File size: 4,125 Bytes
6064cea
06407c6
9281159
 
e310fed
 
9281159
 
 
 
06407c6
 
6064cea
06407c6
6064cea
 
 
e755476
6064cea
 
 
e755476
6064cea
 
 
6f7fbf8
 
6064cea
 
e755476
6064cea
 
 
6f7fbf8
e755476
6f7fbf8
6064cea
 
6f7fbf8
 
 
 
 
6064cea
 
 
 
 
 
 
e755476
6064cea
6f7fbf8
6064cea
 
e755476
6064cea
 
 
6f7fbf8
 
 
6064cea
 
 
 
e755476
6064cea
 
e755476
6064cea
 
 
 
70c8520
6064cea
 
 
6f7fbf8
6064cea
6f7fbf8
e755476
6064cea
e755476
6064cea
 
 
e755476
6064cea
 
 
6f7fbf8
6064cea
e310fed
6f7fbf8
e310fed
 
 
6f7fbf8
 
 
 
 
 
9281159
665b7f8
6f7fbf8
 
 
665b7f8
 
e755476
 
665b7f8
 
 
 
 
 
 
 
6064cea
6f7fbf8
 
 
 
6064cea
 
 
6f7fbf8
 
665b7f8
 
6f7fbf8
 
 
5f6d20c
 
6f7fbf8
 
 
 
9281159
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
from __future__ import annotations

import logging

from legora.config import get_settings
from legora.logging import log_stage
from retrieval.hybrid_retrieval import HybridRetriever
from retrieval.judgments import KanoonClient

logger = logging.getLogger(__name__)


class LegalRetriever:

    def __init__(self):
        self.hybrid = HybridRetriever()
        self.kanoon = KanoonClient()

    # =====================================================
    # BUILD CONTEXT
    # =====================================================

    def build_context(
        self,
        hybrid_result,
        judgments,
        include_full_docs: bool = False
    ):
        context = []

        # -----------------------------------------
        # Statutory Retrieval
        # -----------------------------------------
        context.append("===== STATUTORY LAW =====\n")

        for point in hybrid_result["reranked_results"]:
            payload = point.payload

            if include_full_docs:
                doc_text = payload.get('document', '')
            else:
                doc_text = payload.get('text', '')

            context.append(
                f"""
DOCUMENT:
{payload.get('document')}

CHUNK:
{payload.get('chunk_id')}

TEXT:
{doc_text}
"""
            )

        # -----------------------------------------
        # Graph Expansion
        # -----------------------------------------
        context.append("\n===== KNOWLEDGE GRAPH =====\n")
        for node_id, graph_data in hybrid_result["graph_context"].items():
            context.append(f"\nNODE: {node_id}")
            context.append(
                f"""
Ancestors:
{graph_data['ancestors']}

Children:
{graph_data['children']}

References:
{graph_data['references']}
"""
            )

        # -----------------------------------------
        # Judgments
        # -----------------------------------------
        context.append("\n===== JUDGMENTS =====\n")
        for judgment in judgments:
            context.append(str(judgment))

        return "\n".join(context)

    # =====================================================
    # RETRIEVE
    # =====================================================

    def retrieve(
        self,
        query: str,
        **kwargs
    ):
        settings = get_settings()
        # Extract known parameters with defaults
        vector_k = kwargs.get('vector_k', settings.top_k)
        rerank_k = kwargs.get('rerank_k', settings.rerank_k)
        judgment_k = kwargs.get('judgment_k', settings.judgment_k)
        include_full_docs = kwargs.get('include_full_docs', False)

        # Allow 'top_k' as alias for vector_k
        if 'top_k' in kwargs:
            vector_k = kwargs['top_k']

        # Forward only options supported by each retriever.
        graph_options = {} if kwargs.get("include_graph", True) else {"include_graph": False}
        hybrid_result = self.hybrid.retrieve(
            query=query,
            vector_k=vector_k,
            rerank_k=rerank_k,
            **graph_options
        )

        judgments = []
        if kwargs.get("include_judgments", True):
            with log_stage("judgments") as metrics:
                judgments = self.kanoon.retrieve(
                    query=query,
                    top_k=judgment_k
                )
                metrics.candidate_count = len(judgments)

        final_context = self.build_context(
            hybrid_result,
            judgments,
            include_full_docs=include_full_docs
        )

        return {
            "query": query,
            "statutory_results": hybrid_result["reranked_results"],
            "vector_results": hybrid_result.get("vector_results", []),
            "node_ids": hybrid_result.get("node_ids", []),
            "graph_context": hybrid_result["graph_context"],
            "judgments": judgments,
            "final_context": final_context
        }


if __name__ == "__main__":
    query = "My bike was stolen from outside my house. What punishment can the offender face?"
    retriever = LegalRetriever()
    logger.debug('%s', retriever.retrieve(query))