trustgraph-flow 0.16.4__tar.gz → 0.16.6__tar.gz

{trustgraph-flow-0.16.4 → trustgraph-flow-0.16.6}/PKG-INFO

@@ -1,12 +1,12 @@
 Metadata-Version: 2.1
 Name: trustgraph-flow
-Version: 0.16.4
+Version: 0.16.6
 Summary: TrustGraph provides a means to run a pipeline of flexible AI processing components in a flexible means to achieve a processing pipeline.
 Home-page: https://github.com/trustgraph-ai/trustgraph
 Author: trustgraph.ai
 Author-email: security@trustgraph.ai
 License: UNKNOWN
-Download-URL: https://github.com/trustgraph-ai/trustgraph/archive/refs/tags/v0.16.4.tar.gz
+Download-URL: https://github.com/trustgraph-ai/trustgraph/archive/refs/tags/v0.16.6.tar.gz
 Platform: UNKNOWN
 Classifier: Programming Language :: Python :: 3
 Classifier: License :: OSI Approved :: GNU General Public License v3 or later (GPLv3+)

trustgraph-flow-0.16.6/scripts/ge-query-pinecone

@@ -0,0 +1,6 @@
+#!/usr/bin/env python3
+
+from trustgraph.query.graph_embeddings.pinecone import run
+
+run()
+

trustgraph-flow-0.16.6/scripts/ge-write-pinecone

@@ -0,0 +1,6 @@
+#!/usr/bin/env python3
+
+from trustgraph.storage.graph_embeddings.pinecone import run
+
+run()
+

{trustgraph-flow-0.16.4 → trustgraph-flow-0.16.6}/setup.py

@@ -60,6 +60,7 @@ setuptools.setup(
         "jsonschema",
         "aiohttp",
         "aiopulsar-py",
+        "pinecone[grpc]",
     ],
     scripts=[
         "scripts/api-gateway",
@@ -74,8 +75,10 @@ setuptools.setup(
         "scripts/embeddings-ollama",
         "scripts/embeddings-vectorize",
         "scripts/ge-query-milvus",
+        "scripts/ge-query-pinecone",
         "scripts/ge-query-qdrant",
         "scripts/ge-write-milvus",
+        "scripts/ge-write-pinecone",
         "scripts/ge-write-qdrant",
         "scripts/graph-rag",
         "scripts/kg-extract-definitions",

{trustgraph-flow-0.16.4 → trustgraph-flow-0.16.6}/trustgraph/api/gateway/service.py

@@ -14,7 +14,7 @@ module = ".".join(__name__.split(".")[1:-1])
 
 import asyncio
 import argparse
-from aiohttp import web
+from aiohttp import web, WSMsgType
 import json
 import logging
 import uuid
@@ -47,9 +47,13 @@ from ... schema import GraphRagQuery, GraphRagResponse
 from ... schema import graph_rag_request_queue
 from ... schema import graph_rag_response_queue
 
-from ... schema import TriplesQueryRequest, TriplesQueryResponse
+from ... schema import TriplesQueryRequest, TriplesQueryResponse, Triples
 from ... schema import triples_request_queue
 from ... schema import triples_response_queue
+from ... schema import triples_store_queue
+
+from ... schema import GraphEmbeddings
+from ... schema import graph_embeddings_store_queue
 
 from ... schema import AgentRequest, AgentResponse
 from ... schema import agent_request_queue
@@ -84,6 +88,11 @@ def to_subgraph(x):
         for t in x
     ]
 
