You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
alpha_tools/rpc_alpha_workflow/reference_02.py

630 lines
23 KiB

# -*- coding: utf-8 -*-
import random
import sys
import uuid
import httpx
from odoo import models, fields
from . import decode_template
class AlphaIdea(models.Model):
_name = 'alpha.idea'
_description = 'Alpha Idea'
_order = 'id desc'
name = fields.Char(string='Name', required=True, default=lambda self: str(uuid.uuid4()))
status = fields.Selection([
('draft', 'Draft'),
('fetched_data', 'Fetched Data'),
('generated_prompt', 'Generated Prompt'),
('posted_to_ms', 'Posted To MS'),
('llm_received', 'LLM Received'),
('decode_template', 'Decode Template'),
('decoded', 'Decoded'),
('done', 'Done'),
('failed', 'Failed'),
('cancel', 'Cancel')
], string='Status', default='draft')
pushed = fields.Boolean(string='Pushed', default=False, readonly=True)
region = fields.Many2one('alpha.region.settings', string='Region', required=True)
universe = fields.Many2one('alpha.universe.settings', string='Universe', required=True)
data_type = fields.Selection([('MATRIX', 'MATRIX'), ('VECTOR', 'VECTOR')], string='Data Type', required=True, default='MATRIX')
delay = fields.Selection([('1', '1'), ('0', '0')], required=True, string='Delay', default='1')
data_sets = fields.Char(string='Data Sets', compute='_compute_data_sets')
meta_prompt = fields.Many2one('alpha.prompt.settings', string='Meta Prompt')
replace_prompt = fields.Text(string='Replace Prompt')
final_prompt = fields.Text(string='Final Prompt')
system_prompt = fields.Text(string='System Prompt')
user_prompt = fields.Text(string='User Prompt')
llm_generated_idea = fields.Text(string='LLM Generated Idea')
result_message = fields.Text(string='Result Message')
llm_settings_line_id = fields.Many2one('llm.settings.line', string='Model', default=lambda self: self._default_llm_settings_line_id())
def _default_llm_settings_line_id(self):
"""默认随机选择一个模型,如果没有则返回 False"""
llm_lines = self.env['llm.settings.line'].search([])
if llm_lines:
return random.choice(llm_lines).id
return False
idea_template_ids = fields.One2many('alpha.idea.template', 'idea_id', string='Idea Templates')
needed_data_set_ids = fields.One2many('alpha.needed.data.set', 'idea_id', string='Needed Data Sets')
final_expression_ids = fields.One2many('alpha.final.expression', 'idea_id', string='Final Expressions')
expression_count = fields.Integer(string='Expression Count', required=True, readonly=True, default=0, compute='_compute_expression_count')
def _compute_data_sets(self):
# 显示使用的数据集
for record in self:
record.data_sets = ''
if record.needed_data_set_ids:
data_sets_list = []
for data_set_name in record.needed_data_set_ids:
data_sets_list.append(data_set_name.name)
if len(data_sets_list) > 0:
record.data_sets = ', '.join(data_sets_list)
def action_cancel(self):
self.status = 'cancel'
def action_reset(self):
self.status = 'fetched_data'
def btn_check_and_fetch_data(self):
if not self.needed_data_set_ids:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'No data',
'message': 'No data set is needed.',
'type': 'danger',
'sticky': False,
}
}
# TODO: 通过 datasets_id, 查找 datasets 中, 是否存在 datasets_id 的数据, 如果不存在, 则在 datasets 模块中创建一条记录, 然后需要过去手动下载(必须)
success = False
for dataset in self.needed_data_set_ids:
dataset_id = self.env['alpha.datasets'].search(
[('datasets_id', '=', dataset.name),
('region', '=', self.region.id),
('universe', '=', self.universe.id),
('delay', '=', self.delay)
], limit=1)
if not dataset_id:
dataset_id = self.env['alpha.datasets'].create({
'name': str(uuid.uuid4()),
'datasets_id': dataset.name,
'region': self.region.id,
'universe': self.universe.id,
'delay': self.delay,
})
# 创建 dataset_id 的记录之后, 执行一下 dataset_id 的 btn_get_datasets 方法
try:
dataset_id.btn_get_datasets()
success = True
except Exception as e:
self.status = 'failed'
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'Failed',
'message': f'Dataset fetch failed: {str(e)}',
'type': 'danger',
'sticky': False,
}
}
else:
if dataset_id.line_ids:
success = True
else:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'No data',
'message': 'Dataset does not exist.',
'type': 'danger',
'sticky': False,
}
}
if success:
