mirror of
https://github.com/DrizzleTime/Foxel.git
synced 2026-06-08 17:09:59 +08:00
feat: Implement AI Agent with enhanced tool processing capabilities (#89)
* feat: Implement AI Agent with tool processing capabilities - Added tools for listing and running processors in the agent. - Created data models for agent chat requests and tool calls. - Developed API integration for agent chat and streaming responses. - Built the AI Agent widget with a user interface for interaction. - Styled the agent components for better user experience. * feat: 增强 AI 助手工具功能，添加文件操作和搜索功能，更新界面显示 * feat: 更新 AI 助手组件 * feat: 更新 AiAgentWidget 组件样式，调整背景和边距以提升界面一致性
This commit is contained in:
时雨
448
domain/agent/service.py
Normal file
448
domain/agent/service.py
Normal file
@@ -0,0 +1,448 @@
|
||||
import asyncio
|
||||
import json
|
||||
import uuid
|
||||
from typing import Any, Dict, List, Optional, Tuple
|
||||
|
||||
import httpx
|
||||
from fastapi import HTTPException
|
||||
|
||||
from domain.agent.tools import get_tool, openai_tools, tool_result_to_content
|
||||
from domain.agent.types import AgentChatRequest, PendingToolCall
|
||||
from domain.ai.inference import MissingModelError, chat_completion, chat_completion_stream
|
||||
from domain.ai.service import AIProviderService
|
||||
from domain.auth.types import User
|
||||
|
||||
|
||||
def _normalize_path(p: Optional[str]) -> Optional[str]:
|
||||
if not p:
|
||||
return None
|
||||
s = str(p).strip()
|
||||
if not s:
|
||||
return None
|
||||
s = s.replace("\\", "/")
|
||||
if not s.startswith("/"):
|
||||
s = "/" + s
|
||||
s = s.rstrip("/") or "/"
|
||||
return s
|
||||
|
||||
|
||||
def _build_system_prompt(current_path: Optional[str]) -> str:
|
||||
lines = [
|
||||
"你是 Foxel 的 AI 助手。",
|
||||
"你可以通过工具对文件/目录进行查询、读写、移动、复制、删除，以及运行处理器（processor）。",
|
||||
"",
|
||||
"可用工具：",
|
||||
"- vfs_list_dir：浏览目录（列出 entries + pagination）。",
|
||||
"- vfs_stat：查看文件/目录信息。",
|
||||
"- vfs_read_text：读取文本文件内容（不支持二进制）。",
|
||||
"- vfs_search：搜索文件（vector/filename）。",
|
||||
"- vfs_write_text：写入文本文件内容（覆盖）。",
|
||||
"- vfs_mkdir：创建目录。",
|
||||
"- vfs_delete：删除文件或目录。",
|
||||
"- vfs_move：移动路径。",
|
||||
"- vfs_copy：复制路径。",
|
||||
"- vfs_rename：重命名路径。",
|
||||
"- processors_list：获取可用处理器列表（含 type/name/config_schema/produces_file/supports_directory）。",
|
||||
"- processors_run：运行处理器处理文件或目录（会返回 task_id 或 task_ids）。",
|
||||
"",
|
||||
"规则：",
|
||||
"1) 读操作（vfs_list_dir/vfs_stat/vfs_read_text/vfs_search）可直接调用工具。",
|
||||
"2) 写/改/删操作（vfs_write_text/vfs_mkdir/vfs_delete/vfs_move/vfs_copy/vfs_rename/processors_run）默认需要用户确认；只有在开启自动执行时才应直接执行。",
|
||||
"3) 用户未给出明确路径时先追问；若提供了“当前文件管理目录”，可以基于它把相对描述补全为绝对路径（以 / 开头）。",
|
||||
"4) 修改文件内容：先读取（vfs_read_text）→给出改动点→确认后再写入（vfs_write_text）。",
|
||||
"5) processors_run 返回任务 id 后，说明任务已提交，可在任务队列查看进度。",
|
||||
"6) 回答保持简洁中文。",
|
||||
]
|
||||
if current_path:
|
||||
lines.append("")
|
||||
lines.append(f"当前文件管理目录：{current_path}")
|
||||
return "\n".join(lines)
|
||||
|
||||
|
||||
def _ensure_tool_call_ids(message: Dict[str, Any]) -> Dict[str, Any]:
|
||||
tool_calls = message.get("tool_calls")
|
||||
if not isinstance(tool_calls, list):
|
||||
return message
|
||||
|
||||
changed = False
|
||||
for idx, call in enumerate(tool_calls):
|
||||
if not isinstance(call, dict):
|
||||
continue
|
||||
call_id = call.get("id")
|
||||
if isinstance(call_id, str) and call_id.strip():
|
||||
continue
|
||||
call["id"] = f"call_{idx}"
|
||||
