ai-pipeline-core 0.1.3__tar.gz → 0.1.5__tar.gz

{ai_pipeline_core-0.1.3 → ai_pipeline_core-0.1.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ai-pipeline-core
-Version: 0.1.3
+Version: 0.1.5
 Summary: Core utilities for AI-powered processing pipelines using prefect
 Project-URL: Homepage, https://github.com/bbarwik/ai-pipeline-core
 Project-URL: Repository, https://github.com/bbarwik/ai-pipeline-core
@@ -20,7 +20,7 @@ Classifier: Typing :: Typed
 Requires-Python: >=3.12
 Requires-Dist: httpx>=0.28.1
 Requires-Dist: jinja2>=3.1.6
-Requires-Dist: lmnr>=0.7.4
+Requires-Dist: lmnr>=0.7.5
 Requires-Dist: openai>=1.99.9
 Requires-Dist: prefect>=3.4.13
 Requires-Dist: pydantic-settings>=2.10.1
@@ -471,7 +471,7 @@ Built with:
 
 ## Stability Notice
 
-**Current Version**: 0.1.2
+**Current Version**: 0.1.5
 **Status**: Internal Preview
 **API Stability**: Unstable - Breaking changes expected
 **Recommended Use**: Learning and reference only

{ai_pipeline_core-0.1.3 → ai_pipeline_core-0.1.5}/README.md

@@ -429,7 +429,7 @@ Built with:
 
 ## Stability Notice
 
-**Current Version**: 0.1.2
+**Current Version**: 0.1.5
 **Status**: Internal Preview
 **API Stability**: Unstable - Breaking changes expected
 **Recommended Use**: Learning and reference only

{ai_pipeline_core-0.1.3 → ai_pipeline_core-0.1.5}/ai_pipeline_core/__init__.py

@@ -16,7 +16,7 @@ from .prompt_manager import PromptManager
 from .settings import settings
 from .tracing import trace
 
-__version__ = "0.1.2"
+__version__ = "0.1.4"
 
 __all__ = [
     "Document",

{ai_pipeline_core-0.1.3 → ai_pipeline_core-0.1.5}/ai_pipeline_core/documents/document.py

@@ -25,7 +25,7 @@ from .mime_type import (
 class Document(BaseModel, ABC):
     """Abstract base class for all documents"""
 
-    MAX_CONTENT_SIZE: ClassVar[int] = 10 * 1024 * 1024  # 10MB default
+    MAX_CONTENT_SIZE: ClassVar[int] = 25 * 1024 * 1024  # 25MB default
     DESCRIPTION_EXTENSION: ClassVar[str] = ".description.md"
     MARKDOWN_LIST_SEPARATOR: ClassVar[str] = "\n\n---\n\n"
 
@@ -105,7 +105,7 @@ class Document(BaseModel, ABC):
         except TypeError:
             raise DocumentNameError(f"{cls.__name__}.FILES must be an Enum of string values")
 
-        if name not in allowed:
+        if len(allowed) > 0 and name not in allowed:
             allowed_str = ", ".join(sorted(allowed))
             raise DocumentNameError(f"Invalid filename '{name}'. Allowed names: {allowed_str}")
 

{ai_pipeline_core-0.1.3 → ai_pipeline_core-0.1.5}/ai_pipeline_core/llm/ai_messages.py

@@ -1,4 +1,5 @@
 import base64
+import hashlib
 import json
 
 from openai.types.chat import (
@@ -59,6 +60,11 @@ class AIMessages(list[AIMessageType]):
                 messages.append(message)
         return messages
 
+    def get_prompt_cache_key(self, system_prompt: str | None = None) -> str:
+        if not system_prompt:
+            system_prompt = ""
+        return hashlib.sha256((system_prompt + json.dumps(self.to_prompt())).encode()).hexdigest()
+
     @staticmethod
     def document_to_prompt(document: Document) -> list[ChatCompletionContentPartParam]:
         """

{ai_pipeline_core-0.1.3 → ai_pipeline_core-0.1.5}/ai_pipeline_core/llm/client.py

@@ -48,15 +48,13 @@ def _process_messages(
         # Use AIMessages.to_prompt() for context
         context_messages = context.to_prompt()
 
-        # Apply caching to context messages
-        for msg in context_messages:
-            if msg.get("role") == "user":
-                # Add cache control to user messages in context
-                msg["cache_control"] = {  # type: ignore
-                    "type": "ephemeral",
-                    "ttl": "120s",  # Cache for 2m
-                }
-            processed_messages.append(msg)
+        # Apply caching to last context message
+        context_messages[-1]["cache_control"] = {  # type: ignore
+            "type": "ephemeral",
+            "ttl": "120s",  # Cache for 2m
+        }
+
+        processed_messages.extend(context_messages)
 
     # Process regular messages without caching
     if messages:
@@ -108,9 +106,14 @@ async def _generate_with_retry(
         **options.to_openai_completion_kwargs(),
     }
 
+    if context:
+        completion_kwargs["prompt_cache_key"] = context.get_prompt_cache_key(options.system_prompt)
+
     for attempt in range(options.retries):
         try:
-            with Laminar.start_as_current_span(model, span_type="LLM", input=messages) as span:
+            with Laminar.start_as_current_span(
+                model, span_type="LLM", input=processed_messages
+            ) as span:
                 response = await _generate(model, processed_messages, completion_kwargs)
                 span.set_attributes(response.get_laminar_metadata())
                 Laminar.set_span_output(response.content)

{ai_pipeline_core-0.1.3 → ai_pipeline_core-0.1.5}/ai_pipeline_core/tracing.py

@@ -103,6 +103,7 @@ def trace(
     ignore_inputs: list[str] | None = None,
     input_formatter: Callable[..., str] | None = None,
     output_formatter: Callable[..., str] | None = None,
+    preserve_global_context: bool = True,
 ) -> Callable[[Callable[P, R]], Callable[P, R]] | Callable[P, R]:
     """Decorator that wires Laminar tracing and observation into a function.
 
@@ -136,6 +137,7 @@ def trace(
         _ignore_inputs = ignore_inputs
         _input_formatter = input_formatter
         _output_formatter = output_formatter
+        _preserve_global_context = preserve_global_context
 
         # --- Check debug_only flag and environment variable ---
         if debug_only and os.getenv("LMNR_DEBUG", "").lower() != "true":
@@ -173,6 +175,8 @@ def trace(
                 observe_params["input_formatter"] = _input_formatter
             if _output_formatter is not None:
                 observe_params["output_formatter"] = _output_formatter
+            if _preserve_global_context:
+                observe_params["preserve_global_context"] = _preserve_global_context
 
             return observe_params
 

{ai_pipeline_core-0.1.3 → ai_pipeline_core-0.1.5}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ai-pipeline-core"
-version = "0.1.3"
+version = "0.1.5"
 description = "Core utilities for AI-powered processing pipelines using prefect"
 readme = "README.md"
 license = {text = "MIT"}
@@ -22,7 +22,7 @@ classifiers = [
 dependencies = [
     "httpx>=0.28.1",
     "Jinja2>=3.1.6",
-    "lmnr>=0.7.4",
+    "lmnr>=0.7.5",
     "openai>=1.99.9",
     "prefect>=3.4.13",
     "pydantic-settings>=2.10.1",
@@ -140,7 +140,7 @@ reportIncompatibleVariableOverride = "error"
 reportMissingParameterType = "warning"
 
 [tool.bumpversion]
-current_version = "0.1.3"
+current_version = "0.1.5"
 commit = true
 tag = true
 tag_name = "v{new_version}"