self.status = 'fetched_data'
return True
def btn_generate_final_prompt(self):
if self.final_prompt:
self.final_prompt = ''
self.system_prompt = ''
self.user_prompt = ''
self.status = 'fetched_data'
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': '成功',
'message': '已清除最终提示词',
'type': 'success',
'sticky': False,
}
}
if not self.meta_prompt:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'No data!',
'message': 'Please select the prompt template first.',
'type': 'danger',
'sticky': False,
}
}
if not self.replace_prompt:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'No data!',
'message': 'Research direction cannot be empty.',
'type': 'danger',
'sticky': False,
}
}
system_prompt = '''You are executing two skills in sequence:
1) brain-data-feature-engineering
2) brain-feature-implementation
The following SKILL.md documents are authoritative; follow them exactly.
'''
data_sets_list = []
dataset_id = ''
category = ''
region = self.region.name
delay = self.delay
universe = self.universe.name
for dataset in self.needed_data_set_ids:
datasets_id = self.env['alpha.datasets'].search(
[('datasets_id', '=', dataset.name)], limit=1)
for datasets_line in datasets_id.line_ids:
data_sets_list.append({
'id': datasets_line.data_field_name,
'description': datasets_line.description
})
if not dataset_id:
dataset_id = datasets_line.dataset_id
if not category:
category = datasets_line.category_name
if not data_sets_list:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'No data!',
'message': 'No related dataset found.',
'type': 'danger',
'sticky': False,
}
}
field_count = len(data_sets_list)
operator_ids = self.env['alpha.operator.line'].search([])
operator_list = []
for operator in operator_ids:
operator_list.append({
'name': operator.name,
'category': operator.category,
'scope': operator.scope,
'description': operator.description,
'definition': operator.definition
})
if not operator_list:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'No data!',
'message': 'No related Operator found.',
'type': 'danger',
'sticky': False,
}
}
fields_json = ','.join([str(field) for field in data_sets_list])
user_prompt = '''{
"instructions": {
"output_format": "Fill OUTPUT_TEMPLATE.md with concrete content.",
"implementation_examples": "Each Implementation Example must be a template with {variable} placeholders. Use only placeholders from allowed_placeholders. Use suffix-only names; do not include dataset code/prefix/horizon.",
"no_code_fences": true,
"do_not_invent_placeholders": true
},
"dataset_context": {
"dataset_id": "''' + dataset_id + '''",
"dataset_name": null,
"dataset_description": null,
"category": "''' + category + '''",
"region": "''' + region + '''",
"delay": ''' + str(delay) + ''',
"universe": "''' + universe + '''",
"field_count": ''' + str(field_count) + '''
},
"fields": [
''' + fields_json + '''
]
}
'''
final_prompt = self.meta_prompt.prompt
final_prompt = final_prompt.replace(
'###question_driven###', self.replace_prompt)
final_prompt = final_prompt.replace('###dataset_id###', dataset_id)
final_prompt = final_prompt.replace('###category###', category)
final_prompt = final_prompt.replace('###region###', region)
final_prompt = final_prompt.replace('###delay###', str(delay))
final_prompt = final_prompt.replace(
'###field_count###', str(field_count))
final_prompt = final_prompt.replace('###universe###', universe)
final_prompt = final_prompt.replace(
'###datasets###', str(data_sets_list))
final_prompt = final_prompt.replace(
'###operators###', str(operator_list))
if self.data_type and self.data_type.upper() == "VECTOR":
vector_prompt = "since all the following the data is vector type data, before you do any process, you should choose a vector operator to generate its statistical feature to use, the data cannot be directly use. for example, if datafieldA and datafieldB are vector type data, you can use vec_avg(datafieldA) - vec_avg(datafieldB), where vec_avg() operator is used to generate the average of the data on a certain date. similarly, vector type operator can only be used on the vector type operator directly and cannot be nested, for example vec_avg(vec_sum(datafield)) is a false use."