changed = True
|
||||
|
||||
if changed:
|
||||
message["tool_calls"] = tool_calls
|
||||
return message
|
||||
|
||||
|
||||
def _extract_pending(tool_call: Dict[str, Any], requires_confirmation: bool) -> PendingToolCall:
|
||||
call_id = str(tool_call.get("id") or "")
|
||||
fn = tool_call.get("function") or {}
|
||||
name = str((fn.get("name") if isinstance(fn, dict) else None) or "")
|
||||
raw_args = fn.get("arguments") if isinstance(fn, dict) else None
|
||||
arguments: Dict[str, Any] = {}
|
||||
if isinstance(raw_args, str) and raw_args.strip():
|
||||
try:
|
||||
parsed = json.loads(raw_args)
|
||||
if isinstance(parsed, dict):
|
||||
arguments = parsed
|
||||
except json.JSONDecodeError:
|
||||
arguments = {}
|
||||
return PendingToolCall(
|
||||
id=call_id,
|
||||
name=name,
|
||||
arguments=arguments,
|
||||
requires_confirmation=requires_confirmation,
|
||||
)
|
||||
|
||||
|
||||
def _find_last_assistant_tool_calls(messages: List[Dict[str, Any]]) -> Tuple[int, Dict[str, Any]]:
|
||||
for idx in range(len(messages) - 1, -1, -1):
|
||||
msg = messages[idx]
|
||||
if not isinstance(msg, dict):
|
||||
continue
|
||||
if msg.get("role") != "assistant":
|
||||
continue
|
||||
tool_calls = msg.get("tool_calls")
|
||||
if isinstance(tool_calls, list) and tool_calls:
|
||||
return idx, msg
|
||||
raise HTTPException(status_code=400, detail="没有可确认的待执行操作")
|
||||
|
||||
|
||||
def _existing_tool_result_ids(messages: List[Dict[str, Any]]) -> set[str]:
|
||||
ids: set[str] = set()
|
||||
for msg in messages:
|
||||
if not isinstance(msg, dict):
|
||||
continue
|
||||
if msg.get("role") != "tool":
|
||||
continue
|
||||
tool_call_id = msg.get("tool_call_id")
|
||||
if isinstance(tool_call_id, str) and tool_call_id.strip():
|
||||
ids.add(tool_call_id)
|
||||
return ids
|
||||
|
||||
|
||||
async def _choose_chat_ability() -> str:
|
||||
tools_model = await AIProviderService.get_default_model("tools")
|
||||
return "tools" if tools_model else "chat"
|
||||
|
||||
|
||||
def _sse(event: str, data: Any) -> bytes:
|
||||
payload = json.dumps(data, ensure_ascii=False, separators=(",", ":"))
|
||||
return f"event: {event}\ndata: {payload}\n\n".encode("utf-8")
|
||||
|
||||
|
||||
class AgentService:
|
||||
@classmethod
|
||||
async def chat(cls, req: AgentChatRequest, user: Optional[User]) -> Dict[str, Any]:
|
||||
history: List[Dict[str, Any]] = list(req.messages or [])
|
||||
current_path = _normalize_path(req.context.current_path if req.context else None)
|
||||
|
||||
system_prompt = _build_system_prompt(current_path)
|
||||
internal_messages: List[Dict[str, Any]] = [{"role": "system", "content": system_prompt}] + history
|
||||
|
||||
new_messages: List[Dict[str, Any]] = []
|
||||
pending: List[PendingToolCall] = []
|
||||
|
||||
approved_ids = {i for i in (req.approved_tool_call_ids or []) if isinstance(i, str) and i.strip()}
|
||||
rejected_ids = {i for i in (req.rejected_tool_call_ids or []) if isinstance(i, str) and i.strip()}
|
||||
|
||||
if approved_ids or rejected_ids:
|
||||
_, last_call_msg = _find_last_assistant_tool_calls(internal_messages)
|
||||
last_call_msg = _ensure_tool_call_ids(last_call_msg)
|
||||
tool_calls = last_call_msg.get("tool_calls") or []
|
||||
call_map: Dict[str, Dict[str, Any]] = {
|
||||
str(c.get("id")): c
|
||||
for c in tool_calls
|
||||
if isinstance(c, dict) and isinstance(c.get("id"), str)
|
||||
}
|
||||
|
||||
