1233 lines
19 KiB
Python
1233 lines
19 KiB
Python
# Auto-extracted from chat.py during Phase 1 refactor (Strangler Fig).
|
||
# Source of truth: scripts/phase1_extract.py. Do not hand-merge back.
|
||
|
||
from fastapi import APIRouter, Depends, HTTPException, Query, Response
|
||
from sqlalchemy.orm import Session
|
||
from sqlalchemy import and_, desc, asc, func
|
||
from pydantic import BaseModel, Field
|
||
from typing import Optional
|
||
import uuid
|
||
from datetime import datetime, timezone, timedelta
|
||
import re
|
||
import json
|
||
import asyncio
|
||
import math
|
||
import httpx
|
||
import os
|
||
import logging
|
||
from ..database import get_db, SessionLocal
|
||
from ..models import (
|
||
|
||
|
||
|
||
User, Conversation, Message, AIModel, ModelProvider, TokenUsageLog,
|
||
|
||
|
||
|
||
KnowledgeBase, KnowledgeDocument, KnowledgeDocumentChunk, EvolutionSkill,
|
||
|
||
|
||
|
||
EvolutionTrajectory, MarketApp, AppPurchase, AppInstallation,
|
||
|
||
|
||
|
||
MCPServer,
|
||
|
||
|
||
|
||
gen_uuid
|
||
|
||
|
||
|
||
)
|
||
from ..auth import get_current_user
|
||
from ..platform_knowledge import get_platform_knowledge
|
||
from ..mcp_client import list_mcp_tools, call_mcp_tool
|
||
from ..middleware.log_service import LogService
|
||
from ..capability_registry import resolve_auto_model
|
||
from app.api.chat_skill import (_bm25_search)
|
||
from app.api.chat_utils import (logger)
|
||
|
||
def _retrieve_kb_context(db: Session, kb_ids: list, query: str, max_chars: int = 3000) -> str:
|
||
|
||
|
||
|
||
"""
|
||
|
||
|
||
|
||
Stage C: 增强版知识库检索 — 使用BM25排序替代朴素拼接
|
||
|
||
|
||
|
||
1. 获取每个KB的文档
|
||
|
||
|
||
|
||
2. 构建文档列表(content+title+metadata)
|
||
|
||
|
||
|
||
3. BM25检索最相关的段落
|
||
|
||
|
||
|
||
4. 拼接为上下文注入system prompt
|
||
|
||
|
||
|
||
同时叠加 Dify 外部知识库(RAG)检索结果。
|
||
|
||
|
||
|
||
"""
|
||
|
||
|
||
|
||
local_context = ""
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
if kb_ids:
|
||
|
||
|
||
|
||
kbs = db.query(KnowledgeBase).filter(KnowledgeBase.id.in_(kb_ids)).all()
|
||
|
||
|
||
|
||
if kbs:
|
||
|
||
|
||
|
||
# 收集所有文档段落
|
||
|
||
|
||
|
||
# Stage C.1: 优先使用 KnowledgeDocumentChunk 表的预分段数据
|
||
|
||
|
||
|
||
all_segments = []
|
||
|
||
|
||
|
||
for kb in kbs:
|
||
|
||
|
||
|
||
# 先查chunk表
|
||
|
||
|
||
|
||
chunks = db.query(KnowledgeDocumentChunk).filter(
|
||
|
||
|
||
|
||
KnowledgeDocumentChunk.knowledge_base_id == kb.id
|
||
|
||
|
||
|
||
).order_by(KnowledgeDocumentChunk.chunk_index).limit(30).all()
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
if chunks:
|
||
|
||
|
||
|
||
# 使用预分段数据
|
||
|
||
|
||
|
||
for chunk in chunks:
|
||
|
||
|
||
|
||
if chunk.content and len(chunk.content.strip()) >= 20:
|
||
|
||
|
||
|
||
all_segments.append({
|
||
|
||
|
||
|
||
"content": "[{}/{}/段{}]: {}".format(
|
||
|
||
|
||
|
