agent/auto_agent
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

优化agent配置，修复已知问题

Browse Source
This commit is contained in:
minorQAQ committed 2026-08-12 17:48:17 +08:00
1 parent 744cae8f54
commit 9acb38c9e6
55 files changed
+3009 -623
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

src/web/server.py
+295 -19
View File
@@ -64,7 +64,7 @@ def _read_ws_config() -> dict:
return json.load(f)
except (json.JSONDecodeError, OSError):
pass
return {"current": "", "list": []}
return {"current": "", "list": [], "access_mode": "restricted"}
def _write_ws_config(data: dict) -> bool:
@@ -294,7 +294,7 @@ def _serialize_messages(hist: list[dict[str, Any]] | None, session_id: str = "")
except Exception:
return out
# 按 turn_id 为每条 assistant 消息挂载对应轮次的工具调用
# 按 turn_id 为每条 assistant 消息挂载对应轮次的工具调用与轮次级元数据
has_turn_ids = False
for m in out:
if m.get("role") != "assistant":
@@ -303,20 +303,21 @@ def _serialize_messages(hist: list[dict[str, Any]] | None, session_id: str = "")
if not turn_id:
continue
has_turn_ids = True
m_meta = m.setdefault("meta", {})
try:
tool_calls = get_tool_calls_for_turn(session_id, turn_id)
except Exception:
tool_calls = []
if tool_calls:
_inject_sub_tool_calls(tool_calls, session_id, turn_id, get_sub_agent_traces_for_turn)
m_meta = m.setdefault("meta", {})
m_meta["tool_calls"] = tool_calls
try:
turn_meta = get_turn_meta(session_id, turn_id)
except Exception:
turn_meta = None
if turn_meta:
m_meta["turn_meta"] = turn_meta
# 无条件挂载 turn_meta（started_at/ended_at/duration），供前端显示回复开始时间
try:
turn_meta = get_turn_meta(session_id, turn_id)
except Exception:
turn_meta = None
if turn_meta:
m_meta["turn_meta"] = turn_meta
# 兼容旧消息（无 turn_id）：将最新轮工具调用附加到最后一条 assistant
if not has_turn_ids:
@@ -485,6 +486,153 @@ def api_new_session() -> dict[str, Any]:
return {"sessionId": sid, "sessions": ids, "messages": [], "tokens": 0, "pending_actions": []}
def _branch_copy_attachment(abs_src: str, dst_dir: str, new_sid: str) -> str:
"""将附件复制到新会话目录，返回目标绝对路径（必要时避免重名）。
输入:
abs_src: 源附件绝对路径。
dst_dir: 新会话目录。
new_sid: 新会话 ID（用于冲突重命名）。
输出:
目标绝对路径。
"""
base = os.path.basename(abs_src)
dest = os.path.join(dst_dir, base)
if os.path.exists(dest) and os.path.normpath(dest) != os.path.normpath(abs_src):
stem, ext = os.path.splitext(base)
i = 2
while os.path.exists(dest):
dest = os.path.join(dst_dir, f"{stem}_{new_sid}_{i}{ext}")
i += 1
if os.path.normpath(dest) != os.path.normpath(abs_src):
shutil.copy2(abs_src, dest)
return dest
def _branch_rewrite_attachments(dst_dir: str, new_sid: str) -> None:
"""将分支复制后的 turn 文件中的附件引用复制到新会话目录并重写 relpath。
输入:
dst_dir: 新会话目录。
new_sid: 新会话 ID。
"""
from agent.utils.agent_utils import SESSIONS_ROOT, TURN_JSON_PREFIX
for name in os.listdir(dst_dir):
if not (name.startswith(TURN_JSON_PREFIX) and name.endswith(".json")):
continue
tp = os.path.join(dst_dir, name)
try:
with open(tp, "r", encoding="utf-8") as f:
data = json.load(f)
except (json.JSONDecodeError, OSError):
continue
changed = False
for msg in data.get("messages", []):
role = msg.get("role")
if role == "user":
content = (msg.get("user") or {}).get("content") or []
for piece in content:
if not isinstance(piece, dict):
continue
rel = piece.get("relpath")
if not rel:
continue
abs_src = os.path.normpath(os.path.join(SESSIONS_ROOT, rel))
if not os.path.isfile(abs_src):
continue
dest = _branch_copy_attachment(abs_src, dst_dir, new_sid)
