Provide Indexer to index files

backend/src/main/java/com/cloudera/cai/rag/external/RagBackendClient.java

@@ -68,7 +68,8 @@ public void indexFile(
               + "/data_sources/"
               + ragDocument.dataSourceId()
               + "/documents/download-and-index",
-          new IndexRequest(bucketName, ragDocument.s3Path(), configuration));
+          new IndexRequest(
+              ragDocument.documentId(), bucketName, ragDocument.s3Path(), configuration));
     } catch (IOException e) {
       throw new RuntimeException(e);
     }
@@ -97,6 +98,7 @@ public void deleteSession(Long sessionId) {
   }
 
   record IndexRequest(
+      @JsonProperty("document_id") String documentId,
       @JsonProperty("s3_bucket_name") String s3BucketName,
       @JsonProperty("s3_document_key") String s3DocumentKey,
       IndexConfiguration configuration) {}

backend/src/test/java/com/cloudera/cai/rag/external/RagBackendClientTest.java

@@ -57,7 +57,7 @@ void indexFile() {
     Tracker<TrackedHttpRequest<?>> tracker = new Tracker<>();
     RagBackendClient client = new RagBackendClient(SimpleHttpClient.createNull(tracker));
     IndexConfiguration indexConfiguration = new IndexConfiguration(123, 2);
-    RagDocument document = indexRequest("s3Path", 1234L);
+    RagDocument document = indexRequest("documentId", "s3Path", 1234L);
 
     client.indexFile(document, "bucketName", indexConfiguration);
 
@@ -68,14 +68,15 @@ void indexFile() {
             new TrackedHttpRequest<>(
                 HttpMethod.POST,
                 "http://rag-backend:8000/data_sources/" + 1234L + "/documents/download-and-index",
-                new RagBackendClient.IndexRequest("bucketName", "s3Path", indexConfiguration)));
+                new RagBackendClient.IndexRequest(
+                    "documentId", "bucketName", "s3Path", indexConfiguration)));
   }
 
   @Test
   void createSummary() {
     Tracker<TrackedHttpRequest<?>> tracker = new Tracker<>();
     RagBackendClient client = new RagBackendClient(SimpleHttpClient.createNull(tracker));
-    RagDocument document = indexRequest("s3Path", 1234L);
+    RagDocument document = indexRequest("documentId", "s3Path", 1234L);
 
     client.createSummary(document, "bucketName");
 
@@ -134,13 +135,25 @@ void deleteSession() {
   void null_handlesThrowable() {
     RagBackendClient client =
         RagBackendClient.createNull(new Tracker<>(), new NotFound("not found"));
-    RagDocument document = indexRequest("s3Path", 1234L);
+    RagDocument document = indexRequest("documentId", "s3Path", 1234L);
     assertThatThrownBy(() -> client.indexFile(document, "fakeit", null))
         .isInstanceOf(NotFound.class);
   }
 
-  private static RagDocument indexRequest(String s3Path, Long dataSourceId) {
+  private static RagDocument indexRequest(String documentId, String s3Path, Long dataSourceId) {
     return new RagDocument(
-        null, null, dataSourceId, null, s3Path, null, null, null, null, null, null, null, null);
+        null,
+        null,
+        dataSourceId,
+        documentId,
+        s3Path,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null);
   }
 }