||
kb.name, chunk.document_id[:8], chunk.chunk_index, chunk.content[:500]
|
||
|
||
|
||
|
||
),
|
||
|
||
|
||
|
||
"raw_content": chunk.content,
|
||
|
||
|
||
|
||
"kb_name": kb.name,
|
||
|
||
|
||
|
||
})
|
||
|
||
|
||
|
||
else:
|
||
|
||
|
||
|
||
# 退化: 使用原始文档分段
|
||
|
||
|
||
|
||
docs = db.query(KnowledgeDocument).filter(
|
||
|
||
|
||
|
||
KnowledgeDocument.knowledge_base_id == kb.id,
|
||
|
||
|
||
|
||
KnowledgeDocument.status == "completed"
|
||
|
||
|
||
|
||
).order_by(KnowledgeDocument.updated_at.desc()).limit(20).all()
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
for doc in docs:
|
||
|
||
|
||
|
||
content = doc.content or ""
|
||
|
||
|
||
|
||
if not content:
|
||
|
||
|
||
|
||
continue
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# 如果文档较长, 分段处理 (每段~800字)
|
||
|
||
|
||
|
||
if len(content) > 800:
|
||
|
||
|
||
|
||
paragraphs = re.split(r'\n\n+', content)
|
||
|
||
|
||
|
||
for pi, para in enumerate(paragraphs):
|
||
|
||
|
||
|
||
if len(para.strip()) >= 20:
|
||
|
||
|
||
|
||
all_segments.append({
|
||
|
||
|
||
|
||
"content": "[{}/{}/段{}]: {}".format(
|
||
|
||
|
||
|
||
kb.name, doc.name, pi+1, para.strip()[:500]
|
||
|
||
|
||
|
||
),
|
||
|
||
|
||
|
||
"raw_content": para.strip(),
|
||
|
||
|
||
|
||
"kb_name": kb.name,
|
||
|
||
|
||
|
||
"doc_name": doc.name,
|
||
|
||
|
||
|
||
})
|
||
|
||
|
||
|
||
else:
|
||
|
||
|
||
|
||
all_segments.append({
|
||
|
||
|
||
|
||
"content": "[{}/{}]: {}".format(kb.name, doc.name, content[:500]),
|
||
|
||
|
||
|
||
"raw_content": content,
|
||
|
||
|
||
|
||
"kb_name": kb.name,
|
||
|
||
|
||
|
||
"doc_name": doc.name,
|
||
|
||
|
||
|
||
})
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
if not all_segments:
|
||
|
||
|
||
|
||
# 退化: 使用原始朴素拼接
|
||
|
||
|
||
|
||
kb_names = ", ".join([k.name for k in kbs])
|
||
|
||
|
||
|
||
local_context = f"参考知识库: {kb_names}"
|
||
|
||
|
||
|
||
else:
|
||
|
||
|
||
|
||
# BM25检索最相关的段落
|
||
|
||
|
||
|
||
top_segments = _bm25_search(query, all_segments, top_k=8)
|
||
|
||
|
||
|
||
context_parts = [seg["content"] for seg in top_segments]
|
||
|
||
|
||
|
||
local_context = "\n".join(context_parts)
|
||
|
||
|
||
|
||
if len(local_context) > max_chars:
|
||
|
||
|
||
|
||
local_context = local_context[:max_chars] + "...(内容截断)"
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# 叠加 Dify 外部知识库(RAG)检索
|
||
|
||
|
||
|
||
dify_context = ""
|
||
|
||
|
||
|
||
try:
|
||
|
||
|
||
|
||
from app.dify_kb import get_dify_kb_context
|
||
|
||
|
||
|
||
dify_context = get_dify_kb_context(db, query, top_k=5)
|
||
|
||
|
||
|
||
except Exception as e:
|
||
|
||
|
||
|
||
logger.warning(f"Dify KB context retrieval failed: {e}")
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
parts = []
|
||
|
||
|
||
|
||
if local_context:
|
||
|
||
|
||
|
||
