File size: 8,117 Bytes
34b531b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
from __future__ import annotations

from collections.abc import Callable

from qdrant_client import QdrantClient
from qdrant_client.http.models import Distance, FieldCondition, Filter, MatchValue, PointStruct, VectorParams

from app.config import (
    EMBEDDING_API_URL,
    EMBEDDING_BATCH_SIZE,
    EMBEDDING_CACHE_ENABLED,
    EMBEDDING_CACHE_PATH,
    EMBEDDING_MODEL,
    QDRANT_API_KEY,
    QDRANT_COLLECTION,
    QDRANT_URL,
)
from app.embedding_cache import EmbeddingCache, embedding_cache_key
from app.embeddings import get_embedding_model
from app.schemas import Chunk, RetrievedChunk


def get_qdrant_client() -> QdrantClient:
    kwargs = {"url": QDRANT_URL}
    if QDRANT_API_KEY:
        kwargs["api_key"] = QDRANT_API_KEY
    return QdrantClient(**kwargs)


def ensure_collection(client: QdrantClient, vector_size: int) -> None:
    collections = client.get_collections().collections
    if any(collection.name == QDRANT_COLLECTION for collection in collections):
        return
    client.create_collection(
        collection_name=QDRANT_COLLECTION,
        vectors_config=VectorParams(size=vector_size, distance=Distance.COSINE),
    )


def recreate_collection(client: QdrantClient, vector_size: int) -> None:
    collections = client.get_collections().collections
    if any(collection.name == QDRANT_COLLECTION for collection in collections):
        client.delete_collection(collection_name=QDRANT_COLLECTION)
    client.create_collection(
        collection_name=QDRANT_COLLECTION,
        vectors_config=VectorParams(size=vector_size, distance=Distance.COSINE),
    )


def chunk_payload(chunk: Chunk) -> dict:
    return {
        "text": chunk.text,
        "ticker": chunk.ticker,
        "scope": chunk.scope,
        "modality": chunk.modality,
        "source_path": chunk.source_path,
        "chunk_index": chunk.chunk_index,
        "structure_type": chunk.structure_type,
        "heading_path": chunk.heading_path,
        "token_count": chunk.token_count,
        "metadata": chunk.metadata,
    }


def index_chunks(

    chunks: list[Chunk],

    batch_size: int | None = None,

    rebuild: bool = True,

    progress_callback: Callable[[dict], None] | None = None,

) -> int:
    batch_size = batch_size or EMBEDDING_BATCH_SIZE
    embedding_model = get_embedding_model()
    cache = EmbeddingCache(EMBEDDING_CACHE_PATH) if EMBEDDING_CACHE_ENABLED else None
    client = get_qdrant_client()
    try:
        if rebuild:
            recreate_collection(client, embedding_model.dim)
        else:
            ensure_collection(client, embedding_model.dim)

        indexed = 0
        total_batches = (len(chunks) + batch_size - 1) // batch_size if chunks else 0
        for start in range(0, len(chunks), batch_size):
            batch = chunks[start : start + batch_size]
            batch_number = (start // batch_size) + 1
            vectors, cache_hits, cache_misses = embed_index_batch(batch, embedding_model, cache)
            if progress_callback:
                progress_callback(
                    {
                        "stage": "embedding",
                        "batch_number": batch_number,
                        "total_batches": total_batches,
                        "batch_size": len(batch),
                        "indexed_so_far": indexed,
                        "total_chunks": len(chunks),
                        "embedding_dim": embedding_model.dim,
                        "cache_hits": cache_hits,
                        "cache_misses": cache_misses,
                    }
                )
            points = [
                PointStruct(id=chunk.id, vector=vector, payload=chunk_payload(chunk))
                for chunk, vector in zip(batch, vectors)
            ]
            client.upsert(collection_name=QDRANT_COLLECTION, points=points)
            indexed += len(points)
            if progress_callback:
                progress_callback(
                    {
                        "stage": "upsert",
                        "batch_number": batch_number,
                        "total_batches": total_batches,
                        "batch_size": len(points),
                        "indexed_so_far": indexed,
                        "total_chunks": len(chunks),
                        "embedding_dim": embedding_model.dim,
                        "cache_hits": cache_hits,
                        "cache_misses": cache_misses,
                    }
                )
        return indexed
    finally:
        if cache:
            cache.close()


def embed_index_batch(

    batch: list[Chunk],

    embedding_model,

    cache: EmbeddingCache | None,

) -> tuple[list[list[float]], int, int]:
    if not cache:
        return embedding_model.encode([chunk.text for chunk in batch]), 0, len(batch)

    keys = [
        embedding_cache_key(
            chunk.text,
            provider=embedding_model.provider,
            model=EMBEDDING_MODEL,
            dim=embedding_model.dim,
            api_url=EMBEDDING_API_URL,
        )
        for chunk in batch
    ]
    cached = cache.get_many(keys)
    missing_indexes = [index for index, key in enumerate(keys) if key not in cached]

    if missing_indexes:
        missing_vectors = embedding_model.encode([batch[index].text for index in missing_indexes])
        cache.set_many(
            {
                keys[index]: vector
                for index, vector in zip(missing_indexes, missing_vectors)
            }
        )
        for index, vector in zip(missing_indexes, missing_vectors):
            cached[keys[index]] = vector

    return [cached[key] for key in keys], len(batch) - len(missing_indexes), len(missing_indexes)


def search_points(

    client: QdrantClient,

    query_vector: list[float],

    query_filter: Filter | None,

    limit: int,

):
    if hasattr(client, "search"):
        return client.search(
            collection_name=QDRANT_COLLECTION,
            query_vector=query_vector,
            query_filter=query_filter,
            limit=limit,
            with_payload=True,
        )

    response = client.query_points(
        collection_name=QDRANT_COLLECTION,
        query=query_vector,
        query_filter=query_filter,
        limit=limit,
        with_payload=True,
    )
    return response.points


def retrieve(query: str, top_k: int, ticker: str | None = None) -> list[RetrievedChunk]:
    embedding_model = get_embedding_model()
    client = get_qdrant_client()
    ensure_collection(client, embedding_model.dim)

    query_filter = None
    if ticker:
        query_filter = Filter(
            must=[FieldCondition(key="ticker", match=MatchValue(value=ticker.upper()))]
        )

    hits = search_points(
        client=client,
        query_vector=embedding_model.encode([query])[0],
        query_filter=query_filter,
        limit=top_k,
    )
    retrieved: list[RetrievedChunk] = []
    for hit in hits:
        payload = hit.payload or {}
        source_path = str(payload.get("source_path", ""))
        ticker_value = str(payload.get("ticker", ""))
        scope = str(payload.get("scope") or ticker_value or "")
        if ticker_value.upper() == "MARKET" or "world_market" in source_path or "/market/" in source_path.replace("\\", "/"):
            ticker_value = ""
            scope = "market"
        retrieved.append(
            RetrievedChunk(
                id=str(hit.id),
                text=str(payload.get("text", "")),
                score=float(hit.score),
                ticker=ticker_value,
                modality=str(payload.get("modality", "")),
                source_path=source_path,
                structure_type=str(payload.get("structure_type", "")),
                heading_path=list(payload.get("heading_path") or []),
                metadata=dict(payload.get("metadata") or {}),
                scope=scope,
            )
        )
    return retrieved