trustgraph-flow 0.20.0__tar.gz → 0.20.2__tar.gz

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: trustgraph-flow
-Version: 0.20.0
+Version: 0.20.2
 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
-Download-URL: https://github.com/trustgraph-ai/trustgraph/archive/refs/tags/v0.20.0.tar.gz
+Download-URL: https://github.com/trustgraph-ai/trustgraph/archive/refs/tags/v0.20.2.tar.gz
 Author: trustgraph.ai
 Author-email: security@trustgraph.ai
 Classifier: Programming Language :: Python :: 3

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/setup.py

@@ -35,32 +35,33 @@ setuptools.setup(
     download_url = "https://github.com/trustgraph-ai/trustgraph/archive/refs/tags/v" + version + ".tar.gz",
     install_requires=[
         "trustgraph-base>=0.20,<0.21",
-        "urllib3",
-        "rdflib",
-        "pymilvus",
+        "aiohttp",
+        "anthropic",
+        "cassandra-driver",
+        "cohere",
+        "cryptography",
+        "falkordb",
+        "google-generativeai",
+        "ibis",
+        "jsonschema",
         "langchain",
+        "langchain-community",
         "langchain-core",
         "langchain-text-splitters",
-        "langchain-community",
-        "requests",
-        "cassandra-driver",
+        "neo4j",
+        "openai",
+        "pinecone[grpc]",
+        "prometheus-client",
         "pulsar-client",
+        "pymilvus",
         "pypdf",
+        "pyyaml",
         "qdrant-client",
+        "rdflib",
+        "requests",
         "tabulate",
-        "anthropic",
-        "pyyaml",
-        "prometheus-client",
-        "cohere",
-        "openai",
-        "neo4j",
         "tiktoken",
-        "google-generativeai",
-        "ibis",
-        "jsonschema",
-        "aiohttp",
-        "pinecone[grpc]",
-        "falkordb",
+        "urllib3",
     ],
     scripts=[
         "scripts/agent-manager-react",

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/agent/react/service.py

@@ -14,8 +14,6 @@ from ... schema import AgentRequest, AgentResponse, AgentStep
 from ... schema import agent_request_queue, agent_response_queue
 from ... schema import prompt_request_queue as pr_request_queue
 from ... schema import prompt_response_queue as pr_response_queue
-from ... schema import text_completion_request_queue as tc_request_queue
-from ... schema import text_completion_response_queue as tc_response_queue
 from ... schema import graph_rag_request_queue as gr_request_queue
 from ... schema import graph_rag_response_queue as gr_response_queue
 from ... clients.prompt_client import PromptClient
@@ -133,12 +131,6 @@ class Processor(ConsumerProducer):
         prompt_response_queue = params.get(
             "prompt_response_queue", pr_response_queue
         )
-        text_completion_request_queue = params.get(
-            "text_completion_request_queue", tc_request_queue
-        )
-        text_completion_response_queue = params.get(
-            "text_completion_response_queue", tc_response_queue
-        )
         graph_rag_request_queue = params.get(
             "graph_rag_request_queue", gr_request_queue
         )
@@ -155,8 +147,6 @@ class Processor(ConsumerProducer):
                 "output_schema": AgentResponse,
                 "prompt_request_queue": prompt_request_queue,
                 "prompt_response_queue": prompt_response_queue,
-                "text_completion_request_queue": tc_request_queue,
-                "text_completion_response_queue": tc_response_queue,
                 "graph_rag_request_queue": gr_request_queue,
                 "graph_rag_response_queue": gr_response_queue,
             }
@@ -169,13 +159,6 @@ class Processor(ConsumerProducer):
             pulsar_host = self.pulsar_host
         )
 
-        self.llm = LlmClient(
-            subscriber=subscriber,
-            input_queue=text_completion_request_queue,
-            output_queue=text_completion_response_queue,
-            pulsar_host = self.pulsar_host
-        )
-
         self.graph_rag = GraphRagClient(
             subscriber=subscriber,
             input_queue=graph_rag_request_queue,
@@ -340,18 +323,6 @@ class Processor(ConsumerProducer):
             help=f'Prompt response queue (default: {pr_response_queue})',
         )
 
-        parser.add_argument(
-            '--text-completion-request-queue',
-            default=tc_request_queue,
-            help=f'Text completion request queue (default: {tc_request_queue})',
-        )
-
-        parser.add_argument(
-            '--text-completion-response-queue',
-            default=tc_response_queue,
-            help=f'Text completion response queue (default: {tc_response_queue})',
-        )
-
         parser.add_argument(
             '--graph-rag-request-queue',
             default=gr_request_queue,

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/agent/react/tools.py

@@ -14,6 +14,6 @@ class TextCompletionImpl:
         self.context = context
     def invoke(self, **arguments):
         return self.context.prompt.request(
-            "question", { "question": arguments.get("computation") }
+            "question", { "question": arguments.get("question") }
         )
 

trustgraph-flow-0.20.2/trustgraph/flow_version.py

@@ -0,0 +1 @@
+__version__ = "0.20.2"

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/model/text_completion/azure/llm.py

@@ -158,25 +158,15 @@ class Processor(ConsumerProducer):
 
         except TooManyRequests:
 
-            print("Send rate limit response...", flush=True)
+            print("Rate limit...")
 