+class Running:
+    def __init__(self): self.running = True
+    def get(self): return self.running
+    def stop(self): self.running = False
+
 class Publisher:
 
     def __init__(self, pulsar_host, topic, schema=None, max_size=10,
@@ -132,6 +141,7 @@ class Subscriber:
         self.consumer_name = consumer_name
         self.schema = schema
         self.q = {}
+        self.full = {}
 
     async def run(self):
         while True:
@@ -145,10 +155,19 @@ class Subscriber:
                     ) as consumer:
                         while True:
                             msg = await consumer.receive()
-                            id = msg.properties()["id"]
+
+                            try:
+                                id = msg.properties()["id"]
+                            except:
+                                id = None
+
                             value = msg.value()
                             if id in self.q:
                                 await self.q[id].put(value)
+
+                            for q in self.full.values():
+                                await q.put(value)
+
             except Exception as e:
                 print("Exception:", e, flush=True)
          
@@ -164,6 +183,59 @@ class Subscriber:
         if id in self.q:
             del self.q[id]
     
+    async def subscribe_all(self, id):
+        q = asyncio.Queue()
+        self.full[id] = q
+        return q
+
+    async def unsubscribe_all(self, id):
+        if id in self.full:
+            del self.full[id]
+
+def serialize_triples(message):
+    return {
+        "metadata": {
+            "id": message.metadata.id,
+            "metadata": [
+                {
+                    "s": t.s.value,
+                    "p": t.p.value,
+                    "o": t.o.value,
+                }
+                for t in message.metadata.metadata
+            ],
+            "user": message.metadata.user,
+            "collection": message.metadata.collection,
+        },
+        "triples": [        
+            {
+                "s": t.s.value,
+                "p": t.p.value,
+                "o": t.o.value,
+            }
+            for t in message.triples
+        ]
+    }
+    
+def serialize_graph_embeddings(message):
+    return {
+        "metadata": {
+            "id": message.metadata.id,
+            "metadata": [
+                {
+                    "s": t.s.value,
+                    "p": t.p.value,
+                    "o": t.o.value,
+                }
+                for t in message.metadata.metadata
+            ],
+            "user": message.metadata.user,
+            "collection": message.metadata.collection,
+        },
+        "vectors": message.vectors,
+        "entity": message.entity.value,
+    }
+    
 class Api:
 
     def __init__(self, **config):
@@ -243,6 +315,28 @@ class Api:
             JsonSchema(EmbeddingsResponse)
         )
 
+        self.triples_tap = Subscriber(
+            self.pulsar_host, triples_store_queue,
+            "api-gateway", "api-gateway",
+            schema=JsonSchema(Triples)
+        )
+
+        self.triples_pub = Publisher(
+            self.pulsar_host, triples_store_queue,
+            schema=JsonSchema(Triples)
+        )
+
+        self.graph_embeddings_tap = Subscriber(
+            self.pulsar_host, graph_embeddings_store_queue,
+            "api-gateway", "api-gateway",
+            schema=JsonSchema(GraphEmbeddings)
+        )
+
+        self.graph_embeddings_pub = Publisher(
+            self.pulsar_host, graph_embeddings_store_queue,
+            schema=JsonSchema(GraphEmbeddings)
+        )
+
         self.document_out = Publisher(
             self.pulsar_host, document_ingest_queue,
             schema=JsonSchema(Document),
@@ -264,6 +358,20 @@ class Api:
             web.post("/api/v1/embeddings", self.embeddings),
             web.post("/api/v1/load/document", self.load_document),
             web.post("/api/v1/load/text", self.load_text),
+            web.get("/api/v1/ws", self.socket),
+
+            web.get("/api/v1/stream/triples", self.stream_triples),
+            web.get(
+                "/api/v1/stream/graph-embeddings",
+                self.stream_graph_embeddings
+            ),
+
+            web.get("/api/v1/load/triples", self.load_triples),
+            web.get(
+                "/api/v1/load/graph-embeddings",
+                self.load_graph_embeddings
+            ),
+
         ])
 
     async def llm(self, request):
@@ -660,6 +768,169 @@ class Api:
                 { "error": str(e) }
             )
 
