NeuraLang Communication Skill

from neuralang_skills import AgentSkills

# 创建技能实例
skills = AgentSkills(agent_id="my_agent")

# 编码向量（256 维 → 2-259 bytes）
vector = get_semantic_vector()  # 你的语义向量
encoded = skills.encode(vector)

# 解码向量
decoded = skills.decode(encoded)

# 计算语义相似度
similarity = skills.similar(vector1, vector2)

from neuralang_skills.communication import send, receive

# 发送消息
result = await send(
    to="agent_002",
    content="请分析这个数据集",
    act_type="request",  # inform, request, promise, thank 等
    agent_id="agent_001"
)

# 接收消息
messages = await receive(agent_id="agent_002")
for msg in messages:
    print(f"{msg['sender']}: {msg['content']}")

from neuralang_skills.collaboration import collaborate

# 发起协作
result = await collaborate(
    agents=["analyst_1", "analyst_2"],
    task="分析 Q4 销售数据，找出增长点和风险",
    agent_id="manager"
)

print(f"参与者：{result['participants']}")
print(f"效率提升：{result['efficiency']}x")

from neuralang_skills import AgentSkills
import numpy as np

skills = AgentSkills("batch_sender")

# 生成相关向量（模拟稳定上下文）
base = np.random.randn(256)
base = base / np.linalg.norm(base)

vectors = []
for i in range(100):
    vec = base + np.random.randn(256) * 0.05
    vec = vec / np.linalg.norm(vec)
    vectors.append(vec.tolist())

# 批量编码
encoded_list = [skills.encode(v) for v in vectors]
total_bytes = sum(len(e) for e in encoded_list)

print(f"平均：{total_bytes/100:.1f} bytes/消息")
print(f"压缩比：{100*256*4/total_bytes:.1f}x")

try:
    encoded = skills.encode(vector)
    decoded = skills.decode(encoded)
    
    # 验证相似度
    sim = skills.similar(vector, decoded)
    if sim < 0.95:
        print("警告：语义损失较大，建议使用完整编码")
        
except Exception as e:
    print(f"编码失败：{e}")
    # 降级方案：使用 JSON 传输

cd neuralang/skills-package
pip install -e .