promptlayer 1.0.64__tar.gz → 1.0.66__tar.gz

{promptlayer-1.0.64 → promptlayer-1.0.66}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: promptlayer
-Version: 1.0.64
+Version: 1.0.66
 Summary: PromptLayer is a platform for prompt engineering and tracks your LLM requests.
 License: Apache-2.0
 Author: Magniv

{promptlayer-1.0.64 → promptlayer-1.0.66}/promptlayer/__init__.py

@@ -1,4 +1,4 @@
 from .promptlayer import AsyncPromptLayer, PromptLayer
 
-__version__ = "1.0.64"
+__version__ = "1.0.66"
 __all__ = ["PromptLayer", "AsyncPromptLayer", "__version__"]

{promptlayer-1.0.64 → promptlayer-1.0.66}/promptlayer/promptlayer.py

@@ -171,6 +171,11 @@ class PromptLayer(PromptLayerMixin):
                 metadata=llm_data["prompt_blueprint"]["metadata"],
             )
 
+        if isinstance(response, dict):
+            request_response = response
+        else:
+            request_response = response.model_dump(mode="json")
+
         request_log = self._track_request_log(
             llm_data,
             tags,
@@ -178,7 +183,7 @@ class PromptLayer(PromptLayerMixin):
             group_id,
             pl_run_span_id,
             metadata=metadata,
-            request_response=response.model_dump(mode="json"),
+            request_response=request_response,
         )
 
         return {
@@ -592,6 +597,11 @@ class AsyncPromptLayer(PromptLayerMixin):
             function_kwargs=llm_data["function_kwargs"],
         )
 
+        if isinstance(response, dict):
+            request_response = response
+        else:
+            request_response = response.model_dump(mode="json")
+
         if stream:
             track_request_callable = await self._create_track_request_callable(
                 request_params=llm_data,
@@ -601,7 +611,7 @@ class AsyncPromptLayer(PromptLayerMixin):
                 pl_run_span_id=pl_run_span_id,
             )
             return astream_response(
-                response,
+                request_response,
                 track_request_callable,
                 llm_data["stream_function"],
                 llm_data["prompt_blueprint"]["metadata"],
@@ -614,11 +624,11 @@ class AsyncPromptLayer(PromptLayerMixin):
             group_id,
             pl_run_span_id,
             metadata=metadata,
-            request_response=response.model_dump(mode="json"),
+            request_response=request_response,
         )
 
         return {
             "request_id": request_log.get("request_id", None),
-            "raw_response": response,
+            "raw_response": request_response,
             "prompt_blueprint": request_log.get("prompt_blueprint", None),
         }

{promptlayer-1.0.64 → promptlayer-1.0.66}/promptlayer/promptlayer_mixins.py

@@ -13,6 +13,7 @@ from promptlayer.span_exporter import PromptLayerSpanExporter
 from promptlayer.streaming import (
     aanthropic_stream_completion,
     aanthropic_stream_message,
+    abedrock_stream_message,
     agoogle_stream_chat,
     agoogle_stream_completion,
     amistral_stream_chat,
@@ -20,6 +21,7 @@ from promptlayer.streaming import (
     anthropic_stream_message,
     aopenai_stream_chat,
     aopenai_stream_completion,
+    bedrock_stream_message,
     google_stream_chat,
     google_stream_completion,
     mistral_stream_chat,
@@ -27,10 +29,12 @@ from promptlayer.streaming import (
     openai_stream_completion,
 )
 from promptlayer.utils import (
+    aamazon_bedrock_request,
     aanthropic_bedrock_request,
     aanthropic_request,
     aazure_openai_request,
     agoogle_request,
+    amazon_bedrock_request,
     amistral_request,
     anthropic_bedrock_request,
     anthropic_request,
@@ -94,6 +98,16 @@ MAP_PROVIDER_TO_FUNCTION_NAME = {
             "stream_function": google_stream_completion,
         },
     },
+    "amazon.bedrock": {
+        "chat": {
+            "function_name": "boto3.bedrock-runtime.converse",
+            "stream_function": bedrock_stream_message,
+        },
+        "completion": {
+            "function_name": "boto3.bedrock-runtime.converse",
+            "stream_function": bedrock_stream_message,
+        },
+    },
     "anthropic.bedrock": {
         "chat": {
             "function_name": "anthropic.messages.create",
@@ -114,6 +128,7 @@ MAP_PROVIDER_TO_FUNCTION = {
     "openai": openai_request,
     "openai.azure": azure_openai_request,
     "vertexai": vertexai_request,
+    "amazon.bedrock": amazon_bedrock_request,
     "anthropic.bedrock": anthropic_bedrock_request,
 }
 
