langtrace-python-sdk 3.3.23__py3-none-any.whl → 3.3.24__py3-none-any.whl

examples/awsbedrock_examples/__init__.py

@@ -1,9 +1,7 @@
 from examples.awsbedrock_examples.converse import (
-    use_converse_stream,
-    use_converse,
     use_invoke_model_anthropic,
-    use_invoke_model_cohere,
-    use_invoke_model_amazon,
+    use_invoke_model_titan,
+    use_invoke_model_llama,
 )
 from langtrace_python_sdk import langtrace, with_langtrace_root_span
 
@@ -12,8 +10,9 @@ class AWSBedrockRunner:
     @with_langtrace_root_span("AWS_Bedrock")
     def run(self):
 
-        use_converse_stream()
-        use_converse()
-        use_invoke_model_anthropic()
-        use_invoke_model_cohere()
-        use_invoke_model_amazon()
+        # use_converse_stream()
+        # use_converse()
+        # use_invoke_model_anthropic(stream=True)
+        # use_invoke_model_cohere()
+        # use_invoke_model_llama(stream=False)
+        use_invoke_model_titan(stream=False)

examples/awsbedrock_examples/converse.py

@@ -88,6 +88,12 @@ def use_invoke_model_titan(stream=False):
             response = brt.invoke_model_with_response_stream(
                 body=body, modelId=modelId, accept=accept, contentType=contentType
             )
+            # Extract and print the response text in real-time.
+            for event in response["body"]:
+                chunk = json.loads(event["chunk"]["bytes"])
+                if "outputText" in chunk:
+                    print(chunk["outputText"], end="")
+
         else:
             response = brt.invoke_model(
                 body=body, modelId=modelId, accept=accept, contentType=contentType
@@ -130,7 +136,8 @@ def use_invoke_model_anthropic(stream=False):
             for event in stream_response:
                 chunk = event.get("chunk")
                 if chunk:
-                    print(json.loads(chunk.get("bytes").decode()))
+                    # print(json.loads(chunk.get("bytes").decode()))
+                    pass
 
     else:
         response = brt.invoke_model(
@@ -141,7 +148,7 @@ def use_invoke_model_anthropic(stream=False):
         print(response_body.get("completion"))
 
 
-def use_invoke_model_llama():
+def use_invoke_model_llama(stream=False):
     model_id = "meta.llama3-8b-instruct-v1:0"
     prompt = "What is the capital of France?"
     max_gen_len = 128
@@ -157,11 +164,18 @@ def use_invoke_model_llama():
             "top_p": top_p,
         }
     )
-    response = brt.invoke_model(body=body, modelId=model_id)
-
-    response_body = json.loads(response.get("body").read())
 
-    return response_body
+    if stream:
+        response = brt.invoke_model_with_response_stream(body=body, modelId=model_id)
+        for event in response["body"]:
+            chunk = json.loads(event["chunk"]["bytes"])
+            if "generation" in chunk:
+                # print(chunk["generation"], end="")
+                pass
+    else:
+        response = brt.invoke_model(body=body, modelId=model_id)
+        response_body = json.loads(response.get("body").read())
+        return response_body
 
 
 # print(get_foundation_models())

langtrace_python_sdk/instrumentation/aws_bedrock/patch.py

@@ -16,9 +16,7 @@ limitations under the License.
 
 import json
 
-from langtrace_python_sdk.instrumentation.aws_bedrock.bedrock_streaming_wrapper import (
-    StreamingWrapper,
-)
+from wrapt import ObjectProxy
 from .stream_body_wrapper import BufferedStreamBody
 from functools import wraps
 from langtrace.trace_attributes import (
@@ -87,6 +85,11 @@ def patch_aws_bedrock(tracer, version):
 
         client = wrapped(*args, **kwargs)
         client.invoke_model = patch_invoke_model(client.invoke_model, tracer, version)
+        client.invoke_model_with_response_stream = (
+            patch_invoke_model_with_response_stream(
+                client.invoke_model_with_response_stream, tracer, version
+            )
+        )
 
         client.converse = patch_converse(client.converse, tracer, version)
         client.converse_stream = patch_converse_stream(
@@ -186,6 +189,56 @@ def patch_invoke_model(original_method, tracer, version):
     return traced_method
 