existing_ids = _existing_tool_result_ids(internal_messages)
|
||||
for call_id in approved_ids | rejected_ids:
|
||||
if call_id in existing_ids:
|
||||
continue
|
||||
tool_call = call_map.get(call_id)
|
||||
if not tool_call:
|
||||
continue
|
||||
fn = tool_call.get("function") or {}
|
||||
name = fn.get("name") if isinstance(fn, dict) else None
|
||||
args_raw = fn.get("arguments") if isinstance(fn, dict) else None
|
||||
args: Dict[str, Any] = {}
|
||||
if isinstance(args_raw, str) and args_raw.strip():
|
||||
try:
|
||||
parsed = json.loads(args_raw)
|
||||
if isinstance(parsed, dict):
|
||||
args = parsed
|
||||
except json.JSONDecodeError:
|
||||
args = {}
|
||||
|
||||
spec = get_tool(str(name or ""))
|
||||
if call_id in rejected_ids:
|
||||
content = tool_result_to_content({"canceled": True, "reason": "user_rejected"})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
continue
|
||||
|
||||
if not spec:
|
||||
content = tool_result_to_content({"error": f"unknown_tool: {name}"})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
continue
|
||||
|
||||
try:
|
||||
result = await spec.handler(args)
|
||||
content = tool_result_to_content(result)
|
||||
except Exception as exc: # noqa: BLE001
|
||||
content = tool_result_to_content({"error": str(exc)})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
|
||||
tools_schema = openai_tools()
|
||||
ability = await _choose_chat_ability()
|
||||
max_loops = 4
|
||||
|
||||
for _ in range(max_loops):
|
||||
try:
|
||||
assistant = await chat_completion(
|
||||
internal_messages,
|
||||
ability=ability,
|
||||
tools=tools_schema,
|
||||
tool_choice="auto",
|
||||
timeout=60.0,
|
||||
)
|
||||
except MissingModelError as exc:
|
||||
raise HTTPException(status_code=400, detail=str(exc)) from exc
|
||||
except httpx.HTTPStatusError as exc:
|
||||
raise HTTPException(status_code=502, detail=f"对话请求失败: {exc}") from exc
|
||||
except httpx.RequestError as exc:
|
||||
raise HTTPException(status_code=502, detail=f"对话请求异常: {exc}") from exc
|
||||
|
||||
assistant = _ensure_tool_call_ids(assistant)
|
||||
internal_messages.append(assistant)
|
||||
new_messages.append(assistant)
|
||||
|
||||
tool_calls = assistant.get("tool_calls")
|
||||
if not isinstance(tool_calls, list) or not tool_calls:
|
||||
break
|
||||
|
||||
pending = []
|
||||
for call in tool_calls:
|
||||
if not isinstance(call, dict):
|
||||
continue
|
||||
call_id = str(call.get("id") or "")
|
||||
fn = call.get("function") or {}
|
||||
name = fn.get("name") if isinstance(fn, dict) else None
|
||||
args_raw = fn.get("arguments") if isinstance(fn, dict) else None
|
||||
args: Dict[str, Any] = {}
|
||||
if isinstance(args_raw, str) and args_raw.strip():
|
||||
try:
|
||||
parsed = json.loads(args_raw)
|
||||
if isinstance(parsed, dict):
|
||||
args = parsed
|
||||
except json.JSONDecodeError:
|
||||
args = {}
|
||||
|
||||
spec = get_tool(str(name or ""))
|
||||
if not spec:
|
||||
content = tool_result_to_content({"error": f"unknown_tool: {name}"})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
continue
|
||||
|
||||
if spec.requires_confirmation and not req.auto_execute:
|
||||
pending.append(_extract_pending(call, True))
|
||||
continue
|
||||
|
||||
try:
|
||||
result = await spec.handler(args)
|
||||
content = tool_result_to_content(result)
|
||||
except Exception as exc: # noqa: BLE001
|
||||
content = tool_result_to_content({"error": str(exc)})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