llm-service/app/ai/__init__.py

@@ -0,0 +1,37 @@
+#
+#  CLOUDERA APPLIED MACHINE LEARNING PROTOTYPE (AMP)
+#  (C) Cloudera, Inc. 2024
+#  All rights reserved.
+#
+#  Applicable Open Source License: Apache 2.0
+#
+#  NOTE: Cloudera open source products are modular software products
+#  made up of hundreds of individual components, each of which was
+#  individually copyrighted.  Each Cloudera open source product is a
+#  collective work under U.S. Copyright Law. Your license to use the
+#  collective work is as provided in your written agreement with
+#  Cloudera.  Used apart from the collective work, this file is
+#  licensed for your use pursuant to the open source license
+#  identified above.
+#
+#  This code is provided to you pursuant a written agreement with
+#  (i) Cloudera, Inc. or (ii) a third-party authorized to distribute
+#  this code. If you do not have a written agreement with Cloudera nor
+#  with an authorized and properly licensed third party, you do not
+#  have any rights to access nor to use this code.
+#
+#  Absent a written agreement with Cloudera, Inc. ("Cloudera") to the
+#  contrary, A) CLOUDERA PROVIDES THIS CODE TO YOU WITHOUT WARRANTIES OF ANY
+#  KIND; (B) CLOUDERA DISCLAIMS ANY AND ALL EXPRESS AND IMPLIED
+#  WARRANTIES WITH RESPECT TO THIS CODE, INCLUDING BUT NOT LIMITED TO
+#  IMPLIED WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY AND
+#  FITNESS FOR A PARTICULAR PURPOSE; (C) CLOUDERA IS NOT LIABLE TO YOU,
+#  AND WILL NOT DEFEND, INDEMNIFY, NOR HOLD YOU HARMLESS FOR ANY CLAIMS
+#  ARISING FROM OR RELATED TO THE CODE; AND (D)WITH RESPECT TO YOUR EXERCISE
+#  OF ANY RIGHTS GRANTED TO YOU FOR THE CODE, CLOUDERA IS NOT LIABLE FOR ANY
+#  DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, PUNITIVE OR
+#  CONSEQUENTIAL DAMAGES INCLUDING, BUT NOT LIMITED TO, DAMAGES
+#  RELATED TO LOST REVENUE, LOST PROFITS, LOSS OF INCOME, LOSS OF
+#  BUSINESS ADVANTAGE OR UNAVAILABILITY, OR LOSS OR CORRUPTION OF
+#  DATA.
+#

llm-service/app/ai/indexing/__init__.py

@@ -0,0 +1,37 @@
+#
+#  CLOUDERA APPLIED MACHINE LEARNING PROTOTYPE (AMP)
+#  (C) Cloudera, Inc. 2024
+#  All rights reserved.
+#
+#  Applicable Open Source License: Apache 2.0
+#
+#  NOTE: Cloudera open source products are modular software products
+#  made up of hundreds of individual components, each of which was
+#  individually copyrighted.  Each Cloudera open source product is a
+#  collective work under U.S. Copyright Law. Your license to use the
+#  collective work is as provided in your written agreement with
+#  Cloudera.  Used apart from the collective work, this file is
+#  licensed for your use pursuant to the open source license
+#  identified above.
+#
+#  This code is provided to you pursuant a written agreement with
+#  (i) Cloudera, Inc. or (ii) a third-party authorized to distribute
+#  this code. If you do not have a written agreement with Cloudera nor
+#  with an authorized and properly licensed third party, you do not
+#  have any rights to access nor to use this code.
+#
+#  Absent a written agreement with Cloudera, Inc. ("Cloudera") to the
+#  contrary, A) CLOUDERA PROVIDES THIS CODE TO YOU WITHOUT WARRANTIES OF ANY
+#  KIND; (B) CLOUDERA DISCLAIMS ANY AND ALL EXPRESS AND IMPLIED
+#  WARRANTIES WITH RESPECT TO THIS CODE, INCLUDING BUT NOT LIMITED TO
+#  IMPLIED WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY AND
+#  FITNESS FOR A PARTICULAR PURPOSE; (C) CLOUDERA IS NOT LIABLE TO YOU,
+#  AND WILL NOT DEFEND, INDEMNIFY, NOR HOLD YOU HARMLESS FOR ANY CLAIMS
+#  ARISING FROM OR RELATED TO THE CODE; AND (D)WITH RESPECT TO YOUR EXERCISE
+#  OF ANY RIGHTS GRANTED TO YOU FOR THE CODE, CLOUDERA IS NOT LIABLE FOR ANY
+#  DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, PUNITIVE OR
+#  CONSEQUENTIAL DAMAGES INCLUDING, BUT NOT LIMITED TO, DAMAGES
+#  RELATED TO LOST REVENUE, LOST PROFITS, LOSS OF INCOME, LOSS OF
+#  BUSINESS ADVANTAGE OR UNAVAILABILITY, OR LOSS OR CORRUPTION OF
+#  DATA.
+#

