ai-pipeline-core 0.2.7__py3-none-any.whl → 0.2.8__py3-none-any.whl

ai_pipeline_core/__init__.py

@@ -118,7 +118,7 @@ from .prompt_manager import PromptManager
 from .settings import Settings
 from .tracing import TraceInfo, TraceLevel, set_trace_cost, trace
 
-__version__ = "0.2.7"
+__version__ = "0.2.8"
 
 __all__ = [
     # Config/Settings

ai_pipeline_core/flow/options.py

@@ -61,11 +61,11 @@ class FlowOptions(BaseSettings):
     """
 
     core_model: ModelName = Field(
-        default="gpt-5",
+        default="gemini-2.5-pro",
         description="Primary model for complex analysis and generation tasks.",
     )
     small_model: ModelName = Field(
-        default="gpt-5-mini",
+        default="grok-4-fast",
         description="Fast, cost-effective model for simple tasks and orchestration.",
     )
 

ai_pipeline_core/llm/client.py

@@ -108,6 +108,38 @@ def _process_messages(
     return processed_messages
 
 
+def _remove_cache_control(
+    messages: list[ChatCompletionMessageParam],
+) -> list[ChatCompletionMessageParam]:
+    """Remove cache control directives from messages.
+
+    Internal utility that strips cache_control fields from both message-level
+    and content-level entries. Used in retry logic when cache-related errors
+    occur during LLM API calls.
+
+    Args:
+        messages: List of messages that may contain cache_control directives.
+
+    Returns:
+        The same message list (modified in-place) with all cache_control
+        fields removed from both messages and their content items.
+
+    Note:
+        This function modifies the input list in-place but also returns it
+        for convenience. Handles both list-based content (multipart) and
+        string content (simple messages).
+    """
+    for message in messages:
+        if content := message.get("content"):
+            if isinstance(content, list):
+                for item in content:
+                    if "cache_control" in item:
+                        del item["cache_control"]
+        if "cache_control" in message:
+            del message["cache_control"]
+    return messages  # type: ignore
+
+
 def _model_name_to_openrouter_model(model: ModelName) -> str:
     """Convert a model name to an OpenRouter model name.
 
@@ -242,7 +274,7 @@ async def _generate_with_retry(
     if not context and not messages:
         raise ValueError("Either context or messages must be provided")
 
-    if "gemini" in model.lower() and context.approximate_tokens_count < 5000:
+    if "gemini" in model.lower() and context.approximate_tokens_count < 10000:
         # Bug fix for minimum explicit context size for Gemini models
         options.cache_ttl = None
 
@@ -272,6 +304,8 @@ async def _generate_with_retry(
             if not isinstance(e, asyncio.TimeoutError):
                 # disable cache if it's not a timeout because it may cause an error
                 completion_kwargs["extra_body"]["cache"] = {"no-cache": True}
+                # sometimes there are issues with cache so cache is removed in case of failure
+                processed_messages = _remove_cache_control(processed_messages)
 
             logger.warning(
                 f"LLM generation failed (attempt {attempt + 1}/{options.retries}): {e}",

{ai_pipeline_core-0.2.7.dist-info → ai_pipeline_core-0.2.8.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ai-pipeline-core
-Version: 0.2.7
+Version: 0.2.8
 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