-            r = TextCompletionResponse(
-                error=Error(
-                    type = "rate-limit",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
-            )
-
-            self.producer.send(r, properties={"id": id})
-
-            self.consumer.acknowledge(msg)
+            # Leave rate limit retries to the base handler
+            raise TooManyRequests()
 
         except Exception as e:
 
+            # Apart from rate limits, treat all exceptions as unrecoverable
+
             print(f"Exception: {e}")
 
             print("Send error response...", flush=True)

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/model/text_completion/azure_openai/llm.py

@@ -4,10 +4,9 @@ Simple LLM service, performs text prompt completion using the Azure
 OpenAI endpoit service.  Input is prompt, output is response.
 """
 
-import requests
 import json
 from prometheus_client import Histogram
-from openai import AzureOpenAI
+from openai import AzureOpenAI, RateLimitError
 import os
 
 from .... schema import TextCompletionRequest, TextCompletionResponse, Error
@@ -126,30 +125,27 @@ class Processor(ConsumerProducer):
             print(f"Output Tokens: {outputtokens}", flush=True)
             print("Send response...", flush=True)
 
-            r = TextCompletionResponse(response=resp.choices[0].message.content, error=None, in_token=inputtokens, out_token=outputtokens, model=self.model)
-            self.producer.send(r, properties={"id": id})
-
-        except TooManyRequests:
-
-            print("Send rate limit response...", flush=True)
-
             r = TextCompletionResponse(
-                error=Error(
-                    type = "rate-limit",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
+                response=resp.choices[0].message.content,
+                error=None,
+                in_token=inputtokens,
+                out_token=outputtokens,
+                model=self.model
             )
 
             self.producer.send(r, properties={"id": id})
 
-            self.consumer.acknowledge(msg)
+        except RateLimitError:
+
+            print("Send rate limit response...", flush=True)
+
+            # Leave rate limit retries to the base handler
+            raise TooManyRequests()
 
         except Exception as e:
 
+            # Apart from rate limits, treat all exceptions as unrecoverable
+
             print(f"Exception: {e}")
 
             print("Send error response...", flush=True)

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/model/text_completion/claude/llm.py

@@ -87,8 +87,6 @@ class Processor(ConsumerProducer):
 
         try:
 
-            # FIXME: Rate limits?
-
             with __class__.text_completion_metric.time():
 
                 response = message = self.claude.messages.create(
@@ -117,34 +115,26 @@ class Processor(ConsumerProducer):
             print(f"Output Tokens: {outputtokens}", flush=True)
 
             print("Send response...", flush=True)
-            r = TextCompletionResponse(response=resp, error=None, in_token=inputtokens, out_token=outputtokens, model=self.model)
+            r = TextCompletionResponse(
+                response=resp,
+                error=None,
+                in_token=inputtokens,
+                out_token=outputtokens,
+                model=self.model
+            )
             self.send(r, properties={"id": id})
 
             print("Done.", flush=True)
 
-        # FIXME: Wrong exception, don't know what this LLM throws
-        # for a rate limit
-        except TooManyRequests:
-
-            print("Send rate limit response...", flush=True)
-
-            r = TextCompletionResponse(
-                error=Error(
-                    type = "rate-limit",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
-            )
-
-            self.producer.send(r, properties={"id": id})
+        except anthropic.RateLimitError:
 
-            self.consumer.acknowledge(msg)
+            # Leave rate limit retries to the base handler
+            raise TooManyRequests()
 
         except Exception as e:
 
+            # Apart from rate limits, treat all exceptions as unrecoverable
+
             print(f"Exception: {e}")
 
             print("Send error response...", flush=True)

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/model/text_completion/cohere/llm.py

@@ -112,27 +112,15 @@ class Processor(ConsumerProducer):
 
         # FIXME: Wrong exception, don't know what this LLM throws
         # for a rate limit
-        except TooManyRequests:
+        except cohere.TooManyRequestsError:
 
-            print("Send rate limit response...", flush=True)
-
-            r = TextCompletionResponse(
-                error=Error(
-                    type = "rate-limit",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
-            )
-
-            self.producer.send(r, properties={"id": id})
-
-            self.consumer.acknowledge(msg)
+            # Leave rate limit retries to the base handler
+            raise TooManyRequests()
 
         except Exception as e:
 
+            # Apart from rate limits, treat all exceptions as unrecoverable
+
             print(f"Exception: {e}")
 
             print("Send error response...", flush=True)

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/model/text_completion/googleaistudio/llm.py

@@ -88,7 +88,8 @@ class Processor(ConsumerProducer):
             HarmCategory.HARM_CATEGORY_HARASSMENT: block_level,
             HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: block_level,
             HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: block_level,
-            # There is a documentation conflict on whether or not CIVIC_INTEGRITY is a valid category
+            # There is a documentation conflict on whether or not
+            # CIVIC_INTEGRITY is a valid category
             # HarmCategory.HARM_CATEGORY_CIVIC_INTEGRITY: block_level,
         }
 
