2025计算机设计大赛
A product proposal for an AI contract-review assistant aimed at flexible workers, developed by the TalentLink team for an AIGC innovation challenge.
19p
competition proposal
k=3
retrieved passages
AIGC
innovation challenge
01 / Overview
LawGuard AI is the product concept and prototype developed by the TalentLink team for an AIGC innovation challenge. It reviews employment contracts for flexible workers, retrieves relevant legal context and returns risks, legal reasoning and suggested edits.
Objectives
Product mechanics
The system is explained through its modules and the behaviour they enable.
Pasted or uploaded agreement text.
Labour-law and contract-reference documents.
Vector retrieval over relevant legal passages.
Prompt combining question and retrieved context.
Risks, legal basis, revisions and disclaimer.
02 / Package highlights
03 / Mechanism
The prototype embeds the legal corpus and retrieves the three most relevant passages for each contract question. The retrieved context is inserted into a structured prompt that asks for risk identification, legal basis and concrete edits.
The prompt separates three review layers: whole-contract completeness, party capability and substantive fairness. It also requires a disclaimer and instructs the model to ignore retrieved context when it is irrelevant to the question.
The competition proposal adds market analysis and product positioning around flexible work. The technical prototype focuses on contract review and risk explanation, not legal representation or final adjudication.
04 / Key code
Only verified source excerpts and pseudocode taken from the project papers are shown.
Original Python excerpt from the LawGuard prototype.
from langchain_community.vectorstores import FAISS
from langchain.chains import RetrievalQA
from langchain_core.prompts import PromptTemplate
embeddings = CustomVivoEmbeddings()
vectorstore = FAISS.from_documents(documents, embeddings)
retriever = vectorstore.as_retriever(
search_kwargs={"k": 3}
)
prompt = PromptTemplate(
template=legal_review_template,
input_variables=["context", "question"],
)
qa_chain = RetrievalQA.from_chain_type(
llm=CustomVivoLLM(),
chain_type="stuff",
retriever=retriever,
chain_type_kwargs={"prompt": prompt},
)Original prompt excerpt defining the required analysis format.
Analyse the contract using the retrieved context.
1. Identify potential legal risks.
2. State the relevant legal basis.
3. Propose concrete contract revisions.
4. Provide risk-control suggestions.
Return the result in this structure:
### Identified risks
- ...
### Legal basis and analysis
- ...
### Suggested revisions
- ...
If the question is outside labour law, contract or
flexible-work topics, ignore the retrieved context
and answer normally.Delivery notes
Limitations / next steps
05 / Downloads
Open the original report, presentation or source archive. PDF cover pages with private details were removed.
19-page competition proposal for LawGuard AI.
Next folder
Credentials