parts.append(local_context)
|
||
|
||
|
||
|
||
if dify_context:
|
||
|
||
|
||
|
||
parts.append("【Dify 知识库检索结果】\n" + dify_context)
|
||
|
||
|
||
|
||
return "\n\n".join(parts)
|
||
|
||
async def _ingest_answer_to_dify(conv_id: str, msg_id: str, content: str):
|
||
|
||
|
||
|
||
"""后台任务：将助手回答入库到 Dify 知识库（量化后可供 RAG 检索复用）。"""
|
||
|
||
|
||
|
||
try:
|
||
|
||
|
||
|
||
from app.dify_kb import aingest_text
|
||
|
||
|
||
|
||
from app.database import SessionLocal
|
||
|
||
|
||
|
||
from sqlalchemy import text as _t
|
||
|
||
|
||
|
||
db = SessionLocal()
|
||
|
||
|
||
|
||
try:
|
||
|
||
|
||
|
||
row = db.execute(_t(
|
||
|
||
|
||
|
||
"SELECT api_url, api_key, config FROM external_kb_services "
|
||
|
||
|
||
|
||
"WHERE service_type='dify' AND is_enabled=1 LIMIT 1"
|
||
|
||
|
||
|
||
)).fetchone()
|
||
|
||
|
||
|
||
finally:
|
||
|
||
|
||
|
||
db.close()
|
||
|
||
|
||
|
||
if not row:
|
||
|
||
|
||
|
||
return
|
||
|
||
|
||
|
||
api_url, api_key, config_raw = row
|
||
|
||
|
||
|
||
dataset_name = None
|
||
|
||
|
||
|
||
if config_raw:
|
||
|
||
|
||
|
||
try:
|
||
|
||
|
||
|
||
import json
|
||
|
||
|
||
|
||
cfg = config_raw if isinstance(config_raw, dict) else json.loads(config_raw)
|
||
|
||
|
||
|
||
dataset_name = cfg.get("dataset_name")
|
||
|
||
|
||
|
||
except Exception:
|
||
|
||
|
||
|
||
pass
|
||
|
||
|
||
|
||
await aingest_text(api_url, api_key, content, name=f"chat_{conv_id}_{msg_id}", dataset_name=dataset_name)
|
||
|
||
|
||
|
||
except Exception as e:
|
||
|
||
|
||
|
||
logger.warning(f"后台入库 Dify 失败: {e}")
|
||
|
||
def _build_system_prompt(
|
||
|
||
|
||
|
||
skills: list,
|
||
|
||
|
||
|
||
agents: list,
|
||
|
||
|
||
|
||
kb_context: str,
|
||
|
||
|
||
|
||
user_content: str,
|
||
|
||
|
||
|
||
mcp: list = None,
|
||
|
||
mcp_functions: list = None,
|
||
|
||
auto_discovered: bool = False,
|
||
all_mcp: list = None,
|
||
|
||
) -> tuple:
|
||
|
||
|
||
|
||
"""
|
||
|
||
|
||
|
||
动态组装system prompt + function definitions。
|
||
|
||
|
||
|
||
返回: (system_prompt_str, function_definitions_list)
|
||
|
||
|
||
|
||
"""
|
||
|
||
|
||
|
||
system_parts = ["你是AIFlow平台的AI助手，具备专业的知识储备和工具调用能力。"]
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# 平台知识库注入（始终开启）：确保AI优先使用平台已有能力
|
||
|
||
|
||
|
||
try:
|
||
|
||
|
||
|
||
pk = get_platform_knowledge()
|
||
|
||
|
||
|
||
if pk:
|
||
|
||
|
||
|
||
system_parts.append(f"【平台知识库——优先使用以下知识回答用户问题】\n{pk}")
|
||
|
||
|
||
|
||
except Exception:
|
||
|
||
|
||
|
||
pass
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# 回复格式要求：强制使用Markdown
|
||
|
||
|
||
|
||
system_parts.append("回复格式要求：使用 Markdown 格式，层次分明即可。对于简短问题，用 1-2 段自然文字直接回答，不必强行分点、加标题或表格；仅在确实需要分步骤或列举时才使用列表，代码/配置用 ``` 代码块包裹。避免冗余空行，保持紧凑、美观、易读。")
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
function_definitions = []
