trustgraph-flow 0.23.1__tar.gz → 0.23.3__tar.gz

{trustgraph-flow-0.23.1 → trustgraph-flow-0.23.3}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: trustgraph-flow
-Version: 0.23.1
+Version: 0.23.3
 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/v0.23.1.tar.gz
+Download-URL: https://github.com/trustgraph-ai/trustgraph/archive/refs/tags/v0.23.3.tar.gz
 Author: trustgraph.ai
 Author-email: security@trustgraph.ai
 Classifier: Programming Language :: Python :: 3

{trustgraph-flow-0.23.1 → trustgraph-flow-0.23.3}/trustgraph/config/service/flow.py

@@ -27,6 +27,8 @@ class FlowConfig:
 
         self.config["flow-classes"][msg.class_name] = msg.class_definition
 
+        self.config.version += 1
+
         await self.config.push()
 
         return FlowResponse(
@@ -39,6 +41,8 @@ class FlowConfig:
 
         del self.config["flow-classes"][msg.class_name]
 
+        self.config.version += 1
+
         await self.config.push()
 
         return FlowResponse(
@@ -135,6 +139,8 @@ class FlowConfig:
             "interfaces": interfaces,
         })
 
+        self.config.version += 1
+
         await self.config.push()
 
         return FlowResponse(
@@ -186,6 +192,8 @@ class FlowConfig:
         if msg.flow_id in self.config["flows"]:
             del self.config["flows"][msg.flow_id]
 
+        self.config.version += 1
+
         await self.config.push()
 
         return FlowResponse(

{trustgraph-flow-0.23.1 → trustgraph-flow-0.23.3}/trustgraph/decoding/pdf/pdf_decoder.py

@@ -9,7 +9,6 @@ import base64
 from langchain_community.document_loaders import PyPDFLoader
 
 from ... schema import Document, TextDocument, Metadata
-from ... log_level import LogLevel
 from ... base import FlowProcessor, ConsumerSpec, ProducerSpec
 
 default_ident = "pdf-decoder"

{trustgraph-flow-0.23.1 → trustgraph-flow-0.23.3}/trustgraph/extract/kg/relationships/extract.py

@@ -173,7 +173,7 @@ class Processor(FlowProcessor):
                         o=Value(value=v.metadata.id, is_uri=True)
                     ))
 
-            await self.emit_edges(
+            await self.emit_triples(
                 flow("triples"),
                 Metadata(
                     id=v.metadata.id,

trustgraph-flow-0.23.3/trustgraph/flow_version.py

@@ -0,0 +1 @@
+__version__ = "0.23.3"

{trustgraph-flow-0.23.1 → trustgraph-flow-0.23.3}/trustgraph/metering/counter.py

@@ -3,22 +3,19 @@ Simple token counter for each LLM response.
 """
 
 from prometheus_client import Counter
-from . pricelist import price_list
+import json
 
 from .. schema import TextCompletionResponse, Error
-from .. schema import text_completion_response_queue
-from .. log_level import LogLevel
-from .. base import Consumer
+from .. base import FlowProcessor, ConsumerSpec
 
-module = "metering"
+default_ident = "metering"
 
-default_input_queue = text_completion_response_queue
-default_subscriber = module
-
-class Processor(Consumer):
+class Processor(FlowProcessor):
 
     def __init__(self, **params):
 
+        id = params.get("id", default_ident)
+
         if not hasattr(__class__, "input_token_metric"):
             __class__.input_token_metric = Counter(
                 'input_tokens', 'Input token count'
@@ -39,40 +36,61 @@ class Processor(Consumer):
                 'output_cost', 'Output cost'
             )
 
-        input_queue = params.get("input_queue", default_input_queue)
-        subscriber = params.get("subscriber", default_subscriber)
-
         super(Processor, self).__init__(
             **params | {
-                "input_queue": input_queue,
-                "subscriber": subscriber,
-                "input_schema": TextCompletionResponse,
+                "id": id,
             }
         )
 
-    def get_prices(self, prices, modelname):
-        for model in prices["price_list"]:
-            if model["model_name"] == modelname:
-                return model["input_price"], model["output_price"]
-        return None, None  # Return None if model is not found
+        self.register_config_handler(self.on_cost_config)
+
+        self.register_specification(
+            ConsumerSpec(
+                name = "input",
+                schema = TextCompletionResponse,
+                handler = self.on_message,
+            )
+        )
 
-    async def handle(self, msg):
+        self.prices = {}
 
-        v = msg.value()
-        modelname = v.model
+        self.config_key = "token-costs"
+
+    # Load token costs from the config service
+    async def on_cost_config(self, config, version):
 
-        # Sender-produced ID
-        id = msg.properties()["id"]
+        print("Loading configuration version", version)
 