+    async def socket(self, request):
+
+        ws = web.WebSocketResponse()
+        await ws.prepare(request)
+
+        async for msg in ws:
+            if msg.type == WSMsgType.TEXT:
+                if msg.data == 'close':
+                    await ws.close()
+                else:
+                    await ws.send_str(msg.data + '/answer')
+            elif msg.type == WSMsgType.ERROR:
+                print('ws connection closed with exception %s' %
+                      ws.exception())
+
+        print('websocket connection closed')
+
+        return ws
+
+    async def stream(self, q, ws, running, fn):
+
+        while running.get():
+            try:
+                resp = await asyncio.wait_for(q.get(), 0.5)
+                await ws.send_json(fn(resp))
+
+            except TimeoutError:
+                continue
+
+            except Exception as e:
+                print(f"Exception: {str(e)}", flush=True)
+
+    async def stream_triples(self, request):
+
+        id = str(uuid.uuid4())
+
+        q = await self.triples_tap.subscribe_all(id)
+        running = Running()
+
+        ws = web.WebSocketResponse()
+        await ws.prepare(request)
+
+        tsk = asyncio.create_task(self.stream(
+            q,
+            ws,
+            running,
+            serialize_triples,
+        ))
+
+        async for msg in ws:
+            if msg.type == WSMsgType.ERROR:
+                break
+            else:
+                # Ignore incoming messages
+                pass
+
+        running.stop()
+
+        await self.triples_tap.unsubscribe_all(id)
+        await tsk
+
+        return ws
+
+    async def stream_graph_embeddings(self, request):
+
+        id = str(uuid.uuid4())
+
+        q = await self.graph_embeddings_tap.subscribe_all(id)
+        running = Running()
+
+        ws = web.WebSocketResponse()
+        await ws.prepare(request)
+
+        tsk = asyncio.create_task(self.stream(
+            q,
+            ws,
+            running,
+            serialize_graph_embeddings,
+        ))
+
+        async for msg in ws:
+            if msg.type == WSMsgType.ERROR:
+                break
+            else:
+                # Ignore incoming messages
+                pass
+
+        running.stop()
+
+        await self.graph_embeddings_tap.unsubscribe_all(id)
+        await tsk
+
+        return ws
+
+    async def load_triples(self, request):
+
+        ws = web.WebSocketResponse()
+        await ws.prepare(request)
+
+        async for msg in ws:
+
+            try:
+
+                if msg.type == WSMsgType.TEXT:
+
+                    data = msg.json()
+
+                    elt = Triples(
+                        metadata=Metadata(
+                            id=data["metadata"]["id"],
+                            metadata=to_subgraph(data["metadata"]["metadata"]),
+                            user=data["metadata"]["user"],
+                            collection=data["metadata"]["collection"],
+                        ),
+                        triples=to_subgraph(data["triples"]),
+                    )
+
+                    await self.triples_pub.send(None, elt)
+
+                elif msg.type == WSMsgType.ERROR:
+                    break
+
+            except Exception as e:
+
+                print("Exception:", e)
+
+        return ws
+
+    async def load_graph_embeddings(self, request):
+
+        ws = web.WebSocketResponse()
+        await ws.prepare(request)
+
+        async for msg in ws:
+
+            try:
+
+                if msg.type == WSMsgType.TEXT:
+
+                    data = msg.json()
+
+                    elt = GraphEmbeddings(
+                        metadata=Metadata(
+                            id=data["metadata"]["id"],
+                            metadata=to_subgraph(data["metadata"]["metadata"]),
+                            user=data["metadata"]["user"],
+                            collection=data["metadata"]["collection"],
+                        ),
+                        entity=to_value(data["entity"]),
+                        vectors=data["vectors"],
+                    )
+
+                    await self.graph_embeddings_pub.send(None, elt)
+
+                elif msg.type == WSMsgType.ERROR:
+                    break
+
+            except Exception as e:
+
+                print("Exception:", e)
+
+        return ws
+
     async def app_factory(self):
 
         self.llm_pub_task = asyncio.create_task(self.llm_in.run())
@@ -688,6 +959,22 @@ class Api:
             self.embeddings_out.run()
         )
 