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# Stage B.1: Agent角色注入（Issue 2：让用户询问 agent 能力时，AI 能“看到”该 agent 资料）
|
||
|
||
|
||
|
||
if agents:
|
||
|
||
|
||
|
||
for agent in agents:
|
||
|
||
|
||
|
||
name = agent.get("name", "")
|
||
|
||
|
||
|
||
role = agent.get("role_prompt", "")
|
||
|
||
|
||
|
||
instructions = agent.get("instructions", "")
|
||
|
||
|
||
|
||
desc = agent.get("description", "")
|
||
|
||
|
||
|
||
block = []
|
||
|
||
|
||
|
||
if name:
|
||
|
||
|
||
|
||
block.append(
|
||
|
||
|
||
|
||
f"用户已选择智能体「{name}」。当用户询问该智能体的功能、能力、用法或相关问题时，"
|
||
|
||
|
||
|
||
f"请基于以下资料准确作答，不要说“不知道这个智能体”。"
|
||
|
||
|
||
|
||
)
|
||
|
||
|
||
|
||
if role:
|
||
|
||
|
||
|
||
block.append(f"该智能体的角色设定: {role}")
|
||
|
||
|
||
|
||
if desc:
|
||
|
||
|
||
|
||
block.append(f"该智能体的简介: {desc}")
|
||
|
||
|
||
|
||
if instructions:
|
||
|
||
|
||
|
||
block.append(f"该智能体的行为指引: {instructions}")
|
||
|
||
|
||
|
||
if block:
|
||
|
||
|
||
|
||
system_parts.append("\n".join(block))
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# Stage A.2: Skill工具注入
|
||
|
||
|
||
|
||
if skills:
|
||
|
||
|
||
|
||
skill_descs = ", ".join([f"{s['name']}({s.get('category', 'general')})" for s in skills])
|
||
|
||
|
||
|
||
system_parts.append(f"可用技能工具: {skill_descs}")
|
||
|
||
|
||
|
||
if auto_discovered:
|
||
|
||
|
||
|
||
# Issue 8：每轮默认先寻找本账号已安装且平台已部署的技能，多个适用时 AI 自行判断选 1 个最优
|
||
|
||
|
||
|
||
system_parts.append(
|
||
|
||
|
||
|
||
"重要规则: 以上技能为本账号已安装且平台已部署的技能。每轮对话请先判断用户请求是否可由其中某个技能处理；"
|
||
|
||
|
||
|
||
"若多个技能都适用，请自行判断并选择最贴切的 1 个优先调用，不要一次性全部调用。"
|
||
|
||
|
||
|
||
"调用对应工具后将获得执行结果，基于结果生成最终回复。若这些技能均不适用，则直接基于你的知识回答。"
|
||
|
||
|
||
|
||
)
|
||
|
||
|
||
|
||
else:
|
||
|
||
|
||
|
||
system_parts.append(
|
||
"重要规则：\n"
|
||
"1) 当用户请求需要由该技能【执行】某项任务时，优先通过 function calling 调用对应工具；调用后基于工具返回结果生成回复。\n"
|
||
"2) 当用户询问所选技能【本身的介绍、功能、用途、用法、特点】（例如“介绍一下这个技能”“它是什么”“怎么用”“能做什么”）时，"
|
||
"必须基于下方【已选技能资料】中的名称/描述/指令直接作答，不要泛泛介绍平台整体，也不要说“不知道这个技能”。\n"
|
||
"3) 若用户问题既不执行也不询问该技能本身，则按常规方式回答。"
|
||
)
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# 为每个skill构建function definition
|
||
|
||
|
||
|
||
for s in skills:
|
||
|
||
|
||
|
||
# 从skill_content提取指令注入system prompt
|
||
|
||
|
||
|
||
instructions = s.get("instructions", "")
|
||
|
||
|
||
|
||
name = s.get("name", "")
|
||
if not auto_discovered:
|
||
# 显式选中技能：注入可引用的技能资料，便于“介绍/说明”类问题据此作答
|
||
ref_parts = [f"【已选技能资料】名称：{name}"]
|
||
desc = s.get("description", "")
|
||
if desc:
|
||
ref_parts.append(f"简介：{desc}")
|
||
if instructions and len(instructions) < 1200:
|
||
ref_parts.append(f"使用指令：{instructions}")
|
||
system_parts.append("\n".join(ref_parts))
|
||
elif instructions and len(instructions) < 800:
|
||
system_parts.append(f"技能「{name}」的详细指令: {instructions}")
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# 构建参数schema
|
||
|
||
|
||
|
||
params = s.get("parameters", {})
