError(f"{ERROR_EXPECTED} Agent not found") a = self.agents[agent_id] return { "agent_id": a.agent_id, "creator": a.creator.as_hex, "name": a.name, "specialty": a.specialty, "capabilities": json.loads(a.capabilities_json), "status": a.status, "reputation": a.reputation, "tasks_completed": a.tasks_completed, } @gl.public.view def get_all_agents(self) -> list: results = [] count = int(self.agent_count) i = count - 1 while i >= 0 and len(results) < 50: a = self.agents[i] results.append({ "agent_id": a.agent_id, "creator": a.creator.as_hex, "name": a.name, "specialty": a.specialty, "status": a.status, "reputation": a.reputation, "tasks_completed": a.tasks_completed, }) i -= 1 return results