@@ -122,8 +123,6 @@ class Processor(ConsumerProducer):
 
         try:
 
-            # FIXME: Rate limits?
-
             with __class__.text_completion_metric.time():
 
                 chat_session = self.llm.start_chat(
@@ -140,35 +139,30 @@ class Processor(ConsumerProducer):
             print(f"Output Tokens: {outputtokens}", flush=True)
 
             print("Send response...", flush=True)
-            r = TextCompletionResponse(response=resp, error=None, in_token=inputtokens, out_token=outputtokens, model=self.model)
+            r = TextCompletionResponse(
+                response=resp,
+                error=None,
+                in_token=inputtokens,
+                out_token=outputtokens,
+                model=self.model
+            )
             self.send(r, properties={"id": id})
 
             print("Done.", flush=True)
 
-        # FIXME: Wrong exception, don't know what this LLM throws
-        # for a rate limit
         except ResourceExhausted as e:
 
-            print("Send rate limit response...", flush=True)
+            print("Hit rate limit:", e, flush=True)
 
-            r = TextCompletionResponse(
-                error=Error(
-                    type = "rate-limit",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
-            )
-
-            self.producer.send(r, properties={"id": id})
-
-            self.consumer.acknowledge(msg)
+            # Leave rate limit retries to the default handler
+            raise TooManyRequests()
 
         except Exception as e:
 
-            print(f"Exception: {e}")
+            # Apart from rate limits, treat all exceptions as unrecoverable
+
+            print(type(e), flush=True)
+            print(f"Exception: {e}", flush=True)
 
             print("Send error response...", flush=True)
 

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/model/text_completion/llamafile/llm.py

@@ -126,26 +126,7 @@ class Processor(ConsumerProducer):
 
             print("Done.", flush=True)
 
-        # FIXME: Wrong exception, don't know what this LLM throws
-        # for a rate limit
-        except TooManyRequests:
-
-            print("Send rate limit response...", flush=True)
-
-            r = TextCompletionResponse(
-                error=Error(
-                    type = "rate-limit",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
-            )
-
-            self.producer.send(r, properties={"id": id})
-
-            self.consumer.acknowledge(msg)
+        # SLM, presumably there aren't rate limits
 
         except Exception as e:
 

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/model/text_completion/ollama/llm.py

@@ -100,26 +100,7 @@ class Processor(ConsumerProducer):
 
             print("Done.", flush=True)
 
-        # FIXME: Wrong exception, don't know what this LLM throws
-        # for a rate limit
-        except TooManyRequests:
-
-            print("Send rate limit response...", flush=True)
-
-            r = TextCompletionResponse(
-                error=Error(
-                    type = "rate-limit",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
-            )
-
-            self.producer.send(r, properties={"id": id})
-
-            self.consumer.acknowledge(msg)
+        # SLM, presumably no rate limits
 
         except Exception as e:
 

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph/model/text_completion/openai/llm.py

@@ -4,7 +4,7 @@ Simple LLM service, performs text prompt completion using OpenAI.
 Input is prompt, output is response.
 """
 
-from openai import OpenAI
+from openai import OpenAI, RateLimitError
 from prometheus_client import Histogram
 import os
 
@@ -87,8 +87,6 @@ class Processor(ConsumerProducer):
 
         try:
 
-            # FIXME: Rate limits
-
             with __class__.text_completion_metric.time():
 
                 resp = self.openai.chat.completions.create(
@@ -134,27 +132,15 @@ class Processor(ConsumerProducer):
 
         # FIXME: Wrong exception, don't know what this LLM throws
         # for a rate limit
-        except TooManyRequests:
-
-            print("Send rate limit response...", flush=True)
-
-            r = TextCompletionResponse(
-                error=Error(
-                    type = "rate-limit",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
-            )
+        except openai.RateLimitError:
 
-            self.producer.send(r, properties={"id": id})
-
-            self.consumer.acknowledge(msg)
+            # Leave rate limit retries to the base handler
+            raise TooManyRequests()
 
         except Exception as e:
 
+            # Apart from rate limits, treat all exceptions as unrecoverable
+
             print(f"Exception: {e}")
 
             print("Send error response...", flush=True)

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph_flow.egg-info/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: trustgraph-flow
-Version: 0.20.0
+Version: 0.20.2
 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
-Download-URL: https://github.com/trustgraph-ai/trustgraph/archive/refs/tags/v0.20.0.tar.gz
+Download-URL: https://github.com/trustgraph-ai/trustgraph/archive/refs/tags/v0.20.2.tar.gz
 Author: trustgraph.ai
 Author-email: security@trustgraph.ai
 Classifier: Programming Language :: Python :: 3

{trustgraph-flow-0.20.0 → trustgraph-flow-0.20.2}/trustgraph_flow.egg-info/requires.txt

@@ -2,6 +2,7 @@ aiohttp
 anthropic
 cassandra-driver
 cohere
+cryptography
 falkordb
 google-generativeai
 ibis

trustgraph-flow-0.20.0/trustgraph/flow_version.py

@@ -1 +0,0 @@
-__version__ = "0.20.0"