@@ -168,6 +183,16 @@ AMAP_PROVIDER_TO_FUNCTION_NAME = {
             "stream_function": agoogle_stream_completion,
         },
     },
+    "amazon.bedrock": {
+        "chat": {
+            "function_name": "boto3.bedrock-runtime.converse",
+            "stream_function": abedrock_stream_message,
+        },
+        "completion": {
+            "function_name": "boto3.bedrock-runtime.converse",
+            "stream_function": abedrock_stream_message,
+        },
+    },
     "anthropic.bedrock": {
         "chat": {
             "function_name": "anthropic.messages.create",
@@ -188,6 +213,7 @@ AMAP_PROVIDER_TO_FUNCTION = {
     "openai": aopenai_request,
     "openai.azure": aazure_openai_request,
     "vertexai": avertexai_request,
+    "amazon.bedrock": aamazon_bedrock_request,
     "anthropic.bedrock": aanthropic_bedrock_request,
 }
 

{promptlayer-1.0.64 → promptlayer-1.0.66}/promptlayer/streaming/__init__.py

@@ -13,6 +13,7 @@ from .blueprint_builder import (
 from .response_handlers import (
     aanthropic_stream_completion,
     aanthropic_stream_message,
+    abedrock_stream_message,
     agoogle_stream_chat,
     agoogle_stream_completion,
     amistral_stream_chat,
@@ -20,6 +21,7 @@ from .response_handlers import (
     anthropic_stream_message,
     aopenai_stream_chat,
     aopenai_stream_completion,
+    bedrock_stream_message,
     google_stream_chat,
     google_stream_completion,
     mistral_stream_chat,
@@ -45,6 +47,8 @@ __all__ = [
     "aopenai_stream_completion",
     "anthropic_stream_completion",
     "aanthropic_stream_completion",
+    "bedrock_stream_message",
+    "abedrock_stream_message",
     "google_stream_chat",
     "google_stream_completion",
     "agoogle_stream_chat",

{promptlayer-1.0.64 → promptlayer-1.0.66}/promptlayer/streaming/blueprint_builder.py

@@ -137,3 +137,38 @@ def build_prompt_blueprint_from_google_event(event, metadata):
 
     assistant_message = _build_assistant_message(assistant_content, tool_calls or None, template_format="f-string")
     return _build_prompt_blueprint(assistant_message, metadata)
+
+
+def build_prompt_blueprint_from_bedrock_event(result, metadata):
+    """
+    Build a prompt blueprint from an Amazon Bedrock streaming event.
+    """
+    assistant_content = []
+    tool_calls = []
+
+    if "contentBlockDelta" in result:
+        delta = result["contentBlockDelta"].get("delta", {})
+
+        if "reasoningContent" in delta:
+            reasoning_text = delta["reasoningContent"].get("text", "")
+            signature = delta["reasoningContent"].get("signature")
+            assistant_content.append(_create_content_item("thinking", thinking=reasoning_text, signature=signature))
+
+        elif "text" in delta:
+            assistant_content.append(_create_content_item("text", text=delta["text"]))
+
+        elif "toolUse" in delta:
+            tool_use = delta["toolUse"]
+            assistant_content.append(
+                _create_tool_call(tool_use.get("toolUseId", ""), tool_use.get("name", ""), tool_use.get("input", ""))
+            )
+
+    elif "contentBlockStart" in result:
+        start_block = result["contentBlockStart"].get("start", {})
+
+        if "toolUse" in start_block:
+            tool_use = start_block["toolUse"]
+            tool_calls.append(_create_tool_call(tool_use.get("toolUseId", ""), tool_use.get("name", ""), ""))
+
+    assistant_message = _build_assistant_message(assistant_content, tool_calls or None)
+    return _build_prompt_blueprint(assistant_message, metadata)

{promptlayer-1.0.64 → promptlayer-1.0.66}/promptlayer/streaming/response_handlers.py

@@ -5,6 +5,7 @@ This module contains handlers that process streaming responses from various
 LLM providers and return both the final response and prompt blueprint.
 """
 
