apunkt/trustgraph
1
0
Fork 0
mirror of https://github.com/trustgraph-ai/trustgraph.git synced 2026-04-27 01:16:22 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

release/v1.4 -> master (#548)

Browse source
This commit is contained in:
cybermaggedon 2025-10-06 17:54:26 +01:00 • committed by GitHub
parent 3ec2cd54f9
commit 2bd68ed7f4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
94 changed files with 8571 additions and 1740 deletions
Download patch file Download diff file Expand all files Collapse all files

1
trustgraph-base/pyproject.toml
View file

@ -12,6 +12,7 @@ requires-python = ">=3.8"
dependencies = [
"pulsar-client",
"prometheus-client",
"requests",
]
classifiers = [
"Programming Language :: Python :: 3",

5
trustgraph-base/trustgraph/api/flow.py
View file

@ -87,7 +87,7 @@ class Flow:
return json.loads(self.request(request = input)["flow"])
def start(self, class_name, id, description):
def start(self, class_name, id, description, parameters=None):
# The input consists of system and prompt strings
input = {
@ -97,6 +97,9 @@ class Flow:
"description": description,
}
if parameters:
input["parameters"] = parameters
self.request(request = input)
def stop(self, id):

3
trustgraph-base/trustgraph/base/__init__.py
View file

@ -8,11 +8,12 @@ from . subscriber import Subscriber
from . metrics import ProcessorMetrics, ConsumerMetrics, ProducerMetrics
from . flow_processor import FlowProcessor
from . consumer_spec import ConsumerSpec
from . setting_spec import SettingSpec
from . parameter_spec import ParameterSpec
from . producer_spec import ProducerSpec
from . subscriber_spec import SubscriberSpec
from . request_response_spec import RequestResponseSpec
from . llm_service import LlmService, LlmResult
from . chunking_service import ChunkingService
from . embeddings_service import EmbeddingsService
from . embeddings_client import EmbeddingsClientSpec
from . text_completion_client import TextCompletionClientSpec

62
trustgraph-base/trustgraph/base/chunking_service.py Normal file
View file

@ -0,0 +1,62 @@
"""
Base chunking service that provides parameter specification functionality
for chunk-size and chunk-overlap parameters
"""
import logging
from .flow_processor import FlowProcessor
from .parameter_spec import ParameterSpec
# Module logger
logger = logging.getLogger(__name__)
class ChunkingService(FlowProcessor):
"""Base service for chunking processors with parameter specification support"""
def __init__(self, **params):
# Call parent constructor
super(ChunkingService, self).__init__(**params)
# Register parameter specifications for chunk-size and chunk-overlap
self.register_specification(
ParameterSpec(name="chunk-size")
)
self.register_specification(
ParameterSpec(name="chunk-overlap")
)
logger.debug("ChunkingService initialized with parameter specifications")
async def chunk_document(self, msg, consumer, flow, default_chunk_size, default_chunk_overlap):
"""
Extract chunk parameters from flow and return effective values
Args:
msg: The message containing the document to chunk
consumer: The consumer spec
flow: The flow context
default_chunk_size: Default chunk size from processor config
default_chunk_overlap: Default chunk overlap from processor config
Returns:
tuple: (chunk_size, chunk_overlap) - effective values to use
"""
# Extract parameters from flow (flow-configurable parameters)
chunk_size = flow("chunk-size")
chunk_overlap = flow("chunk-overlap")
# Use provided values or fall back to defaults
effective_chunk_size = chunk_size if chunk_size is not None else default_chunk_size
effective_chunk_overlap = chunk_overlap if chunk_overlap is not None else default_chunk_overlap
logger.debug(f"Using chunk-size: {effective_chunk_size}")
logger.debug(f"Using chunk-overlap: {effective_chunk_overlap}")
return effective_chunk_size, effective_chunk_overlap
@staticmethod
def add_args(parser):
"""Add chunking service arguments to parser"""
FlowProcessor.add_args(parser)

4
trustgraph-base/trustgraph/base/flow.py
View file

@ -12,7 +12,7 @@ class Flow:
# Consumers and publishers. Is this a bit untidy?
self.consumer = {}
self.setting = {}
self.parameter = {}
for spec in processor.specifications:
spec.add(self, processor, defn)
@ -28,5 +28,5 @@ class Flow:
def __call__(self, key):
if key in self.producer: return self.producer[key]
if key in self.consumer: return self.consumer[key]
if key in self.setting: return self.setting[key].value
if key in self.parameter: return self.parameter[key].value
return None

2
trustgraph-base/trustgraph/base/flow_processor.py
View file

@ -35,7 +35,7 @@ class FlowProcessor(AsyncProcessor):
# These can be overriden by a derived class:
# Array of specifications: ConsumerSpec, ProducerSpec, SettingSpec
# Array of specifications: ConsumerSpec, ProducerSpec, ParameterSpec
self.specifications = []
logger.info("Service initialised.")

