draft

Author: Andrei Bratu

poetry.lock

@@ -56,7 +56,7 @@ client.prompts.log(
     messages=[{"role": "user", "content": "What really happened at Roswell?"}],
     inputs={"person": "Trump"},
     created_at=datetime.datetime.fromisoformat(
-        "2024-07-19 00:29:35.178000+00:00",
+        "2024-07-18 23:29:35.178000+00:00",
     ),
     provider_latency=6.5931549072265625,
     output_message={
@@ -6598,10 +6598,10 @@ client.flows.log(
     output="The patient is likely experiencing a myocardial infarction. Immediate medical attention is required.",
     trace_status="incomplete",
     start_time=datetime.datetime.fromisoformat(
-        "2024-07-08 22:40:35+00:00",
+        "2024-07-08 21:40:35+00:00",
     ),
     end_time=datetime.datetime.fromisoformat(
-        "2024-07-08 22:40:39+00:00",
+        "2024-07-08 21:40:39+00:00",
     ),
 )
 

reference.md

@@ -10,16 +10,16 @@
 from opentelemetry.trace import Tracer
 
 from humanloop.core.client_wrapper import SyncClientWrapper
-from humanloop.decorators.types import DecoratorPromptKernelRequestParams
+from humanloop.utilities.types import DecoratorPromptKernelRequestParams
 from humanloop.eval_utils.context import EVALUATION_CONTEXT_VARIABLE_NAME, EvaluationContext
 
 from humanloop.eval_utils import log_with_evaluation_context, run_eval
 from humanloop.eval_utils.types import Dataset, Evaluator, EvaluatorCheck, File
 
 from humanloop.base_client import AsyncBaseHumanloop, BaseHumanloop
-from humanloop.decorators.flow import flow as flow_decorator_factory
-from humanloop.decorators.prompt import prompt as prompt_decorator_factory
-from humanloop.decorators.tool import tool as tool_decorator_factory
+from humanloop.utilities.flow import flow as flow_decorator_factory
+from humanloop.utilities.prompt import prompt as prompt_decorator_factory
+from humanloop.utilities.tool import tool as tool_decorator_factory
 from humanloop.environment import HumanloopEnvironment
 from humanloop.evaluations.client import EvaluationsClient
 from humanloop.otel import instrument_provider
@@ -49,7 +49,6 @@ def run(
         name: Optional[str],
         dataset: Dataset,
         evaluators: Optional[Sequence[Evaluator]] = None,
-        # logs: typing.Sequence[dict] | None = None,
         workers: int = 4,
     ) -> List[EvaluatorCheck]:
         """Evaluate your function for a given `Dataset` and set of `Evaluators`.
@@ -175,14 +174,14 @@ def prompt(
         path: Optional[str] = None,
         **prompt_kernel: Unpack[DecoratorPromptKernelRequestParams],  # type: ignore
     ):
-        """Decorator for declaring a (Prompt)[https://humanloop.com/docs/explanation/prompts] in code.
+        """Decorator for declaring a [Prompt](https://humanloop.com/docs/explanation/prompts) in code.
 
         The decorator intercepts calls to LLM provider APIs and creates
         a new Prompt file based on the hyperparameters used in the call.
         If a hyperparameter is specified in the `@prompt` decorator, then
         they override any value intercepted from the LLM provider call.
 
-        If the (Prompt)[https://humanloop.com/docs/explanation/prompts] already exists
+        If the [Prompt](https://humanloop.com/docs/explanation/prompts) already exists
         on the specified path, a new version will be upserted when any of the above change.
 
         Here's an example of declaring a (Prompt)[https://humanloop.com/docs/explanation/prompts] in code:
@@ -200,7 +199,7 @@ def call_llm(messages):
             ).choices[0].message.content
         ```
 
-        This will create a (Prompt)[https://humanloop.com/docs/explanation/prompts] with the following attributes:
+        This will create a [Prompt](https://humanloop.com/docs/explanation/prompts] with the following attributes:
 
         ```python
         {
@@ -342,9 +341,9 @@ def flow(
     ):
         """Decorator for declaring a [Flow](https://humanloop.com/docs/explanation/flows) in code.
 
