Dockerfile and other imp files

Dockerfile

@@ -0,0 +1,17 @@
+# Use Python 3.10.6 image
+FROM python:3.10
+
+# Set the working directory in the container
+WORKDIR /app
+
+# Copy the requirements file
+COPY requirements.txt .
+
+# Install dependencies
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy the rest of your application files
+COPY . .
+
+# Command to run your FastAPI app
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

api_docs.md

@@ -0,0 +1,168 @@
+
+---
+
+# API Documentation for Legal Assistance
+
+## Overview
+This API provides legal assistance by allowing users to upload PDF documents and submit queries for various legal services, including legal advisory, report generation, and case outcome prediction. 
+
+## Base URL
+```
+https://navilaw-ai.onrender.com/legal-assistance/
+```
+
+## HTTP Method
+`POST`
+
+## Endpoint
+`/legal-assistance/`
+
+## Request Parameters
+
+### Form Parameters
+
+| Parameter | Type        | Required | Description |
+|-----------|-------------|----------|-------------|
+| `query`   | `string`    | Yes      | A string representing the legal query the user wishes to ask. |
+| `option`  | `string`    | Yes      | A string indicating the type of legal assistance required. Possible values are "Legal Advisory", "Legal Report Generation", and "Case Outcome Prediction". |
+| `files`   | `List[UploadFile]` | Yes      | A list of PDF files containing legal documents to be analyzed. |
+
+### Example Request
+```http
+POST /legal-assistance/ HTTP/1.1
+Host: https://navilaw-ai.onrender.com/
+Content-Type: multipart/form-data
+
+--boundary
+Content-Disposition: form-data; name="query"
+
+What are the possible outcomes of my case?
+--boundary
+Content-Disposition: form-data; name="option"
+
+Case Outcome Prediction
+--boundary
+Content-Disposition: form-data; name="files"; filename="legal_case.pdf"
+Content-Type: application/pdf
+
+<PDF file content>
+--boundary--
+```
+
+## Response Format
+
+### Successful Response
+- **Status Code:** `200 OK`
+- **Content-Type:** `application/json`
+
+#### Response Body
+- **For Legal Advisory:**
+```json
+{
+    "result": "Based on the provided documents and the legal query, here are the considerations to keep in mind regarding your case..."
+}
+```
+
+- **For Legal Report Generation:**
+```json
+{
+    "report": "Legal Report:\n\n1. Introduction\n2. Case Details\n3. Analysis\n4. Conclusion"
+}
+```
+
+- **For Case Outcome Prediction:**
+```json
+{
+    "prediction": "Based on the analysis of the legal precedents and the court's decision in a similar case, there is a 70% chance of a favorable outcome for Fast Retail in their lawsuit against Tech Solutions. The court ruled in favor of Fast Retail, ordering Tech Solutions to pay damages of ₹60,00,000. The uncertainty lies in the court's consideration of not all claimed damages were directly attributable to Tech Solutions' breach. It's crucial to consider this when predicting the outcome."
+}
+```
+
+### Error Response
+- **Status Code:** `400 Bad Request`
+- **Content-Type:** `application/json`
+
+#### Response Body
+```json
+{
+    "detail": "Please upload at least one PDF file."
+}
+```
+
+#### Possible Error Messages
+- **If no files are uploaded:**
+```json
+{
+    "detail": "Please upload at least one PDF file."
+}
+```
+
+- **If no query is provided:**
+```json
+{
+    "detail": "Please enter a query."
+}
+```
+
+- **If an invalid option is selected:**
+```json
+{
+    "detail": "Invalid option selected."
+}
+```
+
+## Sample Inputs and Outputs
+
+### 1. Legal Advisory
+#### Request
+```http
+POST /legal-assistance/ 
+```
+With the following form data:
+- **query:** "What are the implications of the new law on my case?"
+- **option:** "Legal Advisory"
+- **files:** (Upload PDF: `law_document.pdf`)
+
+#### Response
+```json
+{
+    "result": "The new law may affect your case in several ways, particularly regarding..."
+}
+```
+
+### 2. Legal Report Generation
+#### Request
+```http
+POST /legal-assistance/ 
+```
+With the following form data:
+- **query:** "Generate a report on the recent legal changes."
+- **option:** "Legal Report Generation"
+- **files:** (Upload PDF: `legal_changes.pdf`)
+
+#### Response
+```json
+{
+    "report": "Legal Report:\n\n1. Introduction\n2. Summary of Changes\n3. Implications\n4. Conclusion"
+}
+```
+
+### 3. Case Outcome Prediction
+#### Request
+```http
+POST /legal-assistance/ 
+```
+With the following form data:
+- **query:** "What is the likelihood of winning my case based on previous rulings?"
+- **option:** "Case Outcome Prediction"
+- **files:** (Upload PDF: `previous_rulings.pdf`)
+
+#### Response
+```json
+{
+    "prediction": "Based on the analysis of the legal precedents and the court's decision in a similar case, there is a 70% chance of a favorable outcome for Fast Retail in their lawsuit against Tech Solutions..."
+}
+```
+
+---
+
+