38
trustgraph-base/trustgraph/base/llm_service.py
View file

@ -5,11 +5,11 @@ LLM text completion base class
import time
import logging
from prometheus_client import Histogram
from prometheus_client import Histogram, Info
from .. schema import TextCompletionRequest, TextCompletionResponse, Error
from .. exceptions import TooManyRequests
from .. base import FlowProcessor, ConsumerSpec, ProducerSpec
from .. base import FlowProcessor, ConsumerSpec, ProducerSpec, ParameterSpec
# Module logger
logger = logging.getLogger(__name__)
@ -32,7 +32,7 @@ class LlmService(FlowProcessor):
def __init__(self, **params):
id = params.get("id")
id = params.get("id", default_ident)
concurrency = params.get("concurrency", 1)
super(LlmService, self).__init__(**params | {
@ -56,6 +56,18 @@ class LlmService(FlowProcessor):
)
)
self.register_specification(
ParameterSpec(
name = "model",
)
)
self.register_specification(
ParameterSpec(
name = "temperature",
)
)
if not hasattr(__class__, "text_completion_metric"):
__class__.text_completion_metric = Histogram(
'text_completion_duration',
@ -70,6 +82,13 @@ class LlmService(FlowProcessor):
]
)
if not hasattr(__class__, "text_completion_model_metric"):
__class__.text_completion_model_metric = Info(
'text_completion_model',
'Text completion model',
["processor", "flow"]
)
async def on_request(self, msg, consumer, flow):
try:
@ -85,10 +104,21 @@ class LlmService(FlowProcessor):
flow=f"{flow.name}-{consumer.name}",
).time():
model = flow("model")
temperature = flow("temperature")
response = await self.generate_content(
request.system, request.prompt
request.system, request.prompt, model, temperature
)
__class__.text_completion_model_metric.labels(
processor = self.id,
flow = flow.name
).info({
"model": str(model) if model is not None else "",
"temperature": str(temperature) if temperature is not None else "",
})
await flow("response").send(
TextCompletionResponse(
error=None,

8
trustgraph-base/trustgraph/base/setting_spec.py → trustgraph-base/trustgraph/base/parameter_spec.py
View file

@ -1,7 +1,7 @@
from . spec import Spec
class Setting:
class Parameter:
def __init__(self, value):
self.value = value
async def start():
@ -9,11 +9,13 @@ class Setting:
async def stop():
pass
class SettingSpec(Spec):
class ParameterSpec(Spec):
def __init__(self, name):
self.name = name
def add(self, flow, processor, definition):
flow.config[self.name] = Setting(definition[self.name])
value = definition.get(self.name, None)
flow.parameter[self.name] = Parameter(value)

4
trustgraph-base/trustgraph/base/request_response_spec.py
View file

@ -49,8 +49,6 @@ class RequestResponse(Subscriber):
id = str(uuid.uuid4())
logger.debug(f"Sending request {id}...")
q = await self.subscribe(id)
try:
@ -75,8 +73,6 @@ class RequestResponse(Subscriber):
timeout=timeout
)
logger.debug("Received response")
if recipient is None:
# If no recipient handler, just return the first

15
trustgraph-base/trustgraph/messaging/translators/flow.py
View file

@ -12,12 +12,13 @@ class FlowRequestTranslator(MessageTranslator):
class_name=data.get("class-name"),
class_definition=data.get("class-definition"),
description=data.get("description"),
flow_id=data.get("flow-id")
flow_id=data.get("flow-id"),
parameters=data.get("parameters")
)
def from_pulsar(self, obj: FlowRequest) -> Dict[str, Any]:
result = {}
if obj.operation is not None:
result["operation"] = obj.operation
if obj.class_name is not None:
@ -28,7 +29,9 @@ class FlowRequestTranslator(MessageTranslator):
result["description"] = obj.description
if obj.flow_id is not None:
result["flow-id"] = obj.flow_id
if obj.parameters is not None:
result["parameters"] = obj.parameters
return result
@ -40,7 +43,7 @@ class FlowResponseTranslator(MessageTranslator):
def from_pulsar(self, obj: FlowResponse) -> Dict[str, Any]:
result = {}
if obj.class_names is not None:
result["class-names"] = obj.class_names
if obj.flow_ids is not None:
@ -51,7 +54,9 @@ class FlowResponseTranslator(MessageTranslator):
result["flow"] = obj.flow
if obj.description is not None:
result["description"] = obj.description
if obj.parameters is not None:
result["parameters"] = obj.parameters
return result
def from_response_with_completion(self, obj: FlowResponse) -> Tuple[Dict[str, Any], bool]:

6
trustgraph-base/trustgraph/schema/services/flow.py
View file

@ -35,6 +35,9 @@ class FlowRequest(Record):
# get_flow, start_flow, stop_flow
flow_id = String()
# start_flow - optional parameters for flow customization
parameters = Map(String())
class FlowResponse(Record):
# list_classes
@ -52,6 +55,9 @@ class FlowResponse(Record):
# get_flow
description = String()
# get_flow - parameters used when flow was started
parameters = Map(String())
# Everything
error = Error()