bb16d3a5c9
访谈与阶段 - 新增 app/agents/stage_constants.py：集中 CHAT_STAGES、章节分类/顺序、阶段到默认 memoir 类别等，与 MemoirState 默认槽位顺序对齐；减少散落在 prompts 内的重复常量。 - 新增 app/agents/chat/prompt_context.py：以 ChatPromptContext 汇总 guided 系统提示所需字段（阶段、槽位、轮次、人设、记忆证据、回复长度模式、背景声线、职业等），统一走 get_guided_conversation_prompt。 - 大幅收敛 app/agents/chat/prompts_conversation.py；调整 prompts.py、stage_prompts.py、stage_detection.py；同步 interview_agent、profile_agent、helpers 与 state_schema，使对话侧构造提示的方式一致、可测。 回忆录流水线 - memoir/prompts.py 删除已迁至 stage_constants / 独立模板的大段常量与图片占位相关逻辑；classification / extraction / fidelity / narrative agents 与 orchest（全量历史仍可用于计数，注入模型时按轮次与字符上限截断）、image_prompt_fallback_disabled。 - dependencies 增加 get_llm_provider_fast（LRU 缓存，可与默认共用密钥与 base_url）。 任务与编排 - memoir_tasks：prepare_batches 注入 llm_fast；开启独立快档模型时打结构化日志。 - chapter_cover_tasks、story_image_tasks：与图片 prompt / JSON 工具路径或策略变更对齐（import 与行为一致）。 - story_pipeline_sync 等小处同步。 其它核心 - langchain_llm、text_normalize 随上述调用链微调。 开发者体验 - .cursor/settings.json：启用 redis-development、postman 插件。 测试 - 新增 test_image_prompt_policy：覆盖「禁止回退」等图片 prompt 策略。 - 更新 test_interview_prompts、test_interview_reply_length、test_experience_regressions、test_json_and_memory_utils，匹配新常量位置、json_utils 与对话/长度行为。
292 lines
11 KiB
Python
292 lines
11 KiB
Python
"""
|
||
ProfileAgent：用户资料收集 Specialist
|
||
负责提取资料、资料追问、资料收集开场白，不负责 Redis 持久化（由 Orchestrator 统一处理）
|
||
"""
|
||
|
||
import json
|
||
import time
|
||
from typing import Any, Dict, List, Optional
|
||
|
||
from langchain_core.messages import AIMessage, HumanMessage, SystemMessage
|
||
|
||
from app.agents.chat.helpers import format_history_string, get_history_with_window
|
||
from app.agents.chat.prompts_profile import (
|
||
get_profile_extraction_prompt,
|
||
get_profile_followup_prompt,
|
||
get_profile_greeting_prompt,
|
||
)
|
||
from app.core.agent_logging import agent_span, log_agent_payload, log_agent_summary
|
||
from app.core.config import settings
|
||
from app.core.dependencies import get_llm_provider
|
||
from app.core.json_utils import extract_json_payload
|
||
from app.core.langchain_llm import ainvoke_json_object
|
||
from app.core.logging import get_logger
|
||
from app.agents.chat.reply_limits import (
|
||
nonempty_segments_or_fallback,
|
||
segments_from_llm_response,
|
||
truncate_chat_segments,
|
||
)
|
||
|
||
logger = get_logger(__name__)
|
||
|
||
|
||
def _get_langchain_llm():
|
||
try:
|
||
provider = get_llm_provider()
|
||
return getattr(provider, "langchain_llm", None)
|
||
except Exception:
|
||
return None
|
||
|
||
|
||
def _message_contents_char_count(messages: List[Any]) -> int:
|
||
n = 0
|
||
for m in messages:
|
||
c = getattr(m, "content", None)
|
||
if isinstance(c, str):
|
||
n += len(c)
|
||
return n
|
||
|
||
|
||
class ProfileAgent:
|
||
"""用户资料收集 Specialist Agent"""
|
||
|
||
def __init__(self):
|
||
self.llm = _get_langchain_llm()
|
||
|
||
async def extract_profile_from_message(
|
||
self,
|
||
user_message: str,
|
||
missing_fields: List[str],
|
||
conversation_id: Optional[str] = None,
|
||
) -> Dict[str, Any]:
|
||
"""从用户消息中提取资料字段，不持久化"""
|
||
if not self.llm or not missing_fields:
|
||
return {}
|
||
recent_dialogue = ""
|
||
if conversation_id:
|
||
hw = await get_history_with_window(
|
||
conversation_id,
|
||
max_pairs=settings.chat_history_max_pairs,
|
||
max_chars=settings.chat_history_max_chars,
|
||
)
|
||
recent = hw.window[-4:] if len(hw.window) > 4 else hw.window
