feat/vectara-destination-to-v2

CHANGELOG.md

@@ -1,3 +1,10 @@
+## 0.2.0-dev0
+
+### Enhancements
+
+* **Migrate Vectara Destination Connector to v2**
+
+
 ## 0.2.0
 
 ### Enhancements

requirements/connectors/vectara.in

@@ -1,3 +1,5 @@
 -c ../common/constraints.txt
 
 requests
+aiofiles
+httpx

requirements/connectors/vectara.txt

@@ -8,6 +8,8 @@ idna==3.10
     # via requests
 requests==2.32.3
     # via -r ./connectors/vectara.in
+aiofiles==24.1.0
+    # via -r ./connectors/vectara.in
 urllib3==1.26.20
     # via
     #   -c ./connectors/../common/constraints.txt

test/integration/connectors/test_vectara.py

@@ -0,0 +1,267 @@
+import json
+import os
+import time
+from pathlib import Path
+from typing import Generator
+from uuid import uuid4
+
+import pytest
+import requests
+
+from test.integration.connectors.utils.constants import DESTINATION_TAG
+from test.integration.utils import requires_env
+from unstructured_ingest.v2.interfaces.file_data import FileData, SourceIdentifiers
+from unstructured_ingest.v2.logger import logger
+from unstructured_ingest.v2.processes.connectors.vectara import (
+    CONNECTOR_TYPE as VECTARA_CONNECTOR_TYPE,
+)
+from unstructured_ingest.v2.processes.connectors.vectara import (
+    VectaraAccessConfig,
+    VectaraConnectionConfig,
+    VectaraUploader,
+    VectaraUploaderConfig,
+    VectaraUploadStager,
+    VectaraUploadStagerConfig,
+)
+
+
+def validate_upload(response: dict, expected_data: dict):
+    element_id = expected_data["element_id"]
+    expected_text = expected_data["text"]
+    filename = expected_data["metadata"]["filename"]
+    filetype = expected_data["metadata"]["filetype"]
+    page_number = expected_data["metadata"]["page_number"]
+
+    response = response["search_results"][0]
+
+    assert response is not None
+    assert response["text"] == expected_text
+    assert response["part_metadata"]["element_id"] == element_id
+    assert response["part_metadata"]["filename"] == filename
+    assert response["part_metadata"]["filetype"] == filetype
+    assert response["part_metadata"]["page_number"] == page_number
+
+
+@requires_env("VECTARA_OAUTH_CLIENT_ID", "VECTARA_OAUTH_SECRET", "VECTARA_CUSTOMER_ID")
+def _get_jwt_token():
+    """Connect to the server and get a JWT token."""
+    customer_id = os.environ["VECTARA_CUSTOMER_ID"]
+    token_endpoint = (
+        f"https://vectara-prod-{customer_id}.auth.us-west-2.amazoncognito.com/oauth2/token"
+    )
+    headers = {
+        "Content-Type": "application/x-www-form-urlencoded",
+    }
+    data = {
+        "grant_type": "client_credentials",
+        "client_id": os.environ["VECTARA_OAUTH_CLIENT_ID"],
+        "client_secret": os.environ["VECTARA_OAUTH_SECRET"],
+    }
+
+    response = requests.post(token_endpoint, headers=headers, data=data)
+    response.raise_for_status()
+    response_json = response.json()
+
+    return response_json.get("access_token")
+
+
+def query_data(corpus_key: str, element_id: str) -> dict:
+
+    url = f"https://api.vectara.io/v2/corpora/{corpus_key}/query"
+
+    # the query below requires the corpus to have filter attributes for element_id
+
+    data = json.dumps(
+        {
+            "query": "string",
+            "search": {
+                "metadata_filter": f"part.element_id = '{element_id}'",
+                "lexical_interpolation": 1,
+                "limit": 10,
+            },
+        }
+    )
+
+    jwt_token = _get_jwt_token()
+    headers = {
+        "Content-Type": "application/json",
+        "Accept": "application/json",
+        "Authorization": f"Bearer {jwt_token}",
+        "X-source": "unstructured",
+    }
+
+    response = requests.post(url, headers=headers, data=data)
+    response.raise_for_status()
+    response_json = response.json()
+
+    return response_json
+
+
+def create_corpora(corpus_key: str, corpus_name: str) -> None:
+    url = "https://api.vectara.io/v2/corpora"
+    data = json.dumps({"key": corpus_key, "name": corpus_name, "description": "integration test"})
+    jwt_token = _get_jwt_token()
+    headers = {
+        "Content-Type": "application/json",
+        "Accept": "application/json",
+        "Authorization": f"Bearer {jwt_token}",
+        "X-source": "unstructured",
+    }
+
+    response = requests.post(url, headers=headers, data=data)
+    response.raise_for_status()
+
+
+def replace_filter_attributes(corpus_key: str) -> None:
+    url = f"https://api.vectara.io/v2/corpora/{corpus_key}/replace_filter_attributes"
+    data = json.dumps(
+        {
+            "filter_attributes": [
+                {"name": "element_id", "level": "part", "indexed": True, "type": "text"}
+            ]
+        }
+    )
+    jwt_token = _get_jwt_token()
+    headers = {
