Forming embeddings-hf package

trustgraph-embeddings-hf/scripts/embeddings-hf

@@ -0,0 +1,6 @@
+#!/usr/bin/env python3
+
+from trustgraph.embeddings_hf.embeddings.hf import run
+
+run()
+

trustgraph-embeddings-hf/setup.py

@@ -0,0 +1,47 @@
+import setuptools
+import os
+
+with open("../README.md", "r") as fh:
+    long_description = fh.read()
+
+version = "0.11.0"
+
+setuptools.setup(
+    name="trustgraph-embeddings-hf",
+    version=version,
+    author="trustgraph.ai",
+    author_email="security@trustgraph.ai",
+    description="HuggingFace embeddings support for TrustGraph.",
+    long_description=long_description,
+    long_description_content_type="text/markdown",
+    url="https://github.com/trustgraph-ai/trustgraph",
+    packages=setuptools.find_namespace_packages(
+        where='./',
+#         include=['trustgraph.core']
+    ),
+    classifiers=[ 
+        "Programming Language :: Python :: 3",
+        "License :: OSI Approved :: GNU General Public License v3 or later (GPLv3+)",
+        "Operating System :: OS Independent",
+    ],
+    python_requires='>=3.8',
+    download_url = "https://github.com/trustgraph-ai/trustgraph/archive/refs/tags/v" + version + ".tar.gz",
+    install_requires=[
+        "trustgraph-core",
+        "torch",
+        "urllib3",
+        "transformers",
+        "sentence-transformers",
+        "langchain",
+        "langchain-core",
+        "langchain-huggingface",
+        "langchain-community",
+        "huggingface-hub",
+        "pulsar-client",
+        "pyyaml",
+        "prometheus-client",
+    ],
+    scripts=[
+        "scripts/embeddings-hf",
+    ]
+)

trustgraph-embeddings-hf/trustgraph/embeddings_hf/embeddings/hf/__init__.py

@@ -0,0 +1,3 @@
+
+from . hf import *
+

trustgraph-embeddings-hf/trustgraph/embeddings_hf/embeddings/hf/__main__.py

@@ -0,0 +1,7 @@
+#!/usr/bin/env python3
+
+from . hf import run
+
+if __name__ == '__main__':
+    run()
+

trustgraph-embeddings-hf/trustgraph/embeddings_hf/embeddings/hf/hf.py

@@ -0,0 +1,99 @@
+
+"""
+Embeddings service, applies an embeddings model selected from HuggingFace.
+Input is text, output is embeddings vector.
+"""
+
+from langchain_huggingface import HuggingFaceEmbeddings
+
+from ... schema import EmbeddingsRequest, EmbeddingsResponse, Error
+from ... schema import embeddings_request_queue, embeddings_response_queue
+from ... log_level import LogLevel
+from ... base import ConsumerProducer
+
+module = ".".join(__name__.split(".")[1:-1])
+
+default_input_queue = embeddings_request_queue
+default_output_queue = embeddings_response_queue
+default_subscriber = module
+default_model="all-MiniLM-L6-v2"
+
+class Processor(ConsumerProducer):
+
+    def __init__(self, **params):
+
+        input_queue = params.get("input_queue", default_input_queue)
+        output_queue = params.get("output_queue", default_output_queue)
+        subscriber = params.get("subscriber", default_subscriber)
+        model = params.get("model", default_model)
+
+        super(Processor, self).__init__(
+            **params | {
+                "input_queue": input_queue,
+                "output_queue": output_queue,
+                "subscriber": subscriber,
+                "input_schema": EmbeddingsRequest,
+                "output_schema": EmbeddingsResponse,
+            }
+        )
+
+        self.embeddings = HuggingFaceEmbeddings(model_name=model)
+
+    def handle(self, msg):
+
+        v = msg.value()
+
+        # Sender-produced ID
+        id = msg.properties()["id"]
+
+        print(f"Handling input {id}...", flush=True)
+
+        try:
+
+            text = v.text
+            embeds = self.embeddings.embed_documents([text])
+
+            print("Send response...", flush=True)
+            r = EmbeddingsResponse(vectors=embeds, error=None)
+            self.producer.send(r, properties={"id": id})
+
+            print("Done.", flush=True)
+
+
+        except Exception as e:
+
+            print(f"Exception: {e}")
+
+            print("Send error response...", flush=True)
+
+            r = EmbeddingsResponse(
+                error=Error(
+                    type = "llm-error",
+                    message = str(e),
+                ),
+                response=None,
+            )
+
+            self.producer.send(r, properties={"id": id})
+
+            self.consumer.acknowledge(msg)
+            
+
+    @staticmethod
+    def add_args(parser):
+
+        ConsumerProducer.add_args(
+            parser, default_input_queue, default_subscriber,
+            default_output_queue,
+        )
+
+        parser.add_argument(
+            '-m', '--model',
+            default="all-MiniLM-L6-v2",
+            help=f'LLM model (default: all-MiniLM-L6-v2)'
+        )
+
+def run():
+
+    Processor.start(module, __doc__)
+