main.py

@@ -0,0 +1,143 @@
+import os
+from fastapi import FastAPI, UploadFile, File, Form, HTTPException
+from fastapi.middleware.cors import CORSMiddleware
+from langchain_groq import ChatGroq
+from PyPDF2 import PdfReader
+from langchain.docstore.document import Document
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.runnables import RunnablePassthrough, RunnableParallel
+from langgraph.prebuilt import create_react_agent
+from retrieval import create_retriever
+from templates import advisor_template, predictor_template, generator_template
+from langchain.tools.retriever import create_retriever_tool
+from tools import tavily_tool
+from dotenv import load_dotenv
+from typing import List
+
+load_dotenv()
+groq_api_key = os.getenv("GROQ_API_KEY")
+chat = ChatGroq(model = "llama-3.3-70b-versatile", api_key=groq_api_key)
+app = FastAPI()
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.get("/")
+async def read_root():
+    return {"message": "Welcome to the Legal Research API! Please use one of the endpoints for requests."}
+
+def process_files(files: List[UploadFile]):
+    if not files:
+        raise HTTPException(status_code=400, detail="Please upload at least one PDF file.")
+    docs = []
+    for uploaded_file in files:
+        reader = PdfReader(uploaded_file.file)
+        text = ""
+        for page in reader.pages:
+            text += page.extract_text()
+        docs.append(Document(page_content=text, metadata={"source": uploaded_file.filename}))
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
+    pdf_content = text_splitter.split_documents(docs)
+    return pdf_content
+
+def setup_retriever(pdf_content):
+    retriever = create_retriever(pdf_content)
+    retrieval_tool = create_retriever_tool(
+        retriever,
+        "Pdf_content_retriever",
+        "Searches and returns excerpts from the set of PDF docs.",
+    )
+    return retriever, retrieval_tool
+
+def setup_agents(tools):
+    advisor_graph = create_react_agent(chat, tools=tools, state_modifier=advisor_template)
+    predictor_graph = create_react_agent(chat, tools=tools, state_modifier=predictor_template)
+    return advisor_graph, predictor_graph
+
+@app.post("/legal-assistance/")
+async def legal_assistance(
+    query: str = Form(...),
+    option: str = Form(...),
+    files: List[UploadFile] = File(...)
+):
+    if not query:
+        raise HTTPException(status_code=400, detail="Please enter a query.")
+    pdf_content = process_files(files)
+    retriever, retrieval_tool = setup_retriever(pdf_content)
+    tools = [tavily_tool, retrieval_tool]
+    advisor_graph, predictor_graph = setup_agents(tools)
+    inputs = {"messages": [("human", query)]}
+    if option == "Legal Advisory":
+        async for chunk in advisor_graph.astream(inputs, stream_mode="values"):
+            final_result = chunk
+        result = final_result["messages"][-1].content
+        return {"result": result}
+    elif option == "Legal Report Generation":
+        set_ret = RunnableParallel({"context": retriever, "query": RunnablePassthrough()})
+        rag_chain = set_ret | generator_template | chat | StrOutputParser()
+        report = rag_chain.invoke(query)
+        return {"report": report}
+    elif option == "Case Outcome Prediction":
+        async for chunk in predictor_graph.astream(inputs, stream_mode="values"):
+            final_prediction = chunk
+        prediction = final_prediction["messages"][-1]
+        return {"prediction": prediction}
+    else:
+        raise HTTPException(status_code=400, detail="Invalid option selected.")
+
+@app.post("/legal-advisory/")
+async def legal_advisory_endpoint(
+    query: str = Form(...),
+    files: List[UploadFile] = File(...)
+):
+    if not query:
+        raise HTTPException(status_code=400, detail="Please enter a query.")
+    pdf_content = process_files(files)
+    retriever, retrieval_tool = setup_retriever(pdf_content)
+    tools = [tavily_tool, retrieval_tool]
+    advisor_graph, _ = setup_agents(tools)
+    inputs = {"messages": [("human", query)]}
+    async for chunk in advisor_graph.astream(inputs, stream_mode="values"):
+        final_result = chunk
+    result = final_result["messages"][-1].content
+    return {"result": result}
+
+@app.post("/case-outcome-prediction/")
+async def case_outcome_prediction_endpoint(
+    query: str = Form(...),
+    files: List[UploadFile] = File(...)
+):
+    if not query:
+        raise HTTPException(status_code=400, detail="Please enter a query.")
+    pdf_content = process_files(files)
+    retriever, retrieval_tool = setup_retriever(pdf_content)
+    tools = [tavily_tool, retrieval_tool]
+    _, predictor_graph = setup_agents(tools)
+    inputs = {"messages": [("human", query)]}
+    async for chunk in predictor_graph.astream(inputs, stream_mode="values"):
+        final_prediction = chunk
+    prediction = final_prediction["messages"][-1].content
+    return {"prediction": prediction}
+
+@app.post("/report-generator/")
+async def report_generator_endpoint(
+    query: str = Form(...),
+    files: List[UploadFile] = File(...)
+):
+    if not query:
+        raise HTTPException(status_code=400, detail="Please enter a query.")
+    pdf_content = process_files(files)
+    retriever, _ = setup_retriever(pdf_content)
+    set_ret = RunnableParallel({"context": retriever, "query": RunnablePassthrough()})
+    rag_chain = set_ret | generator_template | chat | StrOutputParser()
+    report = rag_chain.invoke(query)
+    return {"report": report}
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=10000)