+        "Content-Type": "application/json",
+        "Accept": "application/json",
+        "Authorization": f"Bearer {jwt_token}",
+        "X-source": "unstructured",
+    }
+
+    response = requests.post(url, headers=headers, data=data)
+    response.raise_for_status()
+
+
+def delete_corpora(corpus_key: str) -> None:
+    url = f"https://api.vectara.io/v2/corpora/{corpus_key}"
+
+    jwt_token = _get_jwt_token()
+    headers = {
+        "Content-Type": "application/json",
+        "Accept": "application/json",
+        "Authorization": f"Bearer {jwt_token}",
+        "X-source": "unstructured",
+    }
+
+    response = requests.delete(url, headers=headers)
+    response.raise_for_status()
+
+
+def list_corpora() -> list:
+    url = "https://api.vectara.io/v2/corpora?limit=100"
+    jwt_token = _get_jwt_token()
+    headers = {
+        "Content-Type": "application/json",
+        "Accept": "application/json",
+        "Authorization": f"Bearer {jwt_token}",
+        "X-source": "unstructured",
+    }
+    response = requests.get(url, headers=headers)
+    response.raise_for_status()
+    response_json = response.json()
+    if response_json.get("corpora"):
+        return [item["key"] for item in response_json.get("corpora")]
+    else:
+        return []
+
+
+def wait_for_ready(corpus_key: str, timeout=60, interval=2) -> None:
+    def is_ready_status():
+        corpora_list = list_corpora()
+        return corpus_key in corpora_list
+
+    start = time.time()
+    is_ready = is_ready_status()
+    while not is_ready and time.time() - start < timeout:
+        time.sleep(interval)
+        is_ready = is_ready_status()
+    if not is_ready:
+        raise TimeoutError("time out waiting for corpus to be ready")
+
+
+def wait_for_delete(corpus_key: str, timeout=60, interval=2) -> None:
+    start = time.time()
+    while time.time() - start < timeout:
+        corpora_list = list_corpora()
+        if corpus_key not in corpora_list:
+            return
+        time.sleep(interval)
+
+    raise TimeoutError("time out waiting for corpus to delete")
+
+
+@pytest.fixture
+def corpora_util() -> Generator[str, None, None]:
+    random_id = str(uuid4()).split("-")[0]
+    corpus_key = f"ingest-test-{random_id}"
+    corpus_name = "ingest-test"
+    logger.info(f"Creating corpus with key: {corpus_key}")
+    try:
+        create_corpora(corpus_key, corpus_name)
+        replace_filter_attributes(corpus_key)
+        wait_for_ready(corpus_key=corpus_key)
+        yield corpus_key
+    except Exception as e:
+        logger.error(f"failed to create corpus {corpus_key}: {e}")
+    finally:
+        logger.info(f"deleting corpus: {corpus_key}")
+        delete_corpora(corpus_key)
+        wait_for_delete(corpus_key=corpus_key)
+
+
+@pytest.mark.asyncio
+@pytest.mark.tags(VECTARA_CONNECTOR_TYPE, DESTINATION_TAG, "vectara")
+@requires_env("VECTARA_OAUTH_CLIENT_ID", "VECTARA_OAUTH_SECRET", "VECTARA_CUSTOMER_ID")
+async def test_vectara_destination(
+    upload_file: Path, tmp_path: Path, corpora_util: str, retries=30, interval=10
+):
+    corpus_key = corpora_util
+    connection_kwargs = {
+        "customer_id": os.environ["VECTARA_CUSTOMER_ID"],
+        "corpus_key": corpus_key,
+    }
+
+    oauth_client_id = os.environ["VECTARA_OAUTH_CLIENT_ID"]
+    oauth_secret = os.environ["VECTARA_OAUTH_SECRET"]
+
+    file_data = FileData(
+        source_identifiers=SourceIdentifiers(fullpath=upload_file.name, filename=upload_file.name),
+        connector_type=VECTARA_CONNECTOR_TYPE,
+        identifier="mock-file-data",
+    )
+
+    stager_config = VectaraUploadStagerConfig(batch_size=10)
+    stager = VectaraUploadStager(upload_stager_config=stager_config)
+    new_upload_file = stager.run(
+        elements_filepath=upload_file,
+        output_dir=tmp_path,
+        output_filename=upload_file.name,
+        file_data=file_data,
+    )
+
+    uploader = VectaraUploader(
+        connection_config=VectaraConnectionConfig(
+            **connection_kwargs,
+            access_config=VectaraAccessConfig(
+                oauth_client_id=oauth_client_id, oauth_secret=oauth_secret
+            ),
+        ),
+        upload_config=VectaraUploaderConfig(),
+    )
+
+    if uploader.is_async():
+        await uploader.run_async(path=new_upload_file, file_data=file_data)
+
+    with upload_file.open() as upload_fp:
+        elements = json.load(upload_fp)
+    first_element = elements[0]
+
+    for i in range(retries):
+        response = query_data(corpus_key, first_element["element_id"])
+        if not response["search_results"]:
+            time.sleep(interval)
+        else:
+            break
+
+    validate_upload(response=response, expected_data=first_element)

