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
"""Simple test for opencode-ai SDK"""
import json
import os
import subprocess
import sys
import time
import traceback
from opencode_ai import Opencode
def execute_cmd(cmd):
try:
result = subprocess.run(
cmd,
shell=True,
text=True,
encoding="utf-8",
stdout=subprocess.PIPE,
stderr=subprocess.PIPE, # 捕获错误输出
check=True, # 等价于check_output的行为,命令失败会抛异常
)
stdout = result.stdout.strip() # 去除空白符(换行/空格)
return stdout
except subprocess.CalledProcessError as e:
# 捕获命令执行失败的异常(返回码非0)
print(f"命令执行失败:{e}")
return None
except Exception as e:
# 捕获其他异常(如命令不存在、超时等)
print(f"执行异常:{str(e)}")
return None
def start_opencode():
"""子进程函数:启动opencode serve并完全脱离子进程控制"""
pid = None
try:
if sys.platform == "win32":
# Windows:使用CREATE_NEW_PROCESS_GROUP创建独立进程组,detach脱离父进程
creationflags = subprocess.CREATE_NEW_PROCESS_GROUP | subprocess.DETACHED_PROCESS
proc = subprocess.Popen(
["opencode", "serve"],
shell=False,
creationflags=creationflags,
stdout=subprocess.DEVNULL, # 重定向输出避免控制台关联
stderr=subprocess.DEVNULL,
stdin=subprocess.DEVNULL,
)
pid = proc.pid
else:
# Linux/macOS:使用os.setsid创建新会话,完全脱离控制终端
# 先fork一次,再启动进程,确保脱离所有父进程关联
cmd = ["opencode", "serve"]
# 创建新会话 + 重定向所有输出
with open("opencode.log", "a", encoding="utf-8") as log_file:
proc = subprocess.Popen(
cmd,
preexec_fn=os.setsid, # 关键:创建新的会话ID
stdout=log_file,
stderr=subprocess.STDOUT,
stdin=subprocess.DEVNULL,
)
pid = proc.pid
print(f"opencode serve已启动,PID: {pid}")
# 短暂等待确保进程启动成功
time.sleep(2)
return pid
except Exception as e:
print(f"启动opencode失败: {e}")
traceback.print_exc()
return None
def check_serve_status():
"""测试opencode连接,失败则启动服务并重试"""
try:
client = Opencode(base_url="http://127.0.0.1:4096")
client.app.modes()
except Exception as e:
print(f"连接opencode失败,错误: {e}")
# 启动服务
pid = start_opencode()
if pid:
# 启动后重试连接
time.sleep(3)
def read_new_messages(client, session_id, last_ts):
"""
读取上一次之后的消息, 通过client.session.messages实现,注意
"""
messages = client.session.messages(id=session_id, extra_query={"limit": 10})
next_ts = last_ts
new_messages = []
has_finished = False
for message in messages:
created_time = 0
if hasattr(message, "info") and message.info:
if hasattr(message.info, "time") and message.info.time:
if hasattr(message.info.time, "created"):
created_time = message.info.time.created
if created_time > last_ts:
new_messages.append(message)
if created_time > next_ts:
next_ts = created_time
if messages:
last_message = messages[-1]
if last_message.parts:
for part in last_message.parts:
if hasattr(part, "type") and part.type == "step-finish":
has_finished = True
status = "finished" if has_finished else "running"
return status, new_messages, next_ts
file_path = "status.json"
def read_status():
# 检查文件是否存在
if not os.path.exists(file_path):
return {}
# 读取并解析JSON文件
try:
with open(file_path, "r", encoding="utf-8") as f:
data = json.load(f)
return data
except Exception as e:
# 捕获其他未知异常
print(f"读取 {file_path} 时发生错误:{e}")
return {}
def write_status(status):
try:
with open(file_path, "w", encoding="utf-8") as f:
f.write(json.dumps(status))
except Exception as e:
# 捕获其他未知异常
print(f"写入 {file_path} 时发生错误:{e}")
def list_project(client):
import httpx
http_client = httpx.Client(base_url="http://127.0.0.1:4096")
response = http_client.get("/project")
projects = response.json()
project_list = []
for p in projects:
project_list.append({"id": p.get("id"), "path": p.get("worktree")})
return project_list