 
+def patch_invoke_model_with_response_stream(original_method, tracer, version):
+    @wraps(original_method)
+    def traced_method(*args, **kwargs):
+        modelId = kwargs.get("modelId")
+        (vendor, _) = modelId.split(".")
+        span_attributes = {
+            **get_langtrace_attributes(version, vendor, vendor_type="framework"),
+            **get_extra_attributes(),
+        }
+        span = tracer.start_span(
+            name=get_span_name("aws_bedrock.invoke_model_with_response_stream"),
+            kind=SpanKind.CLIENT,
+            context=set_span_in_context(trace.get_current_span()),
+        )
+        set_span_attributes(span, span_attributes)
+        response = original_method(*args, **kwargs)
+        if span.is_recording():
+            handle_streaming_call(span, kwargs, response)
+        return response
+
+    return traced_method
+
+
+def handle_streaming_call(span, kwargs, response):
+
+    def stream_finished(response_body):
+        request_body = json.loads(kwargs.get("body"))
+
+        (vendor, model) = kwargs.get("modelId").split(".")
+
+        set_span_attribute(span, SpanAttributes.LLM_REQUEST_MODEL, model)
+        set_span_attribute(span, SpanAttributes.LLM_RESPONSE_MODEL, model)
+
+        if vendor == "amazon":
+            set_amazon_attributes(span, request_body, response_body)
+
+        if vendor == "anthropic":
+            if "prompt" in request_body:
+                set_anthropic_completions_attributes(span, request_body, response_body)
+            elif "messages" in request_body:
+                set_anthropic_messages_attributes(span, request_body, response_body)
+
+        if vendor == "meta":
+            set_llama_meta_attributes(span, request_body, response_body)
+
+        span.end()
+
+    response["body"] = StreamingBedrockWrapper(response["body"], stream_finished)
+
+
 def handle_call(span, kwargs, response):
     modelId = kwargs.get("modelId")
     (vendor, model_name) = modelId.split(".")
@@ -195,7 +248,6 @@ def handle_call(span, kwargs, response):
     request_body = json.loads(kwargs.get("body"))
     response_body = json.loads(response.get("body").read())
 
-    set_span_attribute(span, SpanAttributes.LLM_SYSTEM, vendor)
     set_span_attribute(span, SpanAttributes.LLM_RESPONSE_MODEL, modelId)
     set_span_attribute(span, SpanAttributes.LLM_REQUEST_MODEL, modelId)
 
@@ -222,12 +274,18 @@ def set_llama_meta_attributes(span, request_body, response_body):
     set_span_attribute(
         span, SpanAttributes.LLM_REQUEST_MAX_TOKENS, request_body.get("max_gen_len")
     )
+    if "invocation_metrics" in response_body:
+        input_tokens = response_body.get("invocation_metrics").get("inputTokenCount")
+        output_tokens = response_body.get("invocation_metrics").get("outputTokenCount")
+    else:
+        input_tokens = response_body.get("prompt_token_count")
+        output_tokens = response_body.get("generation_token_count")
 
     set_usage_attributes(
         span,
         {
-            "input_tokens": response_body.get("prompt_token_count"),
-            "output_tokens": response_body.get("generation_token_count"),
+            "input_tokens": input_tokens,
+            "output_tokens": output_tokens,
         },
     )
 
@@ -245,7 +303,6 @@ def set_llama_meta_attributes(span, request_body, response_body):
         }
     ]
     set_span_attribute(span, SpanAttributes.LLM_PROMPTS, json.dumps(prompts))
-    print(completions)
     set_event_completion(span, completions)
 