{ai_pipeline_core-0.2.7.dist-info → ai_pipeline_core-0.2.8.dist-info}/RECORD

@@ -1,4 +1,4 @@
-ai_pipeline_core/__init__.py,sha256=SCyd40nB8yF10KylHMBhEdpF0slV35OTsIePEKi9GP8,5720
+ai_pipeline_core/__init__.py,sha256=NMpeyF-rJdNnNWfi9eoOgWNnk7kOVHqvSZq3cnyDuCI,5720
 ai_pipeline_core/exceptions.py,sha256=vx-XLTw2fJSPs-vwtXVYtqoQUcOc0JeI7UmHqRqQYWU,1569
 ai_pipeline_core/pipeline.py,sha256=fWTVmrnOEIFge6o2NUYW2ndGef5UurpL8_fK5tkXbzI,28700
 ai_pipeline_core/prefect.py,sha256=91ZgLJHsDsRUW77CpNmkKxYs3RCJuucPM3pjKmNBeDg,2199
@@ -16,10 +16,10 @@ ai_pipeline_core/documents/temporary_document.py,sha256=Sam344Mm5AlZTm3_l01YdDWe
 ai_pipeline_core/documents/utils.py,sha256=ZyJNjFN7ihWno0K7dJZed7twYmmPLA0z40UzFw1A3A8,5465
 ai_pipeline_core/flow/__init__.py,sha256=2BfWYMOPYW5teGzwo-qzpn_bom1lxxry0bPsjVgcsCk,188
 ai_pipeline_core/flow/config.py,sha256=3PCDph2n8dj-txqAvd9Wflbi_6lmfXFR9rUhM-szGSQ,18887
-ai_pipeline_core/flow/options.py,sha256=2rKR2GifhXcyw8avI_oiEDMLC2jm5Qzpw8z56pbxUMo,2285
+ai_pipeline_core/flow/options.py,sha256=zn3N5DgYtlxLq0AvXfana3UOhym7A3XCheQSBIIarZE,2295
 ai_pipeline_core/llm/__init__.py,sha256=3B_vtEzxrzidP1qOUNQ4RxlUmxZ2MBKQcUhQiTybM9g,661
 ai_pipeline_core/llm/ai_messages.py,sha256=Onin3UPdbJQNl3WfY3-_jE5KRmF-ciXsa5K6UPOiy5s,14410
-ai_pipeline_core/llm/client.py,sha256=385nKrr5fbDha8lNe9AbGLJ9Eszzo3_ArC22WWd8T_s,23646
+ai_pipeline_core/llm/client.py,sha256=Da1NgGzfIkFRw_aDASK36MMdKe3DXaj5_3wMg0gR-Hk,24999
 ai_pipeline_core/llm/model_options.py,sha256=uRNIHfVeh2sgt1mZBiOUx6hPQ6GKjB8b7TytZJ6afKg,11768
 ai_pipeline_core/llm/model_response.py,sha256=6kEr9ss3UGlykvtWAvh1l55rGw2-wyVup3QJhm0Oggc,13264
 ai_pipeline_core/llm/model_types.py,sha256=2J4Qsb1x21I4eo_VPeaMMOW8shOGPqzJuoGjTLcBFPM,2791
@@ -35,7 +35,7 @@ ai_pipeline_core/storage/storage.py,sha256=ClMr419Y-eU2RuOjZYd51dC0stWQk28Vb56Pv
 ai_pipeline_core/utils/__init__.py,sha256=TJSmEm1Quf-gKwXrxM96u2IGzVolUyeNNfLMPoLstXI,254
 ai_pipeline_core/utils/deploy.py,sha256=rAtRuwkmGkc-fqvDMXpt08OzLrD7KTDMAmLDC9wYg7Y,13147
 ai_pipeline_core/utils/remote_deployment.py,sha256=cPTgnS5InK08qiWnuPz3e8YKjoT3sPBloSaDfNTzghs,10137
-ai_pipeline_core-0.2.7.dist-info/METADATA,sha256=Ec4yFJTtt8qFtQD4Hg7KXT2VO80_i8JmLbwfXNia2GE,15159
-ai_pipeline_core-0.2.7.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-ai_pipeline_core-0.2.7.dist-info/licenses/LICENSE,sha256=kKj8mfbdWwkyG3U6n7ztB3bAZlEwShTkAsvaY657i3I,1074
-ai_pipeline_core-0.2.7.dist-info/RECORD,,
+ai_pipeline_core-0.2.8.dist-info/METADATA,sha256=Sl_8s24ar0lwa7iio4d0QYDhzsmAIGHHqOKuXnFxP7s,15159
+ai_pipeline_core-0.2.8.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+ai_pipeline_core-0.2.8.dist-info/licenses/LICENSE,sha256=kKj8mfbdWwkyG3U6n7ztB3bAZlEwShTkAsvaY657i3I,1074
+ai_pipeline_core-0.2.8.dist-info/RECORD,,