+        self.triples_tap_task = asyncio.create_task(
+            self.triples_tap.run()
+        )
+
+        self.triples_pub_task = asyncio.create_task(
+            self.triples_pub.run()
+        )
+
+        self.graph_embeddings_tap_task = asyncio.create_task(
+            self.graph_embeddings_tap.run()
+        )
+
+        self.graph_embeddings_pub_task = asyncio.create_task(
+            self.graph_embeddings_pub.run()
+        )
+
         self.doc_ingest_pub_task = asyncio.create_task(self.document_out.run())
 
         self.text_ingest_pub_task = asyncio.create_task(self.text_out.run())
@@ -699,7 +986,6 @@ class Api:
 
 def run():
 
-
     parser = argparse.ArgumentParser(
         prog="api-gateway",
         description=__doc__

trustgraph-flow-0.16.6/trustgraph/flow_version.py

@@ -0,0 +1 @@
+__version__ = "0.16.6"

trustgraph-flow-0.16.6/trustgraph/query/doc_embeddings/pinecone/service.py

@@ -0,0 +1,142 @@
+
+"""
+Document embeddings query service.  Input is vector, output is an array
+of chunks.  Pinecone implementation.
+"""
+
+from pinecone import Pinecone, ServerlessSpec
+from pinecone.grpc import PineconeGRPC, GRPCClientConfig
+
+import uuid
+import os
+
+from .... schema import DocumentEmbeddingsRequest, DocumentEmbeddingsResponse
+from .... schema import Error, Value
+from .... schema import document_embeddings_request_queue
+from .... schema import document_embeddings_response_queue
+from .... base import ConsumerProducer
+
+module = ".".join(__name__.split(".")[1:-1])
+
+default_input_queue = document_embeddings_request_queue
+default_output_queue = document_embeddings_response_queue
+default_subscriber = module
+default_api_key = os.getenv("PINECONE_API_KEY", "not-specified")
+
+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)
+
+        self.url = params.get("url", None)
+        self.api_key = params.get("api_key", default_api_key)
+
+        if self.url:
+
+            self.pinecone = PineconeGRPC(
+                api_key = self.api_key,
+                host = self.url
+            )
+
+        else:
+
+            self.pinecone = Pinecone(api_key = self.api_key)
+
+        super(Processor, self).__init__(
+            **params | {
+                "input_queue": input_queue,
+                "output_queue": output_queue,
+                "subscriber": subscriber,
+                "input_schema": DocumentEmbeddingsRequest,
+                "output_schema": DocumentEmbeddingsResponse,
+                "url": self.url,
+            }
+        )
+
+    def handle(self, msg):
+
+        try:
+
+            v = msg.value()
+
+            # Sender-produced ID
+            id = msg.properties()["id"]
+
+            print(f"Handling input {id}...", flush=True)
+
+            chunks = []
+
+            for vec in v.vectors:
+
+                dim = len(vec)
+
+                index_name = (
+                    "d-" + v.user + "-" + str(dim)
+                )
+
+                index = self.pinecone.Index(index_name)
+
+                results = index.query(
+                    namespace=v.collection,
+                    vector=vec,
+                    top_k=v.limit,
+                    include_values=False,
+                    include_metadata=True
+                )
+
+                search_result = self.client.query_points(
+                    collection_name=collection,
+                    query=vec,
+                    limit=v.limit,
+                    with_payload=True,
+                ).points
+
+                for r in results.matches:
+                    doc = r.metadata["doc"]
+                    chunks.add(doc)
+
+            print("Send response...", flush=True)
+            r = DocumentEmbeddingsResponse(documents=chunks, 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 = DocumentEmbeddingsResponse(
+                error=Error(
+                    type = "llm-error",
+                    message = str(e),
+                ),
+                documents=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(
+            '-t', '--store-uri',
+            default=default_store_uri,
+            help=f'Milvus store URI (default: {default_store_uri})'
+        )
+
+def run():
+
+    Processor.start(module, __doc__)
+