new_rel = os.path.relpath(dest, SESSIONS_ROOT).replace("\\", "/")
if new_rel != rel:
piece["relpath"] = new_rel
changed = True
elif role == "assistant":
asst = msg.get("assistant") or {}
audio_rel = asst.get("audio_relpath")
if audio_rel:
abs_src = os.path.normpath(os.path.join(SESSIONS_ROOT, audio_rel.replace("/", os.sep)))
if os.path.isfile(abs_src):
dest = _branch_copy_attachment(abs_src, dst_dir, new_sid)
new_rel = os.path.relpath(dest, SESSIONS_ROOT).replace("\\", "/")
if new_rel != audio_rel:
asst["audio_relpath"] = new_rel
changed = True
if changed:
with open(tp, "w", encoding="utf-8") as f:
json.dump(data, f, ensure_ascii=False, indent=2)
@app.post("/api/sessions/branch")
def api_session_branch(
session_id: str = Form(...),
branch_turn_id: str = Form(""),
) -> dict[str, Any]:
"""将源会话中 branch_turn_id 及之前的所有历史复制到新会话（分支）。
输入:
session_id: 源会话 ID。
branch_turn_id: 分支点回合 ID（该回合及其之前的历史会被复制到新会话）。
输出:
{"sessionId": str, "sessions": list[str], "messages": list[dict], "tokens": int,
"pending_actions": list[dict]}
系统定位:
AI 回复的「分支」按钮：复制一份历史到新会话，用户可在此基础上继续对话。
"""
if not session_id:
raise HTTPException(status_code=400, detail="缺少 session_id")
# mysql 后端不落 turn/tool 文件，暂不支持文件级分支
try:
from agent.history import session_storage
if session_storage.is_mysql():
raise HTTPException(status_code=400, detail="mysql 存储后端暂不支持分支")
except Exception:
pass
from agent.utils.agent_utils import SESSIONS_ROOT, TURN_JSON_PREFIX
src_dir = os.path.join(SESSIONS_ROOT, session_id)
if not os.path.isdir(src_dir):
raise HTTPException(status_code=400, detail="源会话不存在")
# 创建新会话并复制 branch_turn_id 及之前的 turn/tool 文件
new_sid = ui.new_timestamp_session_id()
dst_dir = os.path.join(SESSIONS_ROOT, new_sid)
os.makedirs(dst_dir, exist_ok=True)
branch_turn_full = f"{TURN_JSON_PREFIX}{branch_turn_id}.json"
tool_branch_full = f"tool_{branch_turn_id}.json"
copied_turns = 0
try:
for name in sorted(os.listdir(src_dir)):
src_path = os.path.join(src_dir, name)
if name.startswith(TURN_JSON_PREFIX) and name.endswith(".json") and name <= branch_turn_full:
shutil.copy2(src_path, os.path.join(dst_dir, name))
copied_turns += 1
elif name.startswith("tool_") and name.endswith(".json") and name <= tool_branch_full:
shutil.copy2(src_path, os.path.join(dst_dir, name))
if copied_turns == 0:
raise ValueError("分支点回合不存在")
# 复制附件引用并重写 relpath，使新会话自包含
_branch_rewrite_attachments(dst_dir, new_sid)
except Exception:
shutil.rmtree(dst_dir, ignore_errors=True)
raise HTTPException(status_code=400, detail="无效的分支回合")
ids = _ensure_session_in_list(new_sid, ui.list_session_ids_ordered())
from agent.history.tool_call_recorder import get_session_tokens
return {
"sessionId": new_sid,
"sessions": ids,
"messages": _serialize_messages(ui.load_ui_messages_for_session(new_sid), new_sid),
"tokens": get_session_tokens(new_sid),
"pending_actions": _get_pending_actions(new_sid),
}
@app.get("/api/sessions/{session_id}/messages")
def api_session_messages(session_id: str) -> dict[str, Any]:
"""切换会话：加载指定会话的消息列表与全局会话列表。
@@ -1150,15 +1298,18 @@ def api_save_models(data: dict[str, Any]) -> dict[str, Any]:
def api_get_workspace() -> dict[str, Any]:
"""获取工作空间配置。
输出: {"current": str, "list": list[str], "default": str, "snapshots_base": str}
输出: {"current": str, "list": list[str], "default": str, "snapshots_base": str,
"access_mode": str}
current: 当前选中的工作空间路径（空字符串表示使用默认值）
list: 已保存的工作空间路径列表
default: env_config 中的默认工作空间路径
snapshots_base: 快照存储基础目录