+import json
 from typing import Any, AsyncIterable, List
 
 
@@ -548,3 +549,147 @@ async def amistral_stream_chat(generator: AsyncIterable[Any]) -> Any:
     response.choices[0].message.content = content
     response.choices[0].message.tool_calls = tool_calls
     return response
+
+
+def bedrock_stream_message(results: list):
+    """Process Amazon Bedrock streaming message results and return response + blueprint"""
+
+    response = {"ResponseMetadata": {}, "output": {"message": {}}, "stopReason": "end_turn", "metrics": {}, "usage": {}}
+
+    content_blocks = []
+    current_tool_call = None
+    current_tool_input = ""
+    current_text = ""
+    current_signature = ""
+    current_thinking = ""
+
+    for event in results:
+        if "contentBlockStart" in event:
+            content_block = event["contentBlockStart"]
+            if "start" in content_block and "toolUse" in content_block["start"]:
+                tool_use = content_block["start"]["toolUse"]
+                current_tool_call = {"toolUse": {"toolUseId": tool_use["toolUseId"], "name": tool_use["name"]}}
+                current_tool_input = ""
+
+        elif "contentBlockDelta" in event:
+            delta = event["contentBlockDelta"]["delta"]
+            if "text" in delta:
+                current_text += delta["text"]
+            elif "reasoningContent" in delta:
+                reasoning_content = delta["reasoningContent"]
+                if "text" in reasoning_content:
+                    current_thinking += reasoning_content["text"]
+                elif "signature" in reasoning_content:
+                    current_signature += reasoning_content["signature"]
+            elif "toolUse" in delta:
+                if "input" in delta["toolUse"]:
+                    input_chunk = delta["toolUse"]["input"]
+                    current_tool_input += input_chunk
+                    if not input_chunk.strip():
+                        continue
+
+        elif "contentBlockStop" in event:
+            if current_tool_call and current_tool_input:
+                try:
+                    current_tool_call["toolUse"]["input"] = json.loads(current_tool_input)
+                except json.JSONDecodeError:
+                    current_tool_call["toolUse"]["input"] = {}
+                content_blocks.append(current_tool_call)
+                current_tool_call = None
+                current_tool_input = ""
+            elif current_text:
+                content_blocks.append({"text": current_text})
+                current_text = ""
+            elif current_thinking and current_signature:
+                content_blocks.append(
+                    {
+                        "reasoningContent": {
+                            "reasoningText": {"text": current_thinking, "signature": current_signature},
+                        }
+                    }
+                )
+                current_thinking = ""
+                current_signature = ""
+
+        elif "messageStop" in event:
+            response["stopReason"] = event["messageStop"]["stopReason"]
+
+        elif "metadata" in event:
+            metadata = event["metadata"]
+            response["usage"] = metadata.get("usage", {})
+            response["metrics"] = metadata.get("metrics", {})
+
+    response["output"]["message"] = {"role": "assistant", "content": content_blocks}
+    return response
+
+
+async def abedrock_stream_message(generator: AsyncIterable[Any]) -> Any:
+    """Async version of bedrock_stream_message"""
+
+    response = {"ResponseMetadata": {}, "output": {"message": {}}, "stopReason": "end_turn", "metrics": {}, "usage": {}}
+
+    content_blocks = []
+    current_tool_call = None
+    current_tool_input = ""
+    current_text = ""
+    current_signature = ""
+    current_thinking = ""
+
+    async for event in generator:
+        if "contentBlockStart" in event:
+            content_block = event["contentBlockStart"]
+            if "start" in content_block and "toolUse" in content_block["start"]:
+                tool_use = content_block["start"]["toolUse"]
+                current_tool_call = {"toolUse": {"toolUseId": tool_use["toolUseId"], "name": tool_use["name"]}}
+                current_tool_input = ""
+
+        elif "contentBlockDelta" in event:
+            delta = event["contentBlockDelta"]["delta"]
+            if "text" in delta:
+                current_text += delta["text"]
+            elif "reasoningContent" in delta:
+                reasoning_content = delta["reasoningContent"]
+                if "text" in reasoning_content:
+                    current_thinking += reasoning_content["text"]
+                elif "signature" in reasoning_content:
+                    current_signature += reasoning_content["signature"]
+            elif "toolUse" in delta:
+                if "input" in delta["toolUse"]:
+                    input_chunk = delta["toolUse"]["input"]
+                    current_tool_input += input_chunk
+                    if not input_chunk.strip():
+                        continue
+
+        elif "contentBlockStop" in event:
+            if current_tool_call and current_tool_input:
+                try:
+                    current_tool_call["toolUse"]["input"] = json.loads(current_tool_input)
+                except json.JSONDecodeError:
+                    current_tool_call["toolUse"]["input"] = {}
+                content_blocks.append(current_tool_call)
+                current_tool_call = None
+                current_tool_input = ""
+            elif current_text:
+                content_blocks.append({"text": current_text})
+                current_text = ""
+            elif current_thinking and current_signature:
+                content_blocks.append(
+                    {
+                        "reasoningContent": {
+                            "reasoningText": {"text": current_thinking, "signature": current_signature},
+                        }
+                    }
+                )
+                current_thinking = ""
+                current_signature = ""
+
+        elif "messageStop" in event:
+            response["stopReason"] = event["messageStop"]["stopReason"]
+
+        elif "metadata" in event:
+            metadata = event["metadata"]
+            response["usage"] = metadata.get("usage", {})
+            response["metrics"] = metadata.get("metrics", {})
+
+    response["output"]["message"] = {"role": "assistant", "content": content_blocks}
+    return response