 
@@ -257,13 +314,22 @@ def set_amazon_attributes(span, request_body, response_body):
             "content": request_body.get("inputText"),
         }
     ]
-    completions = [
-        {
-            "role": "assistant",
-            "content": result.get("outputText"),
-        }
-        for result in response_body.get("results")
-    ]
+    if "results" in response_body:
+        completions = [
+            {
+                "role": "assistant",
+                "content": result.get("outputText"),
+            }
+            for result in response_body.get("results")
+        ]
+
+    else:
+        completions = [
+            {
+                "role": "assistant",
+                "content": response_body.get("outputText"),
+            }
+        ]
     set_span_attribute(
         span, SpanAttributes.LLM_REQUEST_MAX_TOKENS, config.get("maxTokenCount")
     )
@@ -272,13 +338,19 @@ def set_amazon_attributes(span, request_body, response_body):
     )
     set_span_attribute(span, SpanAttributes.LLM_REQUEST_TOP_P, config.get("topP"))
     set_span_attribute(span, SpanAttributes.LLM_PROMPTS, json.dumps(prompts))
+    input_tokens = response_body.get("inputTextTokenCount")
+    if "results" in response_body:
+        output_tokens = sum(
+            int(result.get("tokenCount")) for result in response_body.get("results")
+        )
+    else:
+        output_tokens = response_body.get("outputTextTokenCount")
+
     set_usage_attributes(
         span,
         {
-            "input_tokens": response_body.get("inputTextTokenCount"),
-            "output_tokens": sum(
-                int(result.get("tokenCount")) for result in response_body.get("results")
-            ),
+            "input_tokens": input_tokens,
+            "output_tokens": output_tokens,
         },
     )
     set_event_completion(span, completions)
@@ -320,7 +392,7 @@ def set_anthropic_messages_attributes(span, request_body, response_body):
     set_span_attribute(
         span,
         SpanAttributes.LLM_REQUEST_MAX_TOKENS,
-        request_body.get("max_tokens_to_sample"),
+        request_body.get("max_tokens_to_sample") or request_body.get("max_tokens"),
     )
     set_span_attribute(
         span,
@@ -394,3 +466,62 @@ def set_span_streaming_response(span, response):
         set_event_completion(
             span, [{"role": role or "assistant", "content": streaming_response}]
         )
+
+
+class StreamingBedrockWrapper(ObjectProxy):
+    def __init__(
+        self,
+        response,
+        stream_done_callback=None,
+    ):
+        super().__init__(response)
+
+        self._stream_done_callback = stream_done_callback
+        self._accumulating_body = {"generation": ""}
+
+    def __iter__(self):
+        for event in self.__wrapped__:
+            self._process_event(event)
+            yield event
+
+    def _process_event(self, event):
+        chunk = event.get("chunk")
+        if not chunk:
+            return
+
+        decoded_chunk = json.loads(chunk.get("bytes").decode())
+        type = decoded_chunk.get("type")
+
+        if type is None and "outputText" in decoded_chunk:
+            self._stream_done_callback(decoded_chunk)
+            return
+        if "generation" in decoded_chunk:
+            self._accumulating_body["generation"] += decoded_chunk.get("generation")
+
+        if type == "message_start":
+            self._accumulating_body = decoded_chunk.get("message")
+        elif type == "content_block_start":
+            self._accumulating_body["content"].append(
+                decoded_chunk.get("content_block")
+            )
+        elif type == "content_block_delta":
+            self._accumulating_body["content"][-1]["text"] += decoded_chunk.get(
+                "delta"
+            ).get("text")
+
+        elif self.has_finished(type, decoded_chunk):
+            self._accumulating_body["invocation_metrics"] = decoded_chunk.get(
+                "amazon-bedrock-invocationMetrics"
+            )
+            self._stream_done_callback(self._accumulating_body)
+
+    def has_finished(self, type, chunk):
+        if type and type == "message_stop":
+            return True
+
+        if "completionReason" in chunk and chunk.get("completionReason") == "FINISH":
+            return True
+
+        if "stop_reason" in chunk and chunk.get("stop_reason") is not None:
+            return True
+        return False

langtrace_python_sdk/version.py

@@ -1 +1 @@
-__version__ = "3.3.23"
+__version__ = "3.3.24"