llm-service/app/ai/indexing/index.py

@@ -0,0 +1,131 @@
+#
+#  CLOUDERA APPLIED MACHINE LEARNING PROTOTYPE (AMP)
+#  (C) Cloudera, Inc. 2024
+#  All rights reserved.
+#
+#  Applicable Open Source License: Apache 2.0
+#
+#  NOTE: Cloudera open source products are modular software products
+#  made up of hundreds of individual components, each of which was
+#  individually copyrighted.  Each Cloudera open source product is a
+#  collective work under U.S. Copyright Law. Your license to use the
+#  collective work is as provided in your written agreement with
+#  Cloudera.  Used apart from the collective work, this file is
+#  licensed for your use pursuant to the open source license
+#  identified above.
+#
+#  This code is provided to you pursuant a written agreement with
+#  (i) Cloudera, Inc. or (ii) a third-party authorized to distribute
+#  this code. If you do not have a written agreement with Cloudera nor
+#  with an authorized and properly licensed third party, you do not
+#  have any rights to access nor to use this code.
+#
+#  Absent a written agreement with Cloudera, Inc. ("Cloudera") to the
+#  contrary, A) CLOUDERA PROVIDES THIS CODE TO YOU WITHOUT WARRANTIES OF ANY
+#  KIND; (B) CLOUDERA DISCLAIMS ANY AND ALL EXPRESS AND IMPLIED
+#  WARRANTIES WITH RESPECT TO THIS CODE, INCLUDING BUT NOT LIMITED TO
+#  IMPLIED WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY AND
+#  FITNESS FOR A PARTICULAR PURPOSE; (C) CLOUDERA IS NOT LIABLE TO YOU,
+#  AND WILL NOT DEFEND, INDEMNIFY, NOR HOLD YOU HARMLESS FOR ANY CLAIMS
+#  ARISING FROM OR RELATED TO THE CODE; AND (D)WITH RESPECT TO YOUR EXERCISE
+#  OF ANY RIGHTS GRANTED TO YOU FOR THE CODE, CLOUDERA IS NOT LIABLE FOR ANY
+#  DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, PUNITIVE OR
+#  CONSEQUENTIAL DAMAGES INCLUDING, BUT NOT LIMITED TO, DAMAGES
+#  RELATED TO LOST REVENUE, LOST PROFITS, LOSS OF INCOME, LOSS OF
+#  BUSINESS ADVANTAGE OR UNAVAILABILITY, OR LOSS OR CORRUPTION OF
+#  DATA.
+#
+
+from dataclasses import dataclass
+import logging
+import os
+from typing import Dict, List, Type
+
+from .readers.pdf import PDFReader
+from .readers.nop import NopReader
+from llama_index.core.readers.base import BaseReader
+from llama_index.core.node_parser import SentenceSplitter
+from llama_index.core.schema import Document
+from llama_index.core.base.embeddings.base import BaseEmbedding
+from ...services.vector_store import VectorStore
+from llama_index.core.node_parser.interface import BaseNode
+
+logger = logging.getLogger(__name__)
+
+READERS: Dict[str, Type[BaseReader]] = {
+    ".pdf": PDFReader,
+    ".txt": NopReader,
+    ".md": NopReader,
+}
+CHUNKABLE_FILE_EXTENSIONS = set(
+    [
+        ".pdf",
+        ".txt",
+        ".md",
+    ]
+)
+
+@dataclass
+class NotSupportedFileExtensionError(Exception):
+    file_extension: str
+
+class Indexer:
+    def __init__(self, data_source_id: int, splitter: SentenceSplitter, embedding_model: BaseEmbedding, chunks_vector_store: VectorStore):
+        self.data_source_id = data_source_id
+        self.splitter = splitter
+        self.embedding_model = embedding_model
+        self.chunks_vector_store = chunks_vector_store
+
+    def index_file(self, file_path: str, file_id: str):
+        logger.debug(f"Indexing file: {file_path}")
+
+        file_extension = os.path.splitext(file_path)[1]
+        reader_cls = READERS.get(file_extension)
+        if not reader_cls:
+            raise NotSupportedFileExtensionError(file_extension)
+
+        reader = reader_cls()
+
+        logger.debug(f"Parsing file: {file_path}")
+
+        documents = self._documents_in_file(reader, file_path, file_id)
+        if file_extension in CHUNKABLE_FILE_EXTENSIONS:
+            logger.debug(f"Chunking file: {file_path}")
+            chunks = [chunk for document in documents for chunk in self._chunks_in_document(document)]
+        else:
+            chunks = documents
+
+        texts = [chunk.text for chunk in chunks]
+        logger.debug(f"Embedding {len(texts)} chunks")
+        embeddings = self.embedding_model.get_text_embedding_batch(texts)
+
+        for chunk, embedding in zip(chunks, embeddings):
+            chunk.embedding = embedding
+
+        logger.debug(f"Adding {len(chunks)} chunks to vector store")
+        chunks_vector_store = self.chunks_vector_store.access_vector_store()
+        chunks_vector_store.add(chunks)
+
+        logger.debug(f"Indexing file: {file_path} completed")
+
+    def _documents_in_file(self, reader: BaseReader, file_path: str, file_id: str) -> List[Document]:
+        documents = reader.load_data(file_path)
+
+        for i, document in enumerate(documents):
+            # Update the document metadata
+            document.metadata["file_id"] = file_id
+            document.metadata["document_part_number"] = i
+            document.metadata["data_source_id"] = self.data_source_id
+
+        return documents
+
+    def _chunks_in_document(self, document: Document) -> List[BaseNode]:
+        chunks = self.splitter.get_nodes_from_documents([document])
+
+        for j, chunk in enumerate(chunks):
+            chunk.metadata["file_id"] = document.metadata["file_id"]
+            chunk.metadata["document_part_number"] = document.metadata["document_part_number"]
+            chunk.metadata["chunk_number"] = j
+            chunk.metadata["data_source_id"] = document.metadata["data_source_id"]
+
+        return chunks