-        A [Flow](https://humanloop.com/docs/explanation/flows) decorator should be added
-        at the entrypoint of your LLM feature. Call other functions decorated with
-        Humanloop SDK decorators to create a Trace of Logs on Humanloop.
+        A [Flow](https://humanloop.com/docs/explanation/flows) wrapped callable should
+        be used as the entrypoint of your LLM feature. Call other functions wrapped with
+        Humanloop decorators to create a trace of Logs on Humanloop.
 
         Here's an example of declaring a [Flow](https://humanloop.com/docs/explanation/flows) in code:
         ```python

src/humanloop/client.py

@@ -16,7 +16,7 @@ def get_headers(self) -> typing.Dict[str, str]:
         headers: typing.Dict[str, str] = {
             "X-Fern-Language": "Python",
             "X-Fern-SDK-Name": "humanloop",
-            "X-Fern-SDK-Version": "0.8.19",
+            "X-Fern-SDK-Version": "0.8.20",
         }
         headers["X-API-KEY"] = self.api_key
         return headers

src/humanloop/core/client_wrapper.py

@@ -212,10 +212,6 @@ def increment(self):
                 sys.stderr.write("\n")
 
 
-# Module-level so it can be shared by threads.
-_PROGRESS_BAR: Optional[_SimpleProgressBar] = None
-
-
 def run_eval(
     client: "BaseHumanloop",
     file: File,
@@ -236,7 +232,6 @@ def run_eval(
     :param workers: the number of threads to process datapoints using your function concurrently.
     :return: per Evaluator checks.
     """
-    global _PROGRESS_BAR
 
     if hasattr(file["callable"], "file"):
         # When the decorator inside `file` is a decorated function,
@@ -274,7 +269,7 @@ def run_eval(
     try:
         type_ = typing.cast(FileType, file_.pop("type"))
         logger.info(
-            f"{CYAN}Evaluating your {type_} function corresponding to `{file_['path']}` on Humanloop{RESET} \n\n"
+            f"{CYAN}Evaluating your {type_} function corresponding to `{file_.get('path') or file_.get('id')}` on Humanloop{RESET} \n\n"
         )
     except KeyError as _:
         type_ = "flow"

src/humanloop/eval_utils/run.py

@@ -202,10 +202,10 @@ def log(
             output="The patient is likely experiencing a myocardial infarction. Immediate medical attention is required.",
             trace_status="incomplete",
             start_time=datetime.datetime.fromisoformat(
-                "2024-07-08 22:40:35+00:00",
+                "2024-07-08 21:40:35+00:00",
             ),
             end_time=datetime.datetime.fromisoformat(
-                "2024-07-08 22:40:39+00:00",
+                "2024-07-08 21:40:39+00:00",
             ),
         )
         """
@@ -1438,10 +1438,10 @@ async def main() -> None:
                 output="The patient is likely experiencing a myocardial infarction. Immediate medical attention is required.",
                 trace_status="incomplete",
                 start_time=datetime.datetime.fromisoformat(
-                    "2024-07-08 22:40:35+00:00",
+                    "2024-07-08 21:40:35+00:00",
                 ),
                 end_time=datetime.datetime.fromisoformat(
-                    "2024-07-08 22:40:39+00:00",
+                    "2024-07-08 21:40:39+00:00",
                 ),
             )
 

src/humanloop/flows/client.py

@@ -1,7 +1,4 @@
-from typing import Optional, TypedDict
-
 from opentelemetry.sdk.trace import TracerProvider
-from typing_extensions import NotRequired
 
 from humanloop.otel.helpers import module_is_installed
 
@@ -41,12 +38,3 @@ def instrument_provider(provider: TracerProvider):
         from opentelemetry.instrumentation.bedrock import BedrockInstrumentor
 
         BedrockInstrumentor().instrument(tracer_provider=provider)
-
-
-class FlowContext(TypedDict):
-    trace_id: NotRequired[str]
-    trace_parent_id: NotRequired[Optional[int]]
-    is_flow_log: NotRequired[bool]
-
-
-TRACE_FLOW_CONTEXT: dict[int, FlowContext] = {}

src/humanloop/otel/__init__.py

@@ -4,3 +4,5 @@
 HUMANLOOP_LOG_KEY = "humanloop.log"
 HUMANLOOP_FILE_TYPE_KEY = "humanloop.file.type"
 HUMANLOOP_PATH_KEY = "humanloop.file.path"
+# Required for the exporter to know when to mark the Flow Log as complete
+HUMANLOOP_FLOW_PREREQUISITES_KEY = "humanloop.flow.prerequisites"