access_mode: 工作空间访问模式（restricted | approval | full）
"""
from agent.utils.env_utils import get_workspace_dir
from agent.memory.system_prompt import _current_workspace_dir
from agent.utils.agent_utils import SESSIONS_ROOT
from agent.core.workspace_policy import get_access_mode
import os
cfg = _read_ws_config()
current = _current_workspace_dir or cfg.get("current", "")
@@ -1175,9 +1326,27 @@ def api_get_workspace() -> dict[str, Any]:
"list": valid,
"default": get_workspace_dir(),
"snapshots_base": os.path.join(SESSIONS_ROOT, ".minor_snapshots").replace("\\", "/"),
"access_mode": get_access_mode(),
}
@app.post("/api/workspace/access-mode")
def api_set_access_mode(data: dict[str, Any]) -> dict[str, Any]:
"""设置工作空间访问模式。
输入: {"mode": "restricted"|"approval"|"full"}
输出: {"success": bool, "mode": str}
"""
from agent.core.workspace_policy import _ACCESS_MODES, set_access_mode
mode = data.get("mode", "") if isinstance(data, dict) else ""
if mode not in _ACCESS_MODES:
raise HTTPException(status_code=400, detail="无效的访问模式（restricted | approval | full）")
ok = set_access_mode(mode)
if not ok:
raise HTTPException(status_code=500, detail="保存访问模式失败")
return {"success": True, "mode": mode}
@app.post("/api/workspace/select")
def api_select_workspace(data: dict[str, Any]) -> dict[str, Any]:
"""选择工作空间。
@@ -1331,6 +1500,33 @@ def health() -> dict[str, str]:
return {"status": "ok"}
@app.get("/api/services/status")
def api_services_status() -> dict[str, Any]:
"""探测 ASR 与 TTS 服务的连通性（供前端麦克风/语音模式切换前检查）。
探测方式: 对服务根路径发起 GET，任何 HTTP 响应（含 404）都视为服务在线；
连接失败/超时视为服务不可用。
输出: {"asr": {"ok": bool, "url": str}, "tts": {"ok": bool, "url": str}}
"""
import requests as _req
from agent.utils.env_utils import ASR_BASE_URL, STREAMING_TTS_URL
def _probe(url: str) -> bool:
if not url or not str(url).strip():
return False
try:
_req.get(str(url).rstrip("/") + "/", timeout=3)
return True
except Exception:
return False
return {
"asr": {"ok": _probe(ASR_BASE_URL), "url": ASR_BASE_URL or ""},
"tts": {"ok": _probe(STREAMING_TTS_URL), "url": STREAMING_TTS_URL or ""},
}
@app.get("/api/pick-folder")
def api_pick_folder() -> dict[str, Any]:
"""打开原生文件夹选择对话框，返回用户选择的文件夹绝对路径。
@@ -1432,6 +1628,37 @@ def api_save_compress_settings(data: dict[str, Any]) -> dict[str, Any]:
return {"success": updated}
# ---------- 深度思考档位 API ----------
@app.post("/api/config/thinking-level")
def api_set_thinking_level(data: dict[str, Any]) -> dict[str, Any]:
"""设置深度思考档位（low | high | xhigh | max | ultra），重建 Agent 运行时。
档位同时决定运行模式与思考开关：
low = agent + 不思考；high = plan + 不思考；
xhigh = agent + 思考；max = plan + 思考；
ultra 为未来 react→审批图预留占位（暂按 agent + 思考）。
输入: {"level": str}
输出: {"success": bool, "level": str}
"""
from agent.utils.env_utils import THINKING_LEVELS, _load_config
level = data.get("level", "") if isinstance(data, dict) else ""
if level not in THINKING_LEVELS:
raise HTTPException(status_code=400, detail="无效的思考档位（low | high | xhigh | max | ultra）")
config = load_env_config()
config["THINKING_LEVEL"] = level
if not save_env_config(config):
raise HTTPException(status_code=500, detail="保存思考档位失败")
_load_config()
# 重建 Agent 运行时使新的 extra_body 生效
try:
from agent.agents.agent_runtime import reload_all_agent_runtimes
reload_all_agent_runtimes()
except Exception:
pass
return {"success": True, "level": level}
# ---------- 会话 Token 用量 API ----------
@app.get("/api/chat/tokens/{session_id}")