|
||
parts = []
|
||
for msg in recent:
|
||
if isinstance(msg, HumanMessage):
|
||
parts.append(f"用户: {msg.content}")
|
||
elif isinstance(msg, AIMessage):
|
||
parts.append(f"助手: {msg.content}")
|
||
recent_dialogue = "\n".join(parts) if parts else ""
|
||
try:
|
||
prompt = get_profile_extraction_prompt(
|
||
user_message, missing_fields, recent_dialogue=recent_dialogue or None
|
||
)
|
||
content = await ainvoke_json_object(
|
||
self.llm,
|
||
prompt,
|
||
max_tokens=512,
|
||
agent="ProfileAgent.extract_profile_from_message",
|
||
)
|
||
parsed = json.loads(extract_json_payload(content))
|
||
result = {}
|
||
if "birth_year" in parsed and parsed["birth_year"] is not None:
|
||
raw = parsed["birth_year"]
|
||
if isinstance(raw, int) and 1900 <= raw <= 2100:
|
||
result["birth_year"] = raw
|
||
elif isinstance(raw, str) and raw.isdigit():
|
||
y = int(raw)
|
||
if y < 100:
|
||
y = 1900 + y if y >= 50 else 2000 + y
|
||
if 1900 <= y <= 2100:
|
||
result["birth_year"] = y
|
||
if "birth_place" in parsed and parsed["birth_place"]:
|
||
result["birth_place"] = str(parsed["birth_place"])
|
||
if "grew_up_place" in parsed and parsed["grew_up_place"]:
|
||
result["grew_up_place"] = str(parsed["grew_up_place"])
|
||
if "occupation" in parsed and parsed["occupation"]:
|
||
result["occupation"] = str(parsed["occupation"])
|
||
return result
|
||
except (json.JSONDecodeError, Exception) as e:
|
||
logger.error("提取资料信息失败: {}", e)
|
||
return {}
|
||
|
||
async def generate_profile_followup(
|
||
self,
|
||
conversation_id: str,
|
||
user_message: str,
|
||
missing_fields: List[str],
|
||
filled_fields: Dict[str, str],
|
||
nickname: str = "",
|
||
interview_stage_hint: str = "",
|
||
) -> List[str]:
|
||
"""生成资料追问回复，不持久化（由 Orchestrator 负责）"""
|
||
if not self.llm:
|
||
return ["谢谢！还能告诉我更多吗？"]
|
||
try:
|
||
prompt = get_profile_followup_prompt(
|
||
missing_fields,
|
||
filled_fields,
|
||
user_message,
|
||
nickname,
|
||
interview_stage_hint=interview_stage_hint,
|
||
)
|
||
hw = await get_history_with_window(
|
||
conversation_id,
|
||
max_pairs=settings.chat_history_max_pairs,
|
||
max_chars=settings.chat_history_max_chars,
|
||
)
|
||
messages: List[Any] = [SystemMessage(content=prompt)]
|
||
messages.extend(hw.window)
|
||
messages.append(HumanMessage(content=user_message))
|
||
log_agent_payload(
|
||
logger,
|
||
"ProfileAgent.followup.prompt",
|
||
format_history_string(messages),
|
||
)
|
||
chat_llm = self.llm.bind(
|
||
max_tokens=settings.chat_profile_followup_max_tokens
|
||
)
|
||
llm_t0 = time.perf_counter()
|
||
with agent_span(
|
||
logger,
|
||
"ProfileAgent.followup.llm",
|
||
conversation_id=conversation_id,
|
||
):
|
||
logger.info(
|
||
"event=chat_prompt_built agent=ProfileAgent.generate_profile_followup "
|
||
"prompt_chars={} history_pairs_total={} history_pairs_windowed={}",
|
||
_message_contents_char_count(messages),
|
||
hw.turn_total,
|
||
len(hw.window) // 2,
|
||
)
|
||
response = await chat_llm.ainvoke(messages)
|
||
logger.info(
|
||
"event=chat_llm_done agent=ProfileAgent.generate_profile_followup "
|
||
"response_latency_ms={:.2f}",
|
||
(time.perf_counter() - llm_t0) * 1000,
|
||
)
|
||
response_text = (
|
||
response.content if hasattr(response, "content") else str(response)
|
||
)
|
||
log_agent_payload(
|
||
logger, "ProfileAgent.followup.raw_response", response_text
|
||
)
|
||