|
||
|
||
|
||
|
||
if isinstance(params, dict) and len(params) > 0:
|
||
|
||
|
||
|
||
properties = {}
|
||
|
||
|
||
|
||
required = []
|
||
|
||
|
||
|
||
for pname, pdef in params.items():
|
||
|
||
|
||
|
||
if isinstance(pdef, dict):
|
||
|
||
|
||
|
||
properties[pname] = {
|
||
|
||
|
||
|
||
"type": pdef.get("type", "string"),
|
||
|
||
|
||
|
||
"description": pdef.get("description", f"传递给{s['name']}的{pname}参数"),
|
||
|
||
|
||
|
||
}
|
||
|
||
|
||
|
||
if pdef.get("required", True):
|
||
|
||
|
||
|
||
required.append(pname)
|
||
|
||
|
||
|
||
else:
|
||
|
||
|
||
|
||
properties[pname] = {"type": "string", "description": f"传递给{s['name']}的{pname}参数"}
|
||
|
||
|
||
|
||
required.append(pname)
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
if not properties:
|
||
|
||
|
||
|
||
properties = {"input": {"type": "string", "description": f"需要传递给{s['name']}的完整用户输入"}}
|
||
|
||
|
||
|
||
required = ["input"]
|
||
|
||
|
||
|
||
else:
|
||
|
||
|
||
|
||
properties = {"input": {"type": "string", "description": f"需要传递给{s['name']}的完整用户输入"}}
|
||
|
||
|
||
|
||
required = ["input"]
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
desc = s.get("description", "")
|
||
|
||
|
||
|
||
category = s.get("category", "general")
|
||
|
||
|
||
|
||
func_desc = f"调用技能「{s['name']}」处理用户请求。使用场景: {desc}。当用户请求涉及{category}类任务时优先调用此工具。"
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
function_definitions.append({
|
||
|
||
|
||
|
||
"name": s["func_name"],
|
||
|
||
|
||
|
||
"description": func_desc,
|
||
|
||
|
||
|
||
"parameters": {
|
||
|
||
|
||
|
||
"type": "object",
|
||
|
||
|
||
|
||
"properties": properties,
|
||
|
||
|
||
|
||
"required": required,
|
||
|
||
|
||
|
||
},
|
||
|
||
|
||
|
||
})
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# KB context injection
|
||
|
||
|
||
|
||
if kb_context:
|
||
|
||
|
||
|
||
system_parts.append(f"参考以下知识库内容回答用户问题:\n{kb_context}")
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# 平台全量 MCP 清单（Issue：让用户询问"平台有哪些 MCP"时 AI 能准确列举，而非凭空编造）
|
||
if all_mcp:
|
||
_mcp_inventory = "\n".join([f"- {m.get('name','')}: {m.get('description','') or '（未填写描述）'}" for m in all_mcp])
|
||
system_parts.append(
|
||
f"【平台已安装的 MCP 应用清单（共 {len(all_mcp)} 个，均为平台真实接入并启用）】\n{_mcp_inventory}\n"
|
||
"当用户询问“平台有哪些 MCP / 接入了哪些 MCP / 有哪些工具服务”时，请严格依据以上清单准确列举名称与功能，"
|
||
"不要编造或遗漏，也不要把未列出的服务说成已接入。"
|
||
)
|
||
|
||
# MCP 应用注入（Issue 3：让用户询问 MCP 能力时，AI 能“看到”该 MCP 的资料与地址）
|
||
|
||
|
||
|
||
if mcp:
|
||
|
||
|
||
|
||
mcp_lines = []
|
||
|
||
|
||
|
||
for m in mcp:
|
||
|
||
|
||
|
||
name = m.get("name", "")
|
||
|
||
|
||
|
||
desc = m.get("description", "")
|
||
|
||
|
||
|
||
transport = m.get("transport", "stdio")
|
||
|
||
|
||
|
||
endpoint = m.get("endpoint", "")
|
||
|
||
|
||
|
||
line = f"用户已选择 MCP 应用「{name}」（传输方式: {transport}）"