def api_get_session_tokens(session_id: str) -> dict[str, Any]:
@@ -1441,7 +1668,7 @@ def api_get_session_tokens(session_id: str) -> dict[str, Any]:
"compress_rate": float}
"""
from agent.utils.env_utils import LLM_CONTEXT_WINDOW, COMPRESS_RATE
from agent.history.tool_call_recorder import get_session_tokens
from agent.history.tool_call_recorder import get_session_tokens, get_session_token_breakdown
# 直接读取 LLM 每次调用后存储的真实 total_tokens
tokens = get_session_tokens(session_id)
@@ -1452,6 +1679,8 @@ def api_get_session_tokens(session_id: str) -> dict[str, Any]:
"context_window": LLM_CONTEXT_WINDOW,
"compress_threshold": compress_threshold,
"compress_rate": COMPRESS_RATE,
# 三类估算占比（消息/工具/系统提示词），供前端环形指示展示
"breakdown": get_session_token_breakdown(session_id),
}
@@ -1613,20 +1842,16 @@ async def api_create_or_update_skill(
name: str = Form(...),
description: str = Form(""),
content: str = Form(""),
version: str = Form("1.0.0"),
author: str = Form(""),
tags: str = Form(""),
enable: str = Form("true"),
) -> dict[str, Any]:
"""创建或更新一个 skill。
"""创建或更新一个 skill（不保留 version/author）。
输入:
name: skill 名称（必填）。
description: 简述。
content: skill.md 的 markdown 内容。
version: 版本号。
author: 作者。
tags: 逗号分隔的标签字符串。
tags: 逗号分隔的标签字符串（可选，便于分级筛选）。
enable: 是否启用，默认 "true"。
输出: {"success": bool, "skill": dict}
@@ -1640,8 +1865,6 @@ async def api_create_or_update_skill(
name=name.strip(),
description=description,
content=content,
version=version,
author=author,
tags=tag_list,
enable=enable_bool,
)
@@ -1652,6 +1875,59 @@ async def api_create_or_update_skill(
return {"success": True, "skill": skill}
# ---------- Skill 导入（zip：先解析展示，保存时才落盘） ----------
@app.post("/api/skills/parse")
async def api_parse_skill_zip(file: UploadFile = File(...)) -> dict[str, Any]:
"""解析 skill zip 包（不落盘），返回 name/description/content 与附件列表。
输入: multipart file（zip）
输出: {"found": bool, "skill": {name, description, content, attachments} | None,
"error": str | None}
"""
if not file.filename:
raise HTTPException(status_code=400, detail="文件名为空")
zip_bytes = await file.read()
from agent.skills import parse_skill_zip
parsed = parse_skill_zip(zip_bytes)
if parsed is None:
return {"found": False, "skill": None,
"error": "未在 zip 的一级子目录下找到 skill.md/SKILL.md"}
return {"found": True, "skill": parsed, "error": None}
@app.post("/api/skills/import")
async def api_import_skill_zip(
file: UploadFile = File(...),
name: str = Form(""),
description: str = Form(""),
content: str = Form(""),
tags: str = Form(""),
enable: str = Form("true"),
) -> dict[str, Any]:
"""导入 skill zip 包到 skills 目录（skill.md/skill.json + 附件保持原相对路径）。
输入: multipart file（zip）+ name/description/content/tags/enable 覆盖值
输出: {"success": bool, "skill": dict | None}
"""
if not file.filename:
raise HTTPException(status_code=400, detail="文件名为空")
zip_bytes = await file.read()
from agent.skills import import_skill_zip, get_skill
tag_list = [t.strip() for t in tags.split(",") if t.strip()] if tags else []
enable_bool = enable.lower() not in ("false", "0", "no", "")
ok = import_skill_zip(
name=name.strip(),
zip_bytes=zip_bytes,
description=description,
content=content,
tags=tag_list,
enable=enable_bool,
)
if not ok:
raise HTTPException(status_code=500, detail="导入 skill 失败（zip 无效或未找到 skill.md）")
return {"success": True, "skill": get_skill(name.strip())}
@app.delete("/api/skills/{name}")
def api_delete_skill(name: str) -> dict[str, Any]:
"""删除一个 skill 及其所有附件。