Mandarin in the AI ​​world, using prompt word engineering to create a "Professional Legal Consultant" application - Enterprise Edition

import osfrom config.load_key import load_key # Load API Keyload_key()print(f'''The API Key you configured is: {os.environ["DASHSCOPE_API_KEY"][:5]+"*"*5}''')

from openai import OpenAIimport osclient = OpenAI( base_url = "https://api.siliconflow.cn/v1", api_key = os.getenv("DASHSCOPE_API_KEY"),)prompt = '''In the civil enforcement procedure, the second auction of the house under the name of the person subject to enforcement still failed, and the applicant purchased it at the failed auction price. After the People's Court ruled that the house belonged to the applicant, it took many years for the People's Court to deliver it. After the enforcement procedure was terminated in accordance with the law, the applicant claimed the loss from the date the ruling took effect to the actual delivery date. Should the People's Court accept it? Should the person subject to enforcement bear liability for compensation for the loss? '''completion = client.chat.completions.create( model="deepseek-ai/DeepSeek-R1", messages=[{"role":"user","content":prompt}], temperature=0.6, top_p=0.7, max_tokens=4096, stream = True # Whether to stream output) for chunk in completion: if chunk.choices[0].delta.content is not None: print(chunk.choices[0].delta.content, end="")

from openai import OpenAIimport osclient = OpenAI( base_url = "https://api.siliconflow.cn/v1", api_key = os.getenv("DASHSCOPE_API_KEY"),)system_prompt = '''# Role You are a senior lawyer. Your goal is to think about the legal questions given by the user step by step, give in-depth answers, and propose the most granular claims, legal basis, factual basis, and executable steps.## Working steps 1. According to the case background and questions given by the user, conduct the first analysis, list the possible claims, legal basis, factual basis, and the next executable actions for the user. If the user does not provide enough factual basis, the factual basis of past similar cases can be provided as a reference. 2. For each step in the first step, further disassemble it as a sub-claim, and give the legal basis, factual basis, and the next executable action. If the user does not provide enough factual basis, the factual basis of past similar cases can be provided as a reference. 3. Repeat the second step until it cannot be further decomposed or it is specific and executable. Through such a conversation, you can make an overall summary and give the customer an overall conclusion according to the specific operation steps. '''def get_completion(prompt): completion = client.chat.completions.create( model="deepseek-ai/DeepSeek-R1", messages=[ {"role":"system", "content":system_prompt}, {"role":"user","content":prompt} ], temperature=0.6, top_p=0.7, max_tokens=4096, stream = True # Whether to stream output ) for chunk in completion: if chunk.choices[0].delta.content is not None: yield chunk.choices[0].delta.contentuser_prompt = '''In the civil enforcement procedure, the second auction of the house under the name of the person subject to enforcement still failed, and the applicant purchased it at the failed auction price. After the People's Court ruled that the house belonged to the applicant, it took many years for the People's Court to deliver it. After the execution procedure is terminated according to law, if the applicant claims losses from the date the ruling takes effect to the date of actual delivery, should the people's court accept it? Should the person subject to execution bear liability for compensation for the losses? '''for result in get_completion(user_prompt): print(result, end="")

from openai import OpenAIimport osclient = OpenAI( base_url = "https://api.siliconflow.cn/v1", api_key = os.getenv("DASHSCOPE_API_KEY"),)system_prompt = '''# Role You are a senior lawyer. Your goal is to think about the legal questions given by the user step by step, give in-depth answers, and propose the most granular claims, legal basis, factual basis, and executable steps.## Working steps 1. According to the case background and questions given by the user, conduct the first analysis, list the possible claims, legal basis, factual basis, and the next executable actions for the user. If the user does not provide enough factual basis, the factual basis of past similar cases can be provided as a reference. 2. For each step in the first step, further disassemble it as a sub-claim, and give the legal basis, factual basis, and the next executable action. If the user does not provide enough factual basis, the factual basis of past similar cases can be provided as a reference. 3. Repeat the second step until it cannot be further decomposed or it is specific and executable. Through such a conversation, you can make an overall summary and give the customer an overall conclusion based on the specific operation steps. '''session = [ { "role": "system", "content": system_prompt }]def get_completion(prompt): session.append({"role":"user","content":prompt}) completion = client.chat.completions.create( model="deepseek-ai/DeepSeek-R1", messages=session, temperature=0.6, top_p=0.7, max_tokens=4096, stream = True # Whether to stream output ) for chunk in completion: if chunk.choices[0].delta.content is not None: yield chunk.choices[0].delta.content # Add historical information to session session.append({"role":"assistant","content":chunk.choices[0].delta.content})print("-----------First round of conversation-----------")user_prompt1 = '''In the civil enforcement procedure, the second auction of the house under the name of the person subject to enforcement still failed, and the applicant purchased it at the failed auction price. After the People's Court ruled that the house belonged to the applicant, it was not delivered by the People's Court until many years later. After the enforcement procedure was terminated according to law, the applicant claimed the loss from the date the ruling took effect to the actual delivery date. Should the People's Court accept it? Should the person subject to enforcement bear the liability for compensation for the loss? '''for result in get_completion(user_prompt1): print(result, end="")print("-----------Second round of dialogue-----------")user_prompt2 = '''Please summarize your conclusion based on the above dialogue.'''for result in get_completion(user_prompt2): print(result, end="")print("-----------Third round of dialogue-----------")user_prompt3 = '''Please output the final result in markdown format.'''for result in get_completion(user_prompt2): print(result, end="")