{promptlayer-1.0.64 → promptlayer-1.0.66}/promptlayer/streaming/stream_processor.py

@@ -2,6 +2,7 @@ from typing import Any, AsyncGenerator, AsyncIterable, Callable, Dict, Generator
 
 from .blueprint_builder import (
     build_prompt_blueprint_from_anthropic_event,
+    build_prompt_blueprint_from_bedrock_event,
     build_prompt_blueprint_from_google_event,
     build_prompt_blueprint_from_openai_chunk,
 )
@@ -18,12 +19,15 @@ def _build_stream_blueprint(result: Any, metadata: Dict) -> Any:
     elif provider == "google" or (provider == "vertexai" and model_name.startswith("gemini")):
         return build_prompt_blueprint_from_google_event(result, metadata)
 
-    elif provider in ["anthropic", "anthropic.bedrock"] or (provider == "vertexai" and model_name.startswith("claude")):
+    elif provider in ("anthropic", "anthropic.bedrock") or (provider == "vertexai" and model_name.startswith("claude")):
         return build_prompt_blueprint_from_anthropic_event(result, metadata)
 
     elif provider == "mistral":
         return build_prompt_blueprint_from_openai_chunk(result.data, metadata)
 
+    elif provider == "amazon.bedrock":
+        return build_prompt_blueprint_from_bedrock_event(result, metadata)
+
     return None
 
 
@@ -37,6 +41,11 @@ def _build_stream_data(result: Any, stream_blueprint: Any, request_id: Any = Non
 
 def stream_response(*, generator: Generator, after_stream: Callable, map_results: Callable, metadata: Dict):
     results = []
+    provider = metadata.get("model", {}).get("provider", "")
+    if provider == "amazon.bedrock":
+        response_metadata = generator.get("ResponseMetadata", {})
+        generator = generator.get("stream", generator)
+
     for result in generator:
         results.append(result)
 
@@ -45,7 +54,12 @@ def stream_response(*, generator: Generator, after_stream: Callable, map_results
         yield data
 
     request_response = map_results(results)
-    response = after_stream(request_response=request_response.model_dump(mode="json"))
+    if provider == "amazon.bedrock":
+        request_response["ResponseMetadata"] = response_metadata
+    else:
+        request_response = request_response.model_dump(mode="json")
+
+    response = after_stream(request_response=request_response)
     data["request_id"] = response.get("request_id")
     data["prompt_blueprint"] = response.get("prompt_blueprint")
     yield data
@@ -58,6 +72,10 @@ async def astream_response(
     metadata: Dict[str, Any] = None,
 ) -> AsyncGenerator[Dict[str, Any], None]:
     results = []
+    provider = metadata.get("model", {}).get("provider", "")
+    if provider == "amazon.bedrock":
+        response_metadata = generator.get("ResponseMetadata", {})
+        generator = generator.get("stream", generator)
 
     async for result in generator:
         results.append(result)
@@ -71,7 +89,13 @@ async def astream_response(
             yield item
 
     request_response = await map_results(async_generator_from_list(results))
-    after_stream_response = await after_stream(request_response=request_response.model_dump(mode="json"))
+
+    if provider == "amazon.bedrock":
+        request_response["ResponseMetadata"] = response_metadata
+    else:
+        request_response = request_response.model_dump(mode="json")
+
+    after_stream_response = await after_stream(request_response=request_response)
     data["request_id"] = after_stream_response.get("request_id")
     data["prompt_blueprint"] = after_stream_response.get("prompt_blueprint")
     yield data

{promptlayer-1.0.64 → promptlayer-1.0.66}/promptlayer/utils.py

@@ -1455,7 +1455,14 @@ MAP_TYPE_TO_GOOGLE_FUNCTION = {
 def google_request(prompt_blueprint: GetPromptTemplateResponse, client_kwargs: dict, function_kwargs: dict):
     from google import genai
 