|
||
|
||
|
||
|
||
if desc:
|
||
|
||
|
||
|
||
line += f"，功能简介: {desc}"
|
||
|
||
|
||
|
||
if endpoint:
|
||
|
||
|
||
|
||
line += f"，服务地址: {endpoint}"
|
||
|
||
|
||
|
||
mcp_lines.append(line)
|
||
|
||
|
||
|
||
system_parts.append("\n".join(mcp_lines))
|
||
|
||
|
||
|
||
system_parts.append(
|
||
|
||
|
||
|
||
"用户已选择上述 MCP 应用。当请求需要借助外部工具/系统（如查询数据、调用 API、操作第三方服务）时，"
|
||
|
||
|
||
|
||
"请优先参考这些 MCP 应用的能力来回答；若平台已接入对应工具，优先通过工具调用完成；"
|
||
|
||
|
||
|
||
"不要说“不了解该 MCP”。"
|
||
|
||
|
||
|
||
)
|
||
|
||
# 高优先级覆盖规则：显式选中工具时，回答必须基于该工具资料
|
||
if (skills and not auto_discovered) or (mcp):
|
||
system_parts.append(
|
||
"【最高优先级覆盖规则】\n"
|
||
"用户在本轮已明确选择了工具（见上文【已选技能资料】或【用户已选择 MCP 应用】）。\n"
|
||
"如果用户的问题是在询问这些已选工具/技能/应用本身的介绍、功能、能做什么、怎么用、特点或用法，"
|
||
"你必须只基于上文提供的该工具具体资料作答；\n"
|
||
"不要引用《平台知识库》中关于技能市场、MCP 协议或平台整体能力的通用描述；"
|
||
"不要列举其它未选中的技能或 MCP 作为示例；也不要说“不知道”或“无法回答”。\n"
|
||
"若用户问题与已选工具本身无关，则按常规方式回答。"
|
||
)
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
system_prompt = "\n".join(system_parts)
|
||
|
||
|
||
|
||
# MCP function definitions 注入（让 LLM 可通过 function calling 真实调用 MCP 工具）
|
||
|
||
if mcp_functions:
|
||
|
||
function_definitions.extend(mcp_functions)
|
||
|
||
|
||
|
||
return system_prompt, function_definitions
|
||
|
||
def _get_conversation_context(db: Session, conv_id: str) -> dict:
|
||
|
||
|
||
|
||
"""
|
||
|
||
|
||
|
||
Stage B.2: 获取对话的持久化上下文
|
||
|
||
|
||
|
||
对话级别的agent/skill/kb选择保存在Conversation字段中。
|
||
|
||
|
||
|
||
"""
|
||
|
||
|
||
|
||
conv = db.query(Conversation).filter(Conversation.id == conv_id).first()
|
||
|
||
|
||
|
||
if not conv:
|
||
|
||
|
||
|
||
return {}
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# Conversation目前没有存储agent/skill/kb的JSON字段
|
||
|
||
|
||
|
||
# 我们从第一条消息中获取(如果存在), 或返回空
|
||
|
||
|
||
|
||
# 后续可通过数据库迁移添加Conversation.context字段
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
# 查询该对话中最早一条含有skill_ids/agent_ids/knowledge_ids的消息
|
||
|
||
|
||
|
||
first_tool_msg = db.query(Message).filter(
|
||
|
||
|
||
|
||
Message.conversation_id == conv_id,
|
||
|
||
|
||
|
||
Message.role == "user",
|
||
|
||
|
||
|
||
).order_by(asc(Message.created_at)).first()
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
context = {}
|
||
|
||
|
||
|
||
if first_tool_msg:
|
||
|
||
|
||
|
||
if first_tool_msg.skill_ids:
|
||
|
||
|
||
|
||
context["skill_ids"] = first_tool_msg.skill_ids
|
||
|
||
|
||
|
||
if first_tool_msg.agent_ids:
|
||
|
||
|
||
|
||
context["agent_ids"] = first_tool_msg.agent_ids
|
||
|
||
|
||
|
||
if first_tool_msg.knowledge_ids:
|
||
|
||
|
||
|
||
context["knowledge_base_ids"] = first_tool_msg.knowledge_ids
|
||
|
||
|
||
|
||
if getattr(first_tool_msg, "mcp_ids", None):
|
||
|
||
|
||
|
||
context["mcp_ids"] = first_tool_msg.mcp_ids
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
return context
|