retrieval.py

@@ -0,0 +1,54 @@
+#RAG method
+from PyPDF2 import PdfReader
+from langchain.document_loaders import PyPDFLoader
+from langchain.docstore.document import Document
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.embeddings import HuggingFaceInferenceAPIEmbeddings
+from langchain_core.vectorstores import InMemoryVectorStore
+from dotenv import load_dotenv
+import os
+
+load_dotenv()
+hf_token = os.getenv("HF_TOKEN")
+
+def load_and_chunk_pdfs(directory_path):
+    docs = []
+    
+    for filename in os.listdir(directory_path):
+        if filename.endswith(".pdf"):
+            file_path = os.path.join(directory_path, filename)
+            
+            reader = PdfReader(file_path)
+            text = ""
+            for page in reader.pages:
+                text += page.extract_text()
+            
+            doc = Document(page_content=text, metadata={"source": filename})
+            docs.append(doc)
+    
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
+    
+    chunked_docs = text_splitter.split_documents(docs)
+    return chunked_docs
+
+
+def create_retriever(documents: list):
+    """
+    Function to create and return a retriever using HuggingFace Embeddings and InMemory VectorStore.
+
+    Args:
+        api_key (str): Hugging Face API key.
+        model_name (str): The model name for sentence transformer embeddings.
+        documents (list): The list of documents to be embedded and added to the vectorstore.
+
+    Returns:
+        retriever: A retriever object to query the vector store.
+    """
+    embeddings = HuggingFaceInferenceAPIEmbeddings(api_key=hf_token, model_name="sentence-transformers/all-MiniLM-l6-v2")
+
+    vectorstore = InMemoryVectorStore(embedding=embeddings)
+    vectorstore.add_documents(documents)
+
+    return vectorstore.as_retriever()
+
+