{langtrace_python_sdk-3.3.23.dist-info → langtrace_python_sdk-3.3.24.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langtrace-python-sdk
-Version: 3.3.23
+Version: 3.3.24
 Summary: Python SDK for LangTrace
 Project-URL: Homepage, https://github.com/Scale3-Labs/langtrace-python-sdk
 Author-email: Scale3 Labs <engineering@scale3labs.com>

{langtrace_python_sdk-3.3.23.dist-info → langtrace_python_sdk-3.3.24.dist-info}/RECORD

@@ -3,8 +3,8 @@ examples/anthropic_example/__init__.py,sha256=03us1YuvAJR6fqXX8NH2kROBfTmyz7KzFV
 examples/anthropic_example/completion.py,sha256=3_YEZrt0BLVNJT_RbLXg6JGP2bweuc_HPC2MWR73tOM,713
 examples/autogen_example/__init__.py,sha256=UJgpzL2yOmzir-DAiGFR1PB1Zz3YcQvYcq5bCN8nl0A,158
 examples/autogen_example/main.py,sha256=6OJ73VCdHgVrqnekF1S1nK8mXCUABLbUUkQtr7wOCdw,2312
-examples/awsbedrock_examples/__init__.py,sha256=lVm2bgPNb4h83xRW2JbTsYf0DrZjrUr29g7-QleO4iQ,505
-examples/awsbedrock_examples/converse.py,sha256=-J0Hz6OSnIpO4idHol9nxL_Ri9GWwq95aqqi3wlCTRI,5317
+examples/awsbedrock_examples/__init__.py,sha256=sQ6dq-K2vunY9Y8r8uS-ZwrqKd8XnvV-BukWQ2Gm0fA,537
+examples/awsbedrock_examples/converse.py,sha256=41CyEla9vFT53ugLJYYHV6kchLEAfXZKMncoJHx1sds,5950
 examples/azureopenai_example/__init__.py,sha256=PaZM90r6VN4eSOXxb6wGsyhf9-RJCNqBypzk1Xa2GJI,271
 examples/azureopenai_example/completion.py,sha256=K_GeU0TfJ9lLDfW5VI0Lmm8_I0JXf1x9Qi83ImJ350c,668
 examples/cerebras_example/__init__.py,sha256=ydfNi0DjFMGVcfo79XVG3VEbzIrHo5wYBgSJzl_asNA,295
@@ -111,7 +111,7 @@ examples/weaviate_example/__init__.py,sha256=8JMDBsRSEV10HfTd-YC7xb4txBjD3la56sn
 examples/weaviate_example/query_text.py,sha256=wPHQTc_58kPoKTZMygVjTj-2ZcdrIuaausJfMxNQnQc,127162
 langtrace_python_sdk/__init__.py,sha256=VZM6i71NR7pBQK6XvJWRelknuTYUhqwqE7PlicKa5Wg,1166
 langtrace_python_sdk/langtrace.py,sha256=jvfrnkAxc41dmIvwNuVcXwCWyV2vqhGV5qAr5KiMHBA,13239
-langtrace_python_sdk/version.py,sha256=Q6BbHmX4u_5OYiZVZRbbX2mSRD5cDGkLiRAb8Pu2590,23
+langtrace_python_sdk/version.py,sha256=BYdEglI7p7QqwHU1MEOvXpBRXVFpeqhDQ3obeVJACVs,23
 langtrace_python_sdk/constants/__init__.py,sha256=3CNYkWMdd1DrkGqzLUgNZXjdAlM6UFMlf_F-odAToyc,146
 langtrace_python_sdk/constants/exporter/langtrace_exporter.py,sha256=EVCrouYCpY98f0KSaKr4PzNxPULTZZO6dSA_crEOyJU,106