-        print(f"Handling response {id}...", flush=True)
+        if self.config_key not in config:
+            print(f"No key {self.config_key} in config", flush=True)
+            return
 
+        config = config[self.config_key]
+
+        self.prices = {
+            k: json.loads(v)
+            for k, v in config.items()
+        }
+
+    def get_prices(self, modelname):
+
+        if modelname in self.prices:
+            model = self.prices[modelname]
+            return model["input_price"], model["output_price"]
+        return None, None  # Return None if model is not found
+
+    async def on_message(self, msg, consumer, flow):
+
+        v = msg.value()
+
+        modelname = v.model
         num_in = v.in_token
         num_out = v.out_token
 
         __class__.input_token_metric.inc(num_in)
         __class__.output_token_metric.inc(num_out)
 
-        model_input_price, model_output_price = self.get_prices(price_list, modelname)
+        model_input_price, model_output_price = self.get_prices(modelname)
 
         if model_input_price == None:
             cost_per_call = f"Model Not Found in Price list"
@@ -91,10 +109,8 @@ class Processor(Consumer):
     @staticmethod
     def add_args(parser):
 
-        Consumer.add_args(
-            parser, default_input_queue, default_subscriber,
-        )
+        FlowProcessor.add_args(parser)
 
 def run():
 
-    Processor.launch(module, __doc__)
+    Processor.launch(default_ident, __doc__)

{trustgraph-flow-0.23.1 → trustgraph-flow-0.23.3}/trustgraph/model/text_completion/azure/llm.py

@@ -9,31 +9,21 @@ import json
 from prometheus_client import Histogram
 import os
 
-from .... schema import TextCompletionRequest, TextCompletionResponse, Error
-from .... schema import text_completion_request_queue
-from .... schema import text_completion_response_queue
-from .... log_level import LogLevel
-from .... base import ConsumerProducer
 from .... exceptions import TooManyRequests
+from .... base import LlmService, LlmResult
 
-module = "text-completion"
+default_ident = "text-completion"
 
-default_input_queue = text_completion_request_queue
-default_output_queue = text_completion_response_queue
-default_subscriber = module
 default_temperature = 0.0
 default_max_output = 4192
 default_model = "AzureAI"
 default_endpoint = os.getenv("AZURE_ENDPOINT")
 default_token = os.getenv("AZURE_TOKEN")
 
-class Processor(ConsumerProducer):
+class Processor(LlmService):
 
     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)
         endpoint = params.get("endpoint", default_endpoint)
         token = params.get("token", default_token)
         temperature = params.get("temperature", default_temperature)
@@ -48,30 +38,13 @@ class Processor(ConsumerProducer):
 
         super(Processor, self).__init__(
             **params | {
-                "input_queue": input_queue,
-                "output_queue": output_queue,
-                "subscriber": subscriber,
-                "input_schema": TextCompletionRequest,
-                "output_schema": TextCompletionResponse,
+                "endpoint": endpoint,
                 "temperature": temperature,
                 "max_output": max_output,
                 "model": model,
             }
         )
 
-        if not hasattr(__class__, "text_completion_metric"):
-            __class__.text_completion_metric = Histogram(
-                'text_completion_duration',
-                'Text completion duration (seconds)',
-                buckets=[
-                    0.25, 0.5, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0,
-                    8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0,
-                    17.0, 18.0, 19.0, 20.0, 21.0, 22.0, 23.0, 24.0, 25.0,
-                    30.0, 35.0, 40.0, 45.0, 50.0, 60.0, 80.0, 100.0,
-                    120.0
-                ]
-            )
-
         self.endpoint = endpoint
         self.token = token
         self.temperature = temperature
@@ -123,25 +96,16 @@ class Processor(ConsumerProducer):
 
         return result
 
-    async def handle(self, msg):
-
-        v = msg.value()
-
-        # Sender-produced ID
-
-        id = msg.properties()["id"]
-
-        print(f"Handling prompt {id}...", flush=True)
+    async def generate_content(self, system, prompt):
 
         try:
 
             prompt = self.build_prompt(
-                v.system,
-                v.prompt
+                system,
+                prompt
             )
 
-            with __class__.text_completion_metric.time():
-                response = self.call_llm(prompt)
+            response = self.call_llm(prompt)
 
             resp = response['choices'][0]['message']['content']
             inputtokens = response['usage']['prompt_tokens']
@@ -153,8 +117,14 @@ class Processor(ConsumerProducer):
 
             print("Send response...", flush=True)
 
-            r = TextCompletionResponse(response=resp, error=None, in_token=inputtokens, out_token=outputtokens, model=self.model)
-            await self.send(r, properties={"id": id})
+            resp = LlmResult(
+                text = resp,
+                in_token = inputtokens,
+                out_token = outputtokens,
+                model = self.model
+            )
+
+            return resp
 
         except TooManyRequests:
 