llm-service/app/ai/indexing/readers/__init__.py

@@ -0,0 +1,37 @@
+#
+#  CLOUDERA APPLIED MACHINE LEARNING PROTOTYPE (AMP)
+#  (C) Cloudera, Inc. 2024
+#  All rights reserved.
+#
+#  Applicable Open Source License: Apache 2.0
+#
+#  NOTE: Cloudera open source products are modular software products
+#  made up of hundreds of individual components, each of which was
+#  individually copyrighted.  Each Cloudera open source product is a
+#  collective work under U.S. Copyright Law. Your license to use the
+#  collective work is as provided in your written agreement with
+#  Cloudera.  Used apart from the collective work, this file is
+#  licensed for your use pursuant to the open source license
+#  identified above.
+#
+#  This code is provided to you pursuant a written agreement with
+#  (i) Cloudera, Inc. or (ii) a third-party authorized to distribute
+#  this code. If you do not have a written agreement with Cloudera nor
+#  with an authorized and properly licensed third party, you do not
+#  have any rights to access nor to use this code.
+#
+#  Absent a written agreement with Cloudera, Inc. ("Cloudera") to the
+#  contrary, A) CLOUDERA PROVIDES THIS CODE TO YOU WITHOUT WARRANTIES OF ANY
+#  KIND; (B) CLOUDERA DISCLAIMS ANY AND ALL EXPRESS AND IMPLIED
+#  WARRANTIES WITH RESPECT TO THIS CODE, INCLUDING BUT NOT LIMITED TO
+#  IMPLIED WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY AND
+#  FITNESS FOR A PARTICULAR PURPOSE; (C) CLOUDERA IS NOT LIABLE TO YOU,
+#  AND WILL NOT DEFEND, INDEMNIFY, NOR HOLD YOU HARMLESS FOR ANY CLAIMS
+#  ARISING FROM OR RELATED TO THE CODE; AND (D)WITH RESPECT TO YOUR EXERCISE
+#  OF ANY RIGHTS GRANTED TO YOU FOR THE CODE, CLOUDERA IS NOT LIABLE FOR ANY
+#  DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, PUNITIVE OR
+#  CONSEQUENTIAL DAMAGES INCLUDING, BUT NOT LIMITED TO, DAMAGES
+#  RELATED TO LOST REVENUE, LOST PROFITS, LOSS OF INCOME, LOSS OF
+#  BUSINESS ADVANTAGE OR UNAVAILABILITY, OR LOSS OR CORRUPTION OF
+#  DATA.
+#

