Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
Python agent-contextagent-memoryagent-pluginsagentic-ragcontext-databasedsh-pluginknowledge-basememorymemory-managementself-evolvingvector-database
#!/usr/bin/env python3
"""
Bob — 新入职成员的使用流程
操作:浏览团队资源 → 回顾团队记忆 → 添加自己的资源 → 对话 → 沉淀记忆 → 带上下文搜索
获取 API Key:
API Key 由租户管理员分配,流程如下:
1. 管理员(如 Alice)用自己的 Key 注册 Bob:
curl -X POST http://localhost:1933/api/v1/admin/accounts/demo-team/users \
-H "X-API-Key: <alice_key>" -H "Content-Type: application/json" \
-d '{"user_id": "bob", "role": "user"}'
返回中的 user_key 就是 Bob 的 API Key
2. 或者运行 setup_users.py 自动完成,Key 写入 user_keys.json
运行(建议在 alice.py 之后执行,这样可以看到 Alice 沉淀的团队记忆):
uv run examples/cloud/bob.py
uv run examples/cloud/bob.py --url http://localhost:1933 --api-key <bob_key>
"""
import argparse
import json
import sys
import time
import openviking as ov
from openviking_cli.utils.async_utils import run_async
def load_key_from_file(user="bob"):
try:
with open("examples/cloud/user_keys.json") as f:
keys = json.load(f)
return keys["url"], keys[f"{user}_key"]
except (FileNotFoundError, KeyError):
return None, None
def main():
parser = argparse.ArgumentParser(description="Bob 的使用流程")
parser.add_argument("--url", default=None, help="Server URL")
parser.add_argument("--api-key", default=None, help="Bob 的 API Key")
args = parser.parse_args()
url, api_key = args.url, args.api_key
if not api_key:
url_from_file, key_from_file = load_key_from_file("bob")
url = url or url_from_file or "http://localhost:1933"
api_key = key_from_file
if not url:
url = "http://localhost:1933"
if not api_key:
print("请通过 --api-key 指定 API Key,或先运行 setup_users.py")
sys.exit(1)
print(f"Server: {url}")
print("User: bob")
print("Key: [hidden]")
client = ov.SyncHTTPClient(url=url, api_key=api_key)
client.initialize()
try:
# ── 1. 浏览团队已有资源 ──
print("\n== 1. 浏览团队资源 ==")
entries = client.ls("viking://")
if not entries:
print(" (空,Alice 还没添加资源)")
for entry in entries:
if isinstance(entry, dict):
kind = "dir " if entry.get("isDir") else "file"
print(f" [{kind}] {entry.get('name', '?')}")
# ── 2. 回顾团队记忆(Alice 沉淀的技术决策) ──
print("\n== 2. 回顾团队记忆: '项目技术选型' ==")
results = client.find(query="项目用了什么技术栈和架构选型", limit=5)
memories = results.get("memories", [])
if memories:
print(" 团队记忆:")
for index, memory in enumerate(memories, 1):
description = (
memory.get("abstract")
or memory.get("overview")
or memory.get("uri", "")
)
print(f" {index}. [{memory.get('score', 0.0):.3f}] {description[:150]}")
else:
print(" 未找到团队记忆(Alice 可能还没执行 commit)")
resources = results.get("resources", [])
if resources:
print(" 相关资源:")
for index, resource in enumerate(resources, 1):
print(
f" {index}. [{resource.get('score', 0.0):.3f}] "
f"{resource['uri']}"
)
# ── 3. 搜索具体决策 ──
print("\n== 3. 搜索: '存储方案 TOS 配置' ==")
results = client.find(query="文件存储方案 TOS bucket 配置", limit=3)
memories = results.get("memories", [])
if memories:
for index, memory in enumerate(memories, 1):
description = (
memory.get("abstract")
or memory.get("overview")
or memory.get("uri", "")
)
print(f" {index}. [{memory.get('score', 0.0):.3f}] {description[:150]}")
else:
print(" 未找到相关记忆")
# ── 4. 添加自己的资源 ──
print("\n== 4. 添加资源: CONTRIBUTING.md ==")
result = client.add_resource(
path="https://raw.githubusercontent.com/volcengine/OpenViking/refs/heads/main/CONTRIBUTING.md",
options={"reason": "贡献指南学习笔记"},
)
bob_uri = result.get("root_uri", "")
print(f" URI: {bob_uri}")
print(" 等待处理...")
client.wait_processed()
print(" 完成")
# ── 5. 创建会话,模拟入职学习 ──
print("\n== 5. 对话: 入职学习 ==")
session_info = client.create_session()
session = client.session(session_id=session_info["session_id"])
print(f" Session: {session.session_id}")
messages = [
("user", "我刚入职,需要了解 OpenViking 的贡献流程"),
(
"assistant",
"欢迎!贡献流程主要是:1) Fork 仓库 2) 创建 feature branch "
"3) 提交 PR 并通过 CI 4) Code Review 后合并。"
"代码规范见 CONTRIBUTING.md。",
),
("user", "本地开发环境怎么搭建?"),
(
"assistant",
"本地开发步骤:1) 安装 Python 3.10+ 和 uv "
"2) git clone 后执行 uv sync 安装依赖 "
"3) 复制 examples/ov.conf.example 为 ~/.openviking/ov.conf 填入 API Key "
"4) 运行 openviking-server 启动开发服务。构建 abi3 C++ 扩展需要 cmake。",
),
("user", "测试怎么跑?"),
(
"assistant",
"运行测试:1) uv run pytest 跑全量测试 "
"2) uv run pytest tests/unit -x 只跑单元测试 "
"3) CI 会自动跑 lint + test,PR 合并前必须全绿。",
),
]
for role, content in messages:
run_async(session.add_message(role=role, content=content))
print(f" 添加了 {len(messages)} 条消息")
# ── 6. 沉淀记忆 ──
print("\n== 6. 沉淀记忆: commit session ==")
print(" 正在提取(开发流程、环境配置等)...")
client.commit_session(session.session_id)
print(" commit 完成")
time.sleep(2)
client.wait_processed()
print(" 记忆向量化完成")
# ── 7. 回顾自己的记忆 ──
print("\n== 7. 回顾记忆: '本地开发环境搭建' ==")
results = client.find(query="本地开发环境搭建步骤", limit=3)
memories = results.get("memories", [])
if memories:
print(" 记忆:")
for index, memory in enumerate(memories, 1):
description = (
memory.get("abstract")
or memory.get("overview")
or memory.get("uri", "")
)
print(f" {index}. [{memory.get('score', 0.0):.3f}] {description[:150]}")
resources = results.get("resources", [])
if resources:
print(" 资源:")
for index, resource in enumerate(resources, 1):
print(
f" {index}. [{resource.get('score', 0.0):.3f}] "
f"{resource['uri']}"
)
# ── 8. 带会话上下文的搜索 ──
print("\n== 8. 带上下文搜索: '还有什么注意事项' ==")
results = client.search(
query="还有什么需要注意的事项",
session_id=session.session_id,
limit=3,
)
resources = results.get("resources", [])
if resources:
for index, resource in enumerate(resources, 1):
print(
f" {index}. [{resource.get('score', 0.0):.3f}] "
f"{resource['uri']}"
)
memories = results.get("memories", [])
if memories:
for index, memory in enumerate(memories, 1):
description = (
memory.get("abstract")
or memory.get("overview")
or memory.get("uri", "")
)
print(f" {index}. [{memory.get('score', 0.0):.3f}] {description[:100]}")
print("\nBob 流程完成")
finally:
client.close()
if __name__ == "__main__":
main()