templates.py

@@ -0,0 +1,83 @@
+from pydantic import BaseModel, Field
+from typing import Literal
+from langchain.prompts import PromptTemplate
+advisor_template = """You are a legal research assistant tasked with providing 
+legal advice based on the given vectorstore context. If needed, conduct 
+additional research using the Tavily Search tool. Analyze the query for 
+specific legal issues, reference relevant sections of legal documents, and 
+ensure jurisdictional relevance. Consider conflicting interpretations or 
+unclear areas of law, and provide practical recommendations or next steps. 
+Include a disclaimer regarding the limitations of AI-generated legal advice.."""
+
+
+predictor_template = """
+You are a legal research assistant tasked with predicting the outcome of a 
+legal case using the provided vectorstore context. If needed, conduct 
+additional research using the Tavily Search tool. Analyze relevant legal 
+precedents, evidence, and arguments, and reference supporting sections from 
+legal documents. Provide a prediction of the case outcome with confidence 
+intervals (e.g., 70 percent hance of a favorable outcome), considering 
+jurisdictional differences. Highlight any uncertainties that could impact the
+ result, and include a disclaimer about the limitations of AI-generated 
+ predictions in real-world legal decisions.
+"""
+
+example_generator_template = """
+---
+### Legal Report template
+**Task Overview:**
+Generate a concise legal report based on the provided vectorstore according to the 
+context and query:
+{context}
+
+query: {query}
+**Report Structure:**
+1. **Title:**
+   - Clear and descriptive.
+2. **Introduction:**
+   - State the legal issue addressed.
+3. **Legal Precedents:**
+   - Summarize relevant precedents that apply.
+4. **Key Findings:**
+   - Present significant evidence and findings.
+5. **Analysis:**
+   - Discuss implications and potential outcomes.
+6. **Conclusion:**
+   - Summarize main points and recommendations.
+7. **Disclaimer:**
+   - Acknowledge that the report is AI-generated and may not account for all legal factors.
+---
+This streamlined template ensures clarity and professionalism without being overly detailed. Let me know if you need any adjustments!
+"""
+
+generator_template = PromptTemplate.from_template(template=example_generator_template)
+
+
+class LegalReportResponse(BaseModel):
+    """Respond to the user with this"""
+    return_direct: bool = False
+    case_summary: str = Field(description="A concise summary of the legal case")
+    relevant_precedents: str = Field(description="Key legal precedents or statutes relevant to the case")
+    evidence_analysis: str = Field(description="Summary of evidence and arguments presented by both sides")
+    key_findings: str = Field(description="Important findings or factors that influence the case")
+    conclusion: str = Field(description="A brief conclusion based on the analysis")
+
+class CaseOutcomePredictionResponse(BaseModel):
+    """Respond to the user with this"""
+    return_direct: bool = False
+    outcome_prediction: str = Field(description="Predicted outcome of the case")
+    confidence_interval: str = Field(description="Confidence interval for the prediction (e.g., 70% chance for the plaintiff)")
+    jurisdiction: str = Field(description="The legal jurisdiction relevant to the prediction")
+    uncertainty_factors: str = Field(description="Factors that might lead to different outcomes")
+    disclaimer: str = Field(description="AI-generated prediction disclaimer for limitations")
+
+class LegalAdviceResponse(BaseModel):
+    """Respond to the user with this"""
+    return_direct: bool = False
+    legal_issue: str = Field(description="The specific legal issue or query addressed")
+    advice: str = Field(description="The legal advice or recommendation provided based on the given context")
+    relevant_sections: str = Field(description="Relevant sections from legal documents or case law supporting the advice")
+    jurisdiction: str = Field(description="The jurisdiction applicable to the legal advice")
+    conflicting_interpretations: str = Field(description="Any conflicting interpretations or unclear areas of law")
+    next_steps: str = Field(description="Practical recommendations or next steps for the user to take")
+    disclaimer: str = Field(description="AI-generated legal advice disclaimer for limitations")

test.py

@@ -0,0 +1,21 @@
+import requests
+
+url = "http://127.0.0.1:8000/legal-assistance/"
+
+# List of files to upload (as a list of tuples)
+files = [
+    ('files', ('Sample Complaint.pdf', open('input/Sample Complaint.pdf', 'rb'), 'application/pdf')),
+    ('files', ('Sample Contract.pdf', open('input/Sample Contract.pdf', 'rb'), 'application/pdf'))
+]
+
+# The form data (query and option)
+data = {
+    'query': 'What are the possible legal options for fast retail pvt ltd.?',
+    'option': 'Legal Advisory'
+}
+
+# Make the request
+response = requests.post(url, files=files, data=data)
+
+# Check the response
+print(response.json())

tools.py

@@ -0,0 +1,29 @@
+from dotenv import load_dotenv
+from langchain_community.tools import TavilySearchResults
+from langchain.tools.retriever import create_retriever_tool
+import os
+load_dotenv()
+
+tavily_api_key = os.getenv("TAVILY_API_KEY")
+
+from langchain_community.tools import TavilySearchResults
+
+tavily_tool = TavilySearchResults(
+    max_results=5,
+    search_depth="advanced",
+    include_answer=True,
+    include_raw_content=True,
+    include_images=False,
+    include_domains=[
+        "indiankanoon.org",        # Indian case law
+        "barandbench.com",         # Legal news and updates
+        "legallyindia.com",        # Legal developments in India
+        "scconline.com",           # Supreme Court Cases Online
+        "lawtimesjournal.in",      # Legal news and case analysis
+        "lawyersclubindia.com",    # Legal community discussions
+        "vlex.in",                 # Global legal information with Indian focus
+        "taxmann.com"              # Taxation and corporate law in India
+    ],
+    exclude_domains=["globalsearch.com", "genericnews.com"]
+)
+