-    client = genai.Client()
+    if os.environ.get("GOOGLE_GENAI_USE_VERTEXAI") == "true":
+        client = genai.Client(
+            vertexai=True,
+            project=os.environ.get("GOOGLE_CLOUD_PROJECT"),
+            location=os.environ.get("GOOGLE_CLOUD_LOCATION"),
+        )
+    else:
+        client = genai.Client(api_key=os.environ.get("GEMINI_API_KEY") or os.environ.get("GOOGLE_API_KEY"))
     request_to_make = MAP_TYPE_TO_GOOGLE_FUNCTION[prompt_blueprint["prompt_template"]["type"]]
     return request_to_make(client, **function_kwargs)
 
@@ -1493,7 +1500,14 @@ AMAP_TYPE_TO_GOOGLE_FUNCTION = {
 async def agoogle_request(prompt_blueprint: GetPromptTemplateResponse, client_kwargs: dict, function_kwargs: dict):
     from google import genai
 
-    client = genai.Client()
+    if os.environ.get("GOOGLE_GENAI_USE_VERTEXAI") == "true":
+        client = genai.Client(
+            vertexai=True,
+            project=os.environ.get("GOOGLE_CLOUD_PROJECT"),
+            location=os.environ.get("GOOGLE_CLOUD_LOCATION"),
+        )
+    else:
+        client = genai.Client(api_key=os.environ.get("GEMINI_API_KEY") or os.environ.get("GOOGLE_API_KEY"))
     request_to_make = AMAP_TYPE_TO_GOOGLE_FUNCTION[prompt_blueprint["prompt_template"]["type"]]
     return await request_to_make(client, **function_kwargs)
 
@@ -1544,6 +1558,51 @@ async def avertexai_request(prompt_blueprint: GetPromptTemplateResponse, client_
     )
 
 
+def amazon_bedrock_request(prompt_blueprint: GetPromptTemplateResponse, client_kwargs: dict, function_kwargs: dict):
+    import boto3
+
+    bedrock_client = boto3.client(
+        "bedrock-runtime",
+        aws_access_key_id=function_kwargs.pop("aws_access_key", None),
+        aws_secret_access_key=function_kwargs.pop("aws_secret_key", None),
+        region_name=function_kwargs.pop("aws_region", "us-east-1"),
+    )
+
+    stream = function_kwargs.pop("stream", False)
+
+    if stream:
+        return bedrock_client.converse_stream(**function_kwargs)
+    else:
+        return bedrock_client.converse(**function_kwargs)
+
+
+async def aamazon_bedrock_request(
+    prompt_blueprint: GetPromptTemplateResponse, client_kwargs: dict, function_kwargs: dict
+):
+    import aioboto3
+
+    aws_access_key = function_kwargs.pop("aws_access_key", None)
+    aws_secret_key = function_kwargs.pop("aws_secret_key", None)
+    aws_region = function_kwargs.pop("aws_region", "us-east-1")
+
+    session_kwargs = {}
+    if aws_access_key:
+        session_kwargs["aws_access_key_id"] = aws_access_key
+    if aws_secret_key:
+        session_kwargs["aws_secret_access_key"] = aws_secret_key
+    if aws_region:
+        session_kwargs["region_name"] = aws_region
+
+    stream = function_kwargs.pop("stream", False)
+    session = aioboto3.Session()
+
+    async with session.client("bedrock-runtime", **session_kwargs) as client:
+        if stream:
+            return await client.converse_stream(**function_kwargs)
+        else:
+            return await client.converse(**function_kwargs)
+
+
 def anthropic_bedrock_request(prompt_blueprint: GetPromptTemplateResponse, client_kwargs: dict, function_kwargs: dict):
     from anthropic import AnthropicBedrock
 

{promptlayer-1.0.64 → promptlayer-1.0.66}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "promptlayer"
-version = "1.0.64"
+version = "1.0.66"
 description = "PromptLayer is a platform for prompt engineering and tracks your LLM requests."
 authors = ["Magniv <hello@magniv.io>"]
 license = "Apache-2.0"
@@ -29,6 +29,8 @@ pytest-network = "^0.0.1"
 pytest-parametrize-cases = "^0.1.2"
 pydantic = "^2.11.7"
 pydantic-settings = "^2.10.1"
+boto3 = "^1.35.0"
+aioboto3 = "^13.0.0"
 
 [build-system]
 requires = ["poetry-core"]