 langtrace_python_sdk/constants/instrumentation/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -147,7 +147,7 @@ langtrace_python_sdk/instrumentation/autogen/patch.py,sha256=7Sq3C8Q5tT27UkWXd1S
 langtrace_python_sdk/instrumentation/aws_bedrock/__init__.py,sha256=IHqPgR1kdDvcoV1nUb-B21PaJ_qbQB0jc011Udi1ioU,96
 langtrace_python_sdk/instrumentation/aws_bedrock/bedrock_streaming_wrapper.py,sha256=_EMlxfBhjNsB-2TrP0ek-vx1CK5bpgjXPmQWNT9eBcE,1370
 langtrace_python_sdk/instrumentation/aws_bedrock/instrumentation.py,sha256=M7Dyw1tG18ptD0ctCT9yHzO4UyKCcS8JkHg9WVaz9Ck,1473
-langtrace_python_sdk/instrumentation/aws_bedrock/patch.py,sha256=cZ20LS0lNCKoEeeJLVEG5aduuGrapu_u4v8Cl-z6gLg,13406
+langtrace_python_sdk/instrumentation/aws_bedrock/patch.py,sha256=KKNp9gKzoj9YuLrqA0VMDe0FUSBzKtrbXiNN1hdciJM,17941
 langtrace_python_sdk/instrumentation/aws_bedrock/stream_body_wrapper.py,sha256=ENdhRVHBhdkIlJIc_tkf8ASijUzZdVZM-oonNLdNM48,1584
 langtrace_python_sdk/instrumentation/cerebras/__init__.py,sha256=9rHNg7PWcZ7a9jExQZlqwWPkvLGcPT-DGWot0_6Bx9k,92
 langtrace_python_sdk/instrumentation/cerebras/instrumentation.py,sha256=WPsaYxHanYnoxGjDk7fILGJSnSRUs_zoQ30JCyPBMII,1927
@@ -277,8 +277,8 @@ tests/pinecone/cassettes/test_query.yaml,sha256=b5v9G3ssUy00oG63PlFUR3JErF2Js-5A
 tests/pinecone/cassettes/test_upsert.yaml,sha256=neWmQ1v3d03V8WoLl8FoFeeCYImb8pxlJBWnFd_lITU,38607
 tests/qdrant/conftest.py,sha256=9n0uHxxIjWk9fbYc4bx-uP8lSAgLBVx-cV9UjnsyCHM,381
 tests/qdrant/test_qdrant.py,sha256=pzjAjVY2kmsmGfrI2Gs2xrolfuaNHz7l1fqGQCjp5_o,3353
-langtrace_python_sdk-3.3.23.dist-info/METADATA,sha256=KuERJmhF9i6f40sX0ty1TVq_6zaiAefIEmwTr78vkz8,15676
-langtrace_python_sdk-3.3.23.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-langtrace_python_sdk-3.3.23.dist-info/entry_points.txt,sha256=1_b9-qvf2fE7uQNZcbUei9vLpFZBbbh9LrtGw95ssAo,70
-langtrace_python_sdk-3.3.23.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
-langtrace_python_sdk-3.3.23.dist-info/RECORD,,
+langtrace_python_sdk-3.3.24.dist-info/METADATA,sha256=V0Xhmeqs2wlRHimItoq_pExSCrYd9JPtWQn-F_brXW8,15676
+langtrace_python_sdk-3.3.24.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+langtrace_python_sdk-3.3.24.dist-info/entry_points.txt,sha256=1_b9-qvf2fE7uQNZcbUei9vLpFZBbbh9LrtGw95ssAo,70
+langtrace_python_sdk-3.3.24.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
+langtrace_python_sdk-3.3.24.dist-info/RECORD,,