llm-service/app/ai/indexing/readers/nop.py

@@ -0,0 +1,46 @@
+#
+#  CLOUDERA APPLIED MACHINE LEARNING PROTOTYPE (AMP)
+#  (C) Cloudera, Inc. 2024
+#  All rights reserved.
+#
+#  Applicable Open Source License: Apache 2.0
+#
+#  NOTE: Cloudera open source products are modular software products
+#  made up of hundreds of individual components, each of which was
+#  individually copyrighted.  Each Cloudera open source product is a
+#  collective work under U.S. Copyright Law. Your license to use the
+#  collective work is as provided in your written agreement with
+#  Cloudera.  Used apart from the collective work, this file is
+#  licensed for your use pursuant to the open source license
+#  identified above.
+#
+#  This code is provided to you pursuant a written agreement with
+#  (i) Cloudera, Inc. or (ii) a third-party authorized to distribute
+#  this code. If you do not have a written agreement with Cloudera nor
+#  with an authorized and properly licensed third party, you do not
+#  have any rights to access nor to use this code.
+#
+#  Absent a written agreement with Cloudera, Inc. ("Cloudera") to the
+#  contrary, A) CLOUDERA PROVIDES THIS CODE TO YOU WITHOUT WARRANTIES OF ANY
+#  KIND; (B) CLOUDERA DISCLAIMS ANY AND ALL EXPRESS AND IMPLIED
+#  WARRANTIES WITH RESPECT TO THIS CODE, INCLUDING BUT NOT LIMITED TO
+#  IMPLIED WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY AND
+#  FITNESS FOR A PARTICULAR PURPOSE; (C) CLOUDERA IS NOT LIABLE TO YOU,
+#  AND WILL NOT DEFEND, INDEMNIFY, NOR HOLD YOU HARMLESS FOR ANY CLAIMS
+#  ARISING FROM OR RELATED TO THE CODE; AND (D)WITH RESPECT TO YOUR EXERCISE
+#  OF ANY RIGHTS GRANTED TO YOU FOR THE CODE, CLOUDERA IS NOT LIABLE FOR ANY
+#  DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, PUNITIVE OR
+#  CONSEQUENTIAL DAMAGES INCLUDING, BUT NOT LIMITED TO, DAMAGES
+#  RELATED TO LOST REVENUE, LOST PROFITS, LOSS OF INCOME, LOSS OF
+#  BUSINESS ADVANTAGE OR UNAVAILABILITY, OR LOSS OR CORRUPTION OF
+#  DATA.
+#
+
+from typing import List
+from llama_index.core.readers.base import BaseReader
+from llama_index.core.schema import Document
+
+class NopReader(BaseReader):
+    def load_data(self, file_path: str) -> List[Document]:
+        with open(file_path, "r") as f:
+            return [Document(text=f.read())]