@@ -168,33 +138,14 @@ class Processor(ConsumerProducer):
             # Apart from rate limits, treat all exceptions as unrecoverable
 
             print(f"Exception: {e}")
-
-            print("Send error response...", flush=True)
-
-            r = TextCompletionResponse(
-                error=Error(
-                    type = "llm-error",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
-            )
-
-            await self.send(r, properties={"id": id})
-
-            self.consumer.acknowledge(msg)
+            raise e
 
         print("Done.", flush=True)
 
     @staticmethod
     def add_args(parser):
 
-        ConsumerProducer.add_args(
-            parser, default_input_queue, default_subscriber,
-            default_output_queue,
-        )
+        LlmService.add_args(parser)
 
         parser.add_argument(
             '-e', '--endpoint',
@@ -224,4 +175,4 @@ class Processor(ConsumerProducer):
 
 def run():
     
-    Processor.launch(module, __doc__)
+    Processor.launch(default_ident, __doc__)

{trustgraph-flow-0.23.1 → trustgraph-flow-0.23.3}/trustgraph/model/text_completion/azure_openai/llm.py

@@ -9,18 +9,11 @@ from prometheus_client import Histogram
 from openai import AzureOpenAI, RateLimitError
 import os
 
-from .... schema import TextCompletionRequest, TextCompletionResponse, Error
-from .... schema import text_completion_request_queue
-from .... schema import text_completion_response_queue
-from .... log_level import LogLevel
-from .... base import ConsumerProducer
 from .... exceptions import TooManyRequests
+from .... base import LlmService, LlmResult
 
-module = "text-completion"
+default_ident = "text-completion"
 
-default_input_queue = text_completion_request_queue
-default_output_queue = text_completion_response_queue
-default_subscriber = module
 default_temperature = 0.0
 default_max_output = 4192
 default_api = "2024-12-01-preview"
@@ -28,13 +21,10 @@ default_endpoint = os.getenv("AZURE_ENDPOINT", None)
 default_token = os.getenv("AZURE_TOKEN", None)
 default_model = os.getenv("AZURE_MODEL", None)
 
-class Processor(ConsumerProducer):
+class Processor(LlmService):
 
     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)
         temperature = params.get("temperature", default_temperature)
         max_output = params.get("max_output", default_max_output)
 
@@ -51,11 +41,6 @@ class Processor(ConsumerProducer):
 
         super(Processor, self).__init__(
             **params | {
-                "input_queue": input_queue,
-                "output_queue": output_queue,
-                "subscriber": subscriber,
-                "input_schema": TextCompletionRequest,
-                "output_schema": TextCompletionResponse,
                 "temperature": temperature,
                 "max_output": max_output,
                 "model": model,
@@ -63,19 +48,6 @@ class Processor(ConsumerProducer):
             }
         )
 
-        if not hasattr(__class__, "text_completion_metric"):
-            __class__.text_completion_metric = Histogram(
-                'text_completion_duration',
-                'Text completion duration (seconds)',
-                buckets=[
-                    0.25, 0.5, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0,
-                    8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0,
-                    17.0, 18.0, 19.0, 20.0, 21.0, 22.0, 23.0, 24.0, 25.0,
-                    30.0, 35.0, 40.0, 45.0, 50.0, 60.0, 80.0, 100.0,
-                    120.0
-                ]
-            )
-
         self.temperature = temperature
         self.max_output = max_output
         self.model = model
@@ -84,41 +56,31 @@ class Processor(ConsumerProducer):
             api_key=token,  
             api_version=api,
             azure_endpoint = endpoint,
-            )
-
-    async def handle(self, msg):
-
-        v = msg.value()
-
-        # Sender-produced ID
-
-        id = msg.properties()["id"]
-
-        print(f"Handling prompt {id}...", flush=True)
+        )
 
-        prompt = v.system + "\n\n" + v.prompt
+    async def generate_content(self, system, prompt):
 
+        prompt = system + "\n\n" + prompt
 
         try:
 
-            with __class__.text_completion_metric.time():
-                resp = self.openai.chat.completions.create(
-                    model=self.model,
-                    messages=[
-                        {
-                            "role": "user",
-                            "content": [
-                                {
-                                    "type": "text",
-                                    "text": prompt
-                                }
-                            ]
-                        }
-                    ],
-                    temperature=self.temperature,
-                    max_tokens=self.max_output,
-                    top_p=1,
-                )
+            resp = self.openai.chat.completions.create(
+                model=self.model,
+                messages=[
+                    {
+                        "role": "user",
+                        "content": [
+                            {
+                                "type": "text",
+                                "text": prompt
+                            }
+                        ]
+                    }
+                ],
+                temperature=self.temperature,
+                max_tokens=self.max_output,
+                top_p=1,
+            )
 
             inputtokens = resp.usage.prompt_tokens
             outputtokens = resp.usage.completion_tokens
@@ -127,15 +89,14 @@ 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
+            r = LlmResult(
+                text = resp.choices[0].message.content,
+                in_token = inputtokens,
+                out_token = outputtokens,
+                model = self.model
             )
 
