From f55fb996f63e95361dcdbf96a889beee2ab78357 Mon Sep 17 00:00:00 2001
From: Jeremy Dyer <jdye64@gmail.com>
Date: Tue, 22 Oct 2024 17:19:21 -0400
Subject: [PATCH 1/2] Adjust NIM endpoint inference endpoints, and adding http
 endpoints

---
 helm/values.yaml          | 20 +++++++++++++++-----
 src/nv_ingest/api/main.py |  4 +++-
 2 files changed, 18 insertions(+), 6 deletions(-)

diff --git a/helm/values.yaml b/helm/values.yaml
index 9cc63cfb..32aef7d0 100644
--- a/helm/values.yaml
+++ b/helm/values.yaml
@@ -75,8 +75,10 @@ resources:
   limits:
     memory: 32Gi
     nvidia.com/gpu: 1
+    cpu: "36000m"
   requests:
     memory: 16Gi
+    cpu: "12000m"
 
 
 ## @param tmpDirSize [default: 8Gi] Specify the amount of space to reserve for temporary storage
@@ -253,15 +255,19 @@ redis:
 ## @skip envVars.REDIS_MORPHEUS_TASK_QUEUE
 ## @skip envVars.CACHED_GRPC_ENDPOINT
 ## @skip envVars.CACHED_HTTP_ENDPOINT
+## @skip envVars.CACHED_INFER_ENDPOINT
 ## @skip envVars.PADDLE_GRPC_ENDPOINT
 ## @skip envVars.PADDLE_HTTP_ENDPOINT
+## @skip envVars.PADDLE_INFER_ENDPOINT
 ## @skip envVars.YOLOX_GRPC_ENDPOINT
 ## @skip envVars.YOLOX_HTTP_ENDPOINT
+## @skip envVars.YOLOX_INFER_ENDPOINT
 ## @skip envVars.DEPLOT_GRPC_ENDPOINT
 ## @skip envVars.DEPLOT_HTTP_ENDPOINT
+## @skip envVars.DEPLOT_INFER_ENDPOINT
 envVars:
-  MESSAGE_CLIENT_HOST: "nv-ingest-ms-runtime"
-  MESSAGE_CLIENT_PORT: "7670"
+  MESSAGE_CLIENT_HOST: "nv-ingest-redis-master"
+  MESSAGE_CLIENT_PORT: "6379"
   REDIS_MORPHEUS_TASK_QUEUE: "morpheus_task_queue"
   NV_INGEST_DEFAULT_TIMEOUT_MS: "1234"
 
@@ -270,13 +276,17 @@ envVars:
   MINIO_BUCKET: nv-ingest
 
   CACHED_GRPC_ENDPOINT: nv-ingest-cached:8001
-  CACHED_HTTP_ENDPOINT: ""
+  CACHED_HTTP_ENDPOINT: http://nv-ingest-cached:8000/v1/infer
+  CACHED_INFER_ENDPOINT: grpc
   PADDLE_GRPC_ENDPOINT: nv-ingest-paddle:8001
-  PADDLE_HTTP_ENDPOINT: ""
+  PADDLE_HTTP_ENDPOINT: http://nv-ingest-paddle:8000/v1/infer
+  PADDLE_INFER_PROTOCOL: grpc
   YOLOX_GRPC_ENDPOINT: nv-ingest-yolox:8001
-  YOLOX_HTTP_ENDPOINT: ""
+  YOLOX_HTTP_ENDPOINT: http://nv-ingest-yolox:8000/v1/infer
+  YOLOX_INFER_PROTOCOL: grpc
   DEPLOT_GRPC_ENDPOINT: ""
   DEPLOT_HTTP_ENDPOINT: http://nv-ingest-deplot:8000/v1/chat/completions
+  DEPLOT_INFER_PROTOCOL: http
 
   EMBEDDING_NIM_ENDPOINT: "http://nv-ingest-embedding:8000/v1"
   MILVUS_ENDPOINT: "http://nv-ingest-milvus:19530"
diff --git a/src/nv_ingest/api/main.py b/src/nv_ingest/api/main.py
index 7800a61e..9beba4f6 100644
--- a/src/nv_ingest/api/main.py
+++ b/src/nv_ingest/api/main.py
@@ -9,6 +9,7 @@
 # its affiliates is strictly prohibited.
 
 import logging
+import os
 
 from opentelemetry import trace
 from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter
@@ -24,7 +25,8 @@
 trace.set_tracer_provider(TracerProvider())
 tracer = trace.get_tracer(__name__)
 
-exporter = OTLPSpanExporter(endpoint="otel-collector:4317", insecure=True)
+otel_endpoint = os.getenv("OTEL_EXPORTER_OTLP_ENDPOINT", "otel-collector:4317")
+exporter = OTLPSpanExporter(endpoint=otel_endpoint, insecure=True)
 span_processor = BatchSpanProcessor(exporter)
 trace.get_tracer_provider().add_span_processor(span_processor)
 

From e2955fcf589c34beb73093dd53d2d01fed0ce854 Mon Sep 17 00:00:00 2001
From: Jeremy Dyer <jdye64@gmail.com>
Date: Tue, 22 Oct 2024 17:23:58 -0400
Subject: [PATCH 2/2] Increase memory footprint

---
 helm/values.yaml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/helm/values.yaml b/helm/values.yaml
index 32aef7d0..d32d5575 100644
--- a/helm/values.yaml
+++ b/helm/values.yaml
@@ -73,12 +73,12 @@ replicaCount: 1
 ## @param resources.requests.memory [default: 16Gi] Specify request for memory
 resources:
   limits:
-    memory: 32Gi
+    memory: 90Gi
     nvidia.com/gpu: 1
     cpu: "36000m"
   requests:
-    memory: 16Gi
-    cpu: "12000m"
+    memory: 24Gi
+    cpu: "16000m"
 
 
 ## @param tmpDirSize [default: 8Gi] Specify the amount of space to reserve for temporary storage