raw_list = segments_from_llm_response(response_text, max_segments=3)
|
||
if not raw_list:
|
||
raw_list = [response_text.strip()]
|
||
out = truncate_chat_segments(
|
||
raw_list,
|
||
max_segments=3,
|
||
max_chars_per_segment=settings.chat_interview_max_chars_per_segment,
|
||
)
|
||
log_agent_summary(
|
||
logger,
|
||
"ProfileAgent.followup segments={} conversation_id={}",
|
||
len(out),
|
||
conversation_id,
|
||
)
|
||
segments = (
|
||
out
|
||
if out
|
||
else [
|
||
response_text.strip()[
|
||
: settings.chat_interview_max_chars_per_segment
|
||
]
|
||
]
|
||
)
|
||
return nonempty_segments_or_fallback(
|
||
segments,
|
||
fallback="谢谢分享！能再告诉我一些吗？",
|
||
)
|
||
except Exception as e:
|
||
logger.error("生成资料跟进回复失败: {}", e)
|
||
return ["谢谢分享！能再告诉我一些吗？"]
|
||
|
||
async def generate_profile_greeting(
|
||
self,
|
||
conversation_id: str,
|
||
missing_fields: List[str],
|
||
nickname: str = "",
|
||
) -> List[str]:
|
||
"""生成资料收集开场白，不持久化（由 Orchestrator 负责）"""
|
||
if not self.llm:
|
||
return ["你好！在开始之前，能告诉我你是哪一年出生的吗？"]
|
||
try:
|
||
prompt = get_profile_greeting_prompt(missing_fields, nickname)
|
||
hw = await get_history_with_window(
|
||
conversation_id,
|
||
max_pairs=settings.chat_history_max_pairs,
|
||
max_chars=settings.chat_history_max_chars,
|
||
)
|
||
messages: List[Any] = [SystemMessage(content=prompt)]
|
||
messages.extend(hw.window)
|
||
if hw.window:
|
||
messages.append(
|
||
HumanMessage(content="（请根据上文自然接话，继续资料收集开场。）")
|
||
)
|
||
else:
|
||
messages.append(HumanMessage(content="（请说出资料收集开场白。）"))
|
||
log_agent_payload(
|
||
logger, "ProfileAgent.greeting.prompt", format_history_string(messages)
|
||
)
|
||
chat_llm = self.llm.bind(
|
||
max_tokens=settings.chat_profile_followup_max_tokens
|
||
)
|
||
llm_t0 = time.perf_counter()
|
||
with agent_span(
|
||
logger,
|
||
"ProfileAgent.greeting.llm",
|
||
conversation_id=conversation_id,
|
||
):
|
||
logger.info(
|
||
"event=chat_prompt_built agent=ProfileAgent.generate_profile_greeting "
|
||
"prompt_chars={} history_pairs_total={} history_pairs_windowed={}",
|
||
_message_contents_char_count(messages),
|
||
hw.turn_total,
|
||
len(hw.window) // 2,
|
||
)
|
||
response = await chat_llm.ainvoke(messages)
|
||
logger.info(
|
||
"event=chat_llm_done agent=ProfileAgent.generate_profile_greeting "
|
||
"response_latency_ms={:.2f}",
|
||
(time.perf_counter() - llm_t0) * 1000,
|
||
)
|
||
response_text = (
|
||
response.content if hasattr(response, "content") else str(response)
|
||
)
|
||
log_agent_payload(
|
||
logger, "ProfileAgent.greeting.raw_response", response_text
|
||
)
|
||
raw_list = segments_from_llm_response(response_text, max_segments=2)
|
||
if not raw_list:
|
||
raw_list = [response_text.strip()]
|
||
out = truncate_chat_segments(
|
||
raw_list,
|
||
max_segments=2,
|
||
max_chars_per_segment=settings.chat_interview_max_chars_per_segment,
|
||
)
|
||
log_agent_summary(
|
||
logger,
|
||
"ProfileAgent.greeting segments={} conversation_id={}",
|
||
len(out),
|
||
conversation_id,
|
||
)
|
||
segments = (
|
||
out
|
||
if out
|
||
else [
|
||
response_text.strip()[
|
||
: settings.chat_interview_max_chars_per_segment
|
||
]
|
||
]
|
||
)
|
||
return nonempty_segments_or_fallback(
|
||
segments,
|
||
fallback="你好！在开始之前，能告诉我你是哪一年出生的吗？",
|
||
)
|
||
except Exception as e:
|
||
logger.error("生成资料收集开场白失败: {}", e)
|
||
return [
|
||
"你好！在我们开始聊人生故事之前，能先简单介绍一下你自己吗？比如你是哪一年出生的？"
|
||
]
|