unstructured_ingest/__version__.py

@@ -1 +1 @@
-__version__ = "0.2.0"  # pragma: no cover
+__version__ = "0.2.0-dev0"  # pragma: no cover

unstructured_ingest/v2/examples/vectara.py

@@ -0,0 +1,54 @@
+from pathlib import Path
+
+from unstructured_ingest.v2.interfaces import ProcessorConfig
+from unstructured_ingest.v2.logger import logger
+from unstructured_ingest.v2.pipeline.pipeline import Pipeline
+from unstructured_ingest.v2.processes.chunker import ChunkerConfig
+from unstructured_ingest.v2.processes.connectors.local import (
+    LocalConnectionConfig,
+    LocalDownloaderConfig,
+    LocalIndexerConfig,
+)
+from unstructured_ingest.v2.processes.connectors.vectara import (
+    CONNECTOR_TYPE,
+    VectaraAccessConfig,
+    VectaraConnectionConfig,
+    VectaraUploaderConfig,
+    VectaraUploadStagerConfig,
+)
+from unstructured_ingest.v2.processes.embedder import EmbedderConfig
+from unstructured_ingest.v2.processes.partitioner import PartitionerConfig
+
+base_path = Path(__file__).parent.parent.parent.parent
+docs_path = base_path / "example-docs"
+work_dir = base_path / "tmp_ingest" / CONNECTOR_TYPE
+output_path = work_dir / "output"
+download_path = work_dir / "download"
+
+if __name__ == "__main__":
+    logger.info(f"writing all content in: {work_dir.resolve()}")
+    Pipeline.from_configs(
+        context=ProcessorConfig(work_dir=str(work_dir.resolve())),
+        indexer_config=LocalIndexerConfig(input_path=str(docs_path.resolve()) + "/multisimple/"),
+        downloader_config=LocalDownloaderConfig(download_dir=download_path),
+        source_connection_config=LocalConnectionConfig(),
+        partitioner_config=PartitionerConfig(strategy="fast"),
+        chunker_config=ChunkerConfig(
+            chunking_strategy="by_title",
+            chunk_include_orig_elements=False,
+            chunk_max_characters=1500,
+            chunk_multipage_sections=True,
+        ),
+        embedder_config=EmbedderConfig(embedding_provider="huggingface"),
+        destination_connection_config=VectaraConnectionConfig(
+            access_config=VectaraAccessConfig(
+                oauth_client_id="fill oauth_client_id", oauth_secret="fill oauth_secret"
+            ),
+            customer_id="fill customer_id",
+            corpus_name="fill corpus_name",
+            corpus_key="fill corpus_key",
+            token_url="fill token_url",
+        ),
+        stager_config=VectaraUploadStagerConfig(batch_size=10),
+        uploader_config=VectaraUploaderConfig(),
+    ).run()

unstructured_ingest/v2/processes/connectors/__init__.py

@@ -48,6 +48,8 @@
 from .singlestore import singlestore_destination_entry
 from .slack import CONNECTOR_TYPE as SLACK_CONNECTOR_TYPE
 from .slack import slack_source_entry
+from .vectara import CONNECTOR_TYPE as VECTARA_CONNECTOR_TYPE
+from .vectara import vectara_destination_entry
 from .weaviate import CONNECTOR_TYPE as WEAVIATE_CONNECTOR_TYPE
 from .weaviate import weaviate_destination_entry
 
@@ -103,3 +105,5 @@
 add_source_entry(source_type=OUTLOOK_CONNECTOR_TYPE, entry=outlook_source_entry)
 
 add_source_entry(source_type=SLACK_CONNECTOR_TYPE, entry=slack_source_entry)
+
+add_destination_entry(destination_type=VECTARA_CONNECTOR_TYPE, entry=vectara_destination_entry)