|
||||
if pending:
|
||||
break
|
||||
|
||||
payload: Dict[str, Any] = {"messages": new_messages}
|
||||
if pending:
|
||||
payload["pending_tool_calls"] = [p.model_dump() for p in pending]
|
||||
return payload
|
||||
|
||||
@classmethod
|
||||
async def chat_stream(cls, req: AgentChatRequest, user: Optional[User]):
|
||||
history: List[Dict[str, Any]] = list(req.messages or [])
|
||||
current_path = _normalize_path(req.context.current_path if req.context else None)
|
||||
|
||||
system_prompt = _build_system_prompt(current_path)
|
||||
internal_messages: List[Dict[str, Any]] = [{"role": "system", "content": system_prompt}] + history
|
||||
|
||||
new_messages: List[Dict[str, Any]] = []
|
||||
pending: List[PendingToolCall] = []
|
||||
|
||||
approved_ids = {i for i in (req.approved_tool_call_ids or []) if isinstance(i, str) and i.strip()}
|
||||
rejected_ids = {i for i in (req.rejected_tool_call_ids or []) if isinstance(i, str) and i.strip()}
|
||||
|
||||
try:
|
||||
if approved_ids or rejected_ids:
|
||||
_, last_call_msg = _find_last_assistant_tool_calls(internal_messages)
|
||||
last_call_msg = _ensure_tool_call_ids(last_call_msg)
|
||||
tool_calls = last_call_msg.get("tool_calls") or []
|
||||
call_map: Dict[str, Dict[str, Any]] = {
|
||||
str(c.get("id")): c
|
||||
for c in tool_calls
|
||||
if isinstance(c, dict) and isinstance(c.get("id"), str)
|
||||
}
|
||||
|
||||
existing_ids = _existing_tool_result_ids(internal_messages)
|
||||
for call_id in approved_ids | rejected_ids:
|
||||
if call_id in existing_ids:
|
||||
continue
|
||||
tool_call = call_map.get(call_id)
|
||||
if not tool_call:
|
||||
continue
|
||||
fn = tool_call.get("function") or {}
|
||||
name = fn.get("name") if isinstance(fn, dict) else None
|
||||
args_raw = fn.get("arguments") if isinstance(fn, dict) else None
|
||||
args: Dict[str, Any] = {}
|
||||
if isinstance(args_raw, str) and args_raw.strip():
|
||||
try:
|
||||
parsed = json.loads(args_raw)
|
||||
if isinstance(parsed, dict):
|
||||
args = parsed
|
||||
except json.JSONDecodeError:
|
||||
args = {}
|
||||
|
||||
spec = get_tool(str(name or ""))
|
||||
if call_id in rejected_ids:
|
||||
content = tool_result_to_content({"canceled": True, "reason": "user_rejected"})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
yield _sse("tool_end", {"tool_call_id": call_id, "name": str(name or ""), "message": tool_msg})
|
||||
continue
|
||||
|
||||
if not spec:
|
||||
content = tool_result_to_content({"error": f"unknown_tool: {name}"})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
yield _sse("tool_end", {"tool_call_id": call_id, "name": str(name or ""), "message": tool_msg})
|
||||
continue
|
||||
|
||||
yield _sse("tool_start", {"tool_call_id": call_id, "name": spec.name})
|
||||
try:
|
||||
result = await spec.handler(args)
|
||||
content = tool_result_to_content(result)
|
||||
except Exception as exc: # noqa: BLE001
|
||||
content = tool_result_to_content({"error": str(exc)})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
yield _sse("tool_end", {"tool_call_id": call_id, "name": spec.name, "message": tool_msg})
|
||||
|
||||
tools_schema = openai_tools()
|
||||
ability = await _choose_chat_ability()
|
||||
max_loops = 4
|
||||
|
||||
for _ in range(max_loops):
|
||||
assistant_event_id = uuid.uuid4().hex
|
||||
yield _sse("assistant_start", {"id": assistant_event_id})
|
||||
|
||||