final_prompt = final_prompt.replace(
'###vector_instruction###', vector_prompt)
else:
final_prompt = final_prompt.replace('###vector_instruction###', '')
self.system_prompt = system_prompt
self.user_prompt = user_prompt
self.final_prompt = final_prompt
self.status = 'generated_prompt'
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': '成功',
'message': f'已生成提示模板,包含 {field_count} 个数据字段',
'type': 'success',
'sticky': False,
}
}
def post_to_ms(self):
if not all([self.final_prompt, self.user_prompt, self.system_prompt]):
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'Error',
'message': 'Please generate a prompt template.',
'type': 'danger',
'sticky': False,
}
}
if not self.llm_settings_line_id:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'Error',
'message': 'Please select a model.',
'type': 'danger',
'sticky': False,
}
}
model_name = self.llm_settings_line_id.model_name
base_url = self.llm_settings_line_id.llm_setting_id.base_url
api_key = self.llm_settings_line_id.llm_setting_id.api_key
# 获取当前 record_id
record_id = self.id
# 获取 Odoo 回调地址
base_url_odoo = self.env['ir.config_parameter'].sudo().get_param('web.base.url')
callback_url = f"{base_url_odoo}/api/alpha-idea/result"
# 组装完整 prompt
full_prompt = f"{self.system_prompt}\n\n{self.user_prompt}\n\n{self.final_prompt}"
# 获取微服务配置
ms_config = self.get_ms_config()
ms_url = ms_config.get('url', '')
if not ms_url:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'Error',
'message': 'Microservice address not configured.',
'type': 'danger',
'sticky': False
}
}
# 组装请求数据
payload = {
'record_id': record_id,
'prompt': full_prompt,
'model_name': model_name,
'base_url': base_url,
'api_key': api_key,
'callback_url': callback_url,
'system_prompt': self.system_prompt,
'user_prompt': self.user_prompt,
'final_prompt': self.final_prompt,
}
# 发送请求到微服务
try:
httpx.post(f"{ms_url}:32004/api_alpha_generate_idea",
json=payload, timeout=0.001)
except httpx.TimeoutException:
pass
except Exception as e:
self.status = 'failed'
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'Error',
'message': f'Failed to send microservice:\n{e}.',
'type': 'danger',
'sticky': False,
}
}
self.status = 'posted_to_ms'
def get_ms_config(self):
# TODO 先从 nacos 获取微服务 url
nacos_url = ''
platform_info = sys.platform
if platform_info == "darwin":
nacos_url = 'http://192.168.31.41:30848/nacos/v1/cs/configs?dataId=microservices_dev&group=quantify'
elif platform_info.startswith("linux"):
nacos_url = 'http://192.168.31.41:30848/nacos/v1/cs/configs?dataId=microservices&group=quantify'
try:
ms_config_resp = httpx.get(nacos_url)
ms_config_resp.raise_for_status()
except Exception as e:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'Error',
'message': f'Nacos request failed:\n{e}.',
'type': 'danger',
'sticky': False,
}
}
ms_config = ms_config_resp.json()
return ms_config
def action_set_llm_received(self):
"""手动设置 LLM 已返回状态
当手动录入 llm_generated_idea 后,调用此函数将状态改为 llm_received
"""
if self.llm_generated_idea and self.status == 'generated_prompt':
self.status = 'llm_received'
return True
else:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'Error',
'message': 'llm_generated_idea 无数据或状态不是 Generated Prompt',
'type': 'danger',
'sticky': False,
}
}
def decode_template(self):
if not self.llm_generated_idea or self.status != 'llm_received':
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'Please use llm to generate the template first.',
'message': 'llm_generated_idea no data or status != llm_received',
'type': 'danger',
'sticky': False,
}
}
if self.status == 'done':