-            await self.send(r, properties={"id": id})
+            return r
 
         except RateLimitError:
 
@@ -147,35 +108,15 @@ class Processor(ConsumerProducer):
         except Exception as e:
 
             # Apart from rate limits, treat all exceptions as unrecoverable
-
             print(f"Exception: {e}")
-
-            print("Send error response...", flush=True)
-
-            r = TextCompletionResponse(
-                error=Error(
-                    type = "llm-error",
-                    message = str(e),
-                ),
-                response=None,
-                in_token=None,
-                out_token=None,
-                model=None,
-            )
-
-            await self.send(r, properties={"id": id})
-
-            self.consumer.acknowledge(msg)
+            raise e
 
         print("Done.", flush=True)
 
     @staticmethod
     def add_args(parser):
 
-        ConsumerProducer.add_args(
-            parser, default_input_queue, default_subscriber,
-            default_output_queue,
-        )
+        LlmService.add_args(parser)
 
         parser.add_argument(
             '-e', '--endpoint',
@@ -217,4 +158,4 @@ class Processor(ConsumerProducer):
 
 def run():
     
-    Processor.launch(module, __doc__)
+    Processor.launch(default_ident, __doc__)

trustgraph-flow-0.23.3/trustgraph/model/text_completion/claude/llm.py

@@ -0,0 +1,130 @@
+
+"""
+Simple LLM service, performs text prompt completion using Claude.
+Input is prompt, output is response.
+"""
+
+import anthropic
+import os
+
+from .... exceptions import TooManyRequests
+from .... base import LlmService, LlmResult
+
+default_ident = "text-completion"
+
+default_model = 'claude-3-5-sonnet-20240620'
+default_temperature = 0.0
+default_max_output = 8192
+default_api_key = os.getenv("CLAUDE_KEY")
+
+class Processor(LlmService):
+
+    def __init__(self, **params):
+    
+        model = params.get("model", default_model)
+        api_key = params.get("api_key", default_api_key)
+        temperature = params.get("temperature", default_temperature)
+        max_output = params.get("max_output", default_max_output)
+
+        if api_key is None:
+            raise RuntimeError("Claude API key not specified")
+
+        super(Processor, self).__init__(
+            **params | {
+                "model": model,
+                "temperature": temperature,
+                "max_output": max_output,
+            }
+        )
+
+        self.model = model
+        self.claude = anthropic.Anthropic(api_key=api_key)
+        self.temperature = temperature
+        self.max_output = max_output
+
+        print("Initialised", flush=True)
+
+    async def generate_content(self, system, prompt):
+
+        try:
+
+            response = message = self.claude.messages.create(
+                model=self.model,
+                max_tokens=self.max_output,
+                temperature=self.temperature,
+                system = system,
+                messages=[
+                    {
+                        "role": "user",
+                        "content": [
+                            {
+                                "type": "text",
+                                "text": prompt
+                            }
+                        ]
+                    }
+                ]
+            )
+
+            resp = response.content[0].text
+            inputtokens = response.usage.input_tokens
+            outputtokens = response.usage.output_tokens
+            print(resp, flush=True)
+            print(f"Input Tokens: {inputtokens}", flush=True)
+            print(f"Output Tokens: {outputtokens}", flush=True)
+
+            resp = LlmResult(
+                text = resp,
+                in_token = inputtokens,
+                out_token = outputtokens,
+                model = self.model
+            )
+
+            return resp
+
+        except anthropic.RateLimitError:
+
+            # 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}")
+            raise e
+
+    @staticmethod
+    def add_args(parser):
+
+        LlmService.add_args(parser)
+
+        parser.add_argument(
+            '-m', '--model',
+            default="claude-3-5-sonnet-20240620",
+            help=f'LLM model (default: claude-3-5-sonnet-20240620)'
+        )
+
+        parser.add_argument(
+            '-k', '--api-key',
+            default=default_api_key,
+            help=f'Claude API key'
+        )
+
+        parser.add_argument(
+            '-t', '--temperature',
+            type=float,
+            default=default_temperature,
+            help=f'LLM temperature parameter (default: {default_temperature})'
+        )
+
+        parser.add_argument(
+            '-x', '--max-output',
+            type=int,
+            default=default_max_output,
+            help=f'LLM max output tokens (default: {default_max_output})'
+        )
+
+def run():
+    
+    Processor.launch(default_ident, __doc__)