assistant_message: Dict[str, Any] | None = None
|
||||
try:
|
||||
async for event in chat_completion_stream(
|
||||
internal_messages,
|
||||
ability=ability,
|
||||
tools=tools_schema,
|
||||
tool_choice="auto",
|
||||
timeout=60.0,
|
||||
):
|
||||
if event.get("type") == "delta":
|
||||
delta = event.get("delta")
|
||||
if isinstance(delta, str) and delta:
|
||||
yield _sse("assistant_delta", {"id": assistant_event_id, "delta": delta})
|
||||
elif event.get("type") == "message":
|
||||
msg = event.get("message")
|
||||
if isinstance(msg, dict):
|
||||
assistant_message = msg
|
||||
except MissingModelError as exc:
|
||||
raise HTTPException(status_code=400, detail=str(exc)) from exc
|
||||
except httpx.HTTPStatusError as exc:
|
||||
raise HTTPException(status_code=502, detail=f"对话请求失败: {exc}") from exc
|
||||
except httpx.RequestError as exc:
|
||||
raise HTTPException(status_code=502, detail=f"对话请求异常: {exc}") from exc
|
||||
|
||||
if not assistant_message:
|
||||
assistant_message = {"role": "assistant", "content": ""}
|
||||
|
||||
assistant_message = _ensure_tool_call_ids(assistant_message)
|
||||
internal_messages.append(assistant_message)
|
||||
new_messages.append(assistant_message)
|
||||
yield _sse("assistant_end", {"id": assistant_event_id, "message": assistant_message})
|
||||
|
||||
tool_calls = assistant_message.get("tool_calls")
|
||||
if not isinstance(tool_calls, list) or not tool_calls:
|
||||
break
|
||||
|
||||
pending = []
|
||||
for call in tool_calls:
|
||||
if not isinstance(call, dict):
|
||||
continue
|
||||
call_id = str(call.get("id") or "")
|
||||
fn = call.get("function") or {}
|
||||
name = fn.get("name") if isinstance(fn, dict) else None
|
||||
args_raw = fn.get("arguments") if isinstance(fn, dict) else None
|
||||
args: Dict[str, Any] = {}
|
||||
if isinstance(args_raw, str) and args_raw.strip():
|
||||
try:
|
||||
parsed = json.loads(args_raw)
|
||||
if isinstance(parsed, dict):
|
||||
args = parsed
|
||||
except json.JSONDecodeError:
|
||||
args = {}
|
||||
|
||||
spec = get_tool(str(name or ""))
|
||||
if not spec:
|
||||
content = tool_result_to_content({"error": f"unknown_tool: {name}"})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
yield _sse("tool_end", {"tool_call_id": call_id, "name": str(name or ""), "message": tool_msg})
|
||||
continue
|
||||
|
||||
if spec.requires_confirmation and not req.auto_execute:
|
||||
pending.append(_extract_pending(call, True))
|
||||
continue
|
||||
|
||||
yield _sse("tool_start", {"tool_call_id": call_id, "name": spec.name})
|
||||
try:
|
||||
result = await spec.handler(args)
|
||||
content = tool_result_to_content(result)
|
||||
except Exception as exc: # noqa: BLE001
|
||||
content = tool_result_to_content({"error": str(exc)})
|
||||
tool_msg = {"role": "tool", "tool_call_id": call_id, "content": content}
|
||||
internal_messages.append(tool_msg)
|
||||
new_messages.append(tool_msg)
|
||||
yield _sse("tool_end", {"tool_call_id": call_id, "name": spec.name, "message": tool_msg})
|
||||
|
||||
if pending:
|
||||
yield _sse("pending", {"pending_tool_calls": [p.model_dump() for p in pending]})
|
||||
break
|
||||
|
||||
payload: Dict[str, Any] = {"messages": new_messages}
|
||||
if pending:
|
||||
payload["pending_tool_calls"] = [p.model_dump() for p in pending]
|
||||
yield _sse("done", payload)
|
||||
|
||||
except asyncio.CancelledError:
|
||||
return
|
||||
Reference in New Issue
Block a user