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'message',
'message': 'This idea has been generated, please do not repeat the operation.',
'type': 'danger',
'sticky': False,
}
}
llm_template = self.llm_generated_idea
data_sets_list = []
for dataset in self.needed_data_set_ids:
datasets_id = self.env['alpha.datasets'].search(
[('datasets_id', '=', dataset.name)], limit=1)
for datasets_line in datasets_id.line_ids:
data_sets_list.append({
'id': datasets_line.data_field_name
})
result_data = decode_template.process(data_sets_list, llm_template)
if result_data['success']:
templates = result_data['templates']
expressions = result_data['expressions']
# 如果没有解出数据,不做任何操作
if len(templates) == 0 and len(expressions) == 0:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': '提示',
'message': '未解码出任何数据',
'type': 'warning',
'sticky': False,
}
}
# 组装模板数据
template_data = []
for template_item in templates:
template_data.append({
'template': template_item['template'],
'original_template': template_item['original_template'],
'idea': template_item.get('idea', ''),
'template_line_ids': [(0, 0, {'expression': line}) for line in template_item['expressions']]
})
# 保存数据
self.write({
'idea_template_ids': [(0, 0, data) for data in template_data],
'final_expression_ids': [(0, 0, {'name': expression}) for expression in expressions],
'status': 'decoded',
})
return True
else:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': '失败',
'message': '模板解码失败, 生成数量为 0',
'type': 'danger',
'sticky': False,
}
}
def combination_settings(self):
self.ensure_one()
if self.idea_template_ids and self.final_expression_ids:
return {
'type': 'ir.actions.act_window',
'name': '组合设置',
'res_model': 'wizard.combination.settings',
'view_mode': 'form',
'target': 'new',
'context': {
'active_model': 'alpha.idea',
'active_id': self.id,
},
}
else:
return {
'type': 'ir.actions.client',
'tag': 'display_notification',
'params': {
'title': 'Execute the decoding template first',
'message': 'idea_template_ids and final_expression_ids no data!',
'type': 'danger',
'sticky': False,
}
}
def _compute_expression_count(self):
for record in self:
if record.final_expression_ids:
record.expression_count = len(record.final_expression_ids)
else:
record.expression_count = 0
class NeededDataSet(models.Model):
_name = 'alpha.needed.data.set'
_description = 'Alpha Needed Data Set'
idea_id = fields.Many2one('alpha.idea', string='Idea', ondelete='cascade')
name = fields.Char(string='Name')
class IdeaTemplate(models.Model):
_name = 'alpha.idea.template'
_description = 'Alpha Idea Template'
idea_id = fields.Many2one('alpha.idea', string='Idea', ondelete='cascade')
template_line_ids = fields.One2many(
'alpha.idea.template.line', 'idea_id', string='Template Lines')
template = fields.Char(string='Template')
original_template = fields.Char(string='Original Template')
expression_count = fields.Integer(
string='Expression Count', compute='_compute_expression_count', default=0)
idea = fields.Text(string='Text')
def _compute_expression_count(self):
for record in self:
record.expression_count = len(record.template_line_ids)
class IdeaTemplateLine(models.Model):
_name = 'alpha.idea.template.line'
_description = 'Alpha Idea Template Line'
idea_id = fields.Many2one('alpha.idea.template',
string='Alpha Idea Template', ondelete='cascade')
expression = fields.Char(string='Expression')
class Final_Expression(models.Model):
_name = 'alpha.final.expression'
_description = 'Alpha Final Expression'
idea_id = fields.Many2one('alpha.idea', string='Idea', ondelete='cascade')
name = fields.Char(string='Name')