Spaces:
Sleeping
Sleeping
File size: 9,240 Bytes
62516b8 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 | from __future__ import annotations
from datetime import UTC, datetime
from typing import Iterable
from uuid import uuid4
from langchain_core.documents import Document
from langchain_core.embeddings import DeterministicFakeEmbedding, Embeddings
from src.config import settings
def _normalize_chunks(chunks: Iterable[str]) -> list[str]:
normalized: list[str] = []
for chunk in chunks:
clean = " ".join(chunk.split())
if clean:
normalized.append(clean)
return normalized
def split_material_text(
text: str,
*,
chunk_size: int,
chunk_overlap: int,
) -> list[str]:
if chunk_size <= 0:
raise ValueError("chunk_size must be greater than zero.")
if chunk_overlap < 0:
raise ValueError("chunk_overlap must be zero or greater.")
if chunk_overlap >= chunk_size:
raise ValueError("chunk_overlap must be smaller than chunk_size.")
clean = " ".join(text.split())
if not clean:
return []
if len(clean) <= chunk_size:
return [clean]
step = chunk_size - chunk_overlap
chunks: list[str] = []
for start in range(0, len(clean), step):
segment = clean[start : start + chunk_size]
if not segment.strip():
continue
chunks.append(segment)
if start + chunk_size >= len(clean):
break
return _normalize_chunks(chunks)
def _to_float_list(values: object) -> list[float]:
if hasattr(values, "tolist"):
values = values.tolist()
if isinstance(values, (list, tuple)):
converted: list[float] = []
for item in values:
if hasattr(item, "item"):
item = item.item()
converted.append(float(item))
return converted
if hasattr(values, "item"):
values = values.item()
return [float(values)]
def _to_float_vectors(values: object) -> list[list[float]]:
if hasattr(values, "tolist"):
values = values.tolist()
if not isinstance(values, (list, tuple)):
return [_to_float_list(values)]
if not values:
return []
first = values[0]
if isinstance(first, (list, tuple)) or hasattr(first, "tolist"):
return [_to_float_list(vector) for vector in values]
return [_to_float_list(values)]
def _short_error_message(exc: Exception, *, max_chars: int = 260) -> str:
compact = " ".join(str(exc).split())
if len(compact) <= max_chars:
return compact
return f"{compact[: max_chars - 3]}..."
def _build_embeddings() -> tuple[Embeddings, str | None]:
try:
from chromadb.utils.embedding_functions import DefaultEmbeddingFunction
class ChromaDefaultEmbeddings(Embeddings):
def __init__(self) -> None:
self._fn = DefaultEmbeddingFunction()
def embed_documents(self, texts: list[str]) -> list[list[float]]:
vectors = self._fn(texts)
return _to_float_vectors(vectors)
def embed_query(self, text: str) -> list[float]:
vector = self._fn([text])[0]
return _to_float_list(vector)
return ChromaDefaultEmbeddings(), None
except Exception as exc:
return (
DeterministicFakeEmbedding(size=256),
f"RAG embedding fallback mode: {exc}",
)
class MaterialRAGStore:
def __init__(self) -> None:
self._warning: str | None = None
self._fallback_docs: list[Document] = []
self._vectorstore = None
self._embeddings, emb_warning = _build_embeddings()
if emb_warning:
self._warning = emb_warning
try:
from langchain_chroma import Chroma
self._vectorstore = Chroma(
collection_name=settings.rag_collection_name,
persist_directory=settings.chroma_persist_dir,
embedding_function=self._embeddings,
)
except Exception as exc:
warning = f"Material RAG vectorstore fallback mode: {exc}"
self._warning = f"{self._warning}; {warning}" if self._warning else warning
@property
def init_warning(self) -> str | None:
return self._warning
def index_material(
self,
*,
user_id: str,
document_id: str,
filename: str,
file_type: str,
text: str,
) -> tuple[int, list[str]]:
warnings: list[str] = []
chunks = split_material_text(
text,
chunk_size=settings.rag_chunk_size,
chunk_overlap=settings.rag_chunk_overlap,
)
if not chunks:
raise ValueError("No chunks produced for RAG indexing.")
now = datetime.now(UTC).isoformat()
docs: list[Document] = []
doc_ids: list[str] = []
for idx, chunk in enumerate(chunks):
chunk_id = f"{document_id}:chunk:{idx}"
metadata = {
"chunk_id": chunk_id,
"user_id": user_id,
"document_id": document_id,
"filename": filename,
"file_type": file_type,
"chunk_index": idx,
"uploaded_at": now,
"source": "uploaded_material_chunk",
}
docs.append(Document(page_content=chunk, metadata=metadata))
doc_ids.append(chunk_id)
self._fallback_docs.extend(docs)
if self._vectorstore is not None:
try:
self._vectorstore.add_documents(docs, ids=doc_ids)
except Exception as exc:
warnings.append(
f"RAG indexing fallback to memory: {_short_error_message(exc)}"
)
else:
warnings.append("RAG vectorstore unavailable; using in-memory fallback.")
return len(docs), warnings
def retrieve_for_generation(
self,
*,
user_id: str,
document_id: str,
queries: list[str],
) -> tuple[list[Document], list[str]]:
warnings: list[str] = []
if not queries:
return [], warnings
where = {"user_id": user_id, "document_id": document_id}
if self._vectorstore is None:
docs = self._fallback_retrieve(
user_id=user_id,
document_id=document_id,
queries=queries,
)
if docs:
return docs, warnings
warnings.append("RAG retrieval returned no chunks from fallback memory.")
return [], warnings
try:
collected: list[Document] = []
for query in queries:
docs = self._vectorstore.max_marginal_relevance_search(
query=query,
k=settings.rag_top_k,
fetch_k=settings.rag_fetch_k,
lambda_mult=settings.rag_mmr_lambda,
filter=where,
)
collected.extend(docs)
deduped = self._dedupe_docs(collected)
if deduped:
return deduped, warnings
warnings.append("RAG retrieval returned no chunks.")
return [], warnings
except Exception as exc:
warnings.append(
f"RAG retrieval failed; fallback to memory: {_short_error_message(exc)}"
)
docs = self._fallback_retrieve(
user_id=user_id,
document_id=document_id,
queries=queries,
)
return docs, warnings
def new_document_id(self) -> str:
return f"doc-{uuid4().hex}"
@staticmethod
def _dedupe_docs(docs: list[Document]) -> list[Document]:
deduped: list[Document] = []
seen: set[str] = set()
for doc in docs:
metadata = doc.metadata or {}
chunk_id = str(metadata.get("chunk_id") or "")
key = chunk_id if chunk_id else doc.page_content[:128]
if key in seen:
continue
seen.add(key)
deduped.append(doc)
return deduped
def _fallback_retrieve(
self,
*,
user_id: str,
document_id: str,
queries: list[str],
) -> list[Document]:
candidates = [
doc
for doc in self._fallback_docs
if (doc.metadata or {}).get("user_id") == user_id
and (doc.metadata or {}).get("document_id") == document_id
]
if not candidates:
return []
query_terms = {
term.lower()
for term in " ".join(queries).split()
if len(term.strip()) >= 3
}
scored: list[tuple[int, Document]] = []
for doc in candidates:
text = doc.page_content.lower()
score = sum(1 for term in query_terms if term in text)
scored.append((score, doc))
scored.sort(
key=lambda item: (
item[0],
-int((item[1].metadata or {}).get("chunk_index", 0)),
),
reverse=True,
)
limit = max(1, settings.rag_top_k)
return [item[1] for item in scored[:limit]]
|