Use large github action machine to run e2e tests (#230)

.github/workflows/e2e_tests.yml

@@ -8,7 +8,8 @@ on:
 
 jobs:
   test:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-latest-m
+    # runs-on: gh-large-150gb-ssd
 
     steps:
       - name: Checkout code
@@ -29,4 +30,5 @@ jobs:
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
           MISTRAL_API_KEY: ${{ secrets.MISTRAL_API_KEY }}
         run: |
-          cd e2e_tests && bash run_e2e_tests.sh
+          python -mvenv venv
+          source venv/bin/activate && cd e2e_tests && bash run_e2e_tests.sh

arch/Dockerfile

@@ -13,7 +13,7 @@ FROM envoyproxy/envoy:v1.31-latest as envoy
 #Build config generator, so that we have a single build image for both Rust and Python
 FROM python:3-slim as arch
 
-RUN apt-get update && apt-get install -y gettext-base && apt-get clean && rm -rf /var/lib/apt/lists/*
+RUN apt-get update && apt-get install -y gettext-base curl && apt-get clean && rm -rf /var/lib/apt/lists/*
 
 COPY --from=builder /arch/target/wasm32-wasi/release/prompt_gateway.wasm /etc/envoy/proxy-wasm-plugins/prompt_gateway.wasm
 COPY --from=builder /arch/target/wasm32-wasi/release/llm_gateway.wasm /etc/envoy/proxy-wasm-plugins/llm_gateway.wasm

arch/docker-compose.yaml

@@ -11,6 +11,11 @@ services:
       - /etc/ssl/cert.pem:/etc/ssl/cert.pem
       - ~/archgw_logs:/var/log/
     env_file:
-      - stage.env
+      - env.list
     extra_hosts:
       - "host.docker.internal:host-gateway"
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:10000/healthz"]
+      interval: 30s
+      timeout: 10s
+      retries: 3

arch/tools/build_cli.sh

@@ -3,7 +3,6 @@
 # Define paths
 source_schema="../arch_config_schema.yaml"
 source_compose="../docker-compose.yaml"
-source_stage_env="../stage.env"
 destination_dir="config"
 
 # Ensure the destination directory exists only if it doesn't already
@@ -15,7 +14,7 @@ fi
 # Copy the files
 cp "$source_schema" "$destination_dir/arch_config_schema.yaml"
 cp "$source_compose" "$destination_dir/docker-compose.yaml"
-cp "$source_stage_env" "$destination_dir/stage.env"
+touch "$destination_dir/env.list"
 
 # Print success message
 echo "Files copied successfully!"

arch/tools/cli/config_generator.py

@@ -34,7 +34,7 @@ def validate_and_render_schema():
     try:
         validate_prompt_config(ARCH_CONFIG_FILE, ARCH_CONFIG_SCHEMA_FILE)
     except Exception as e:
-        print(e)
+        print(str(e))
         exit(1)  # validate_prompt_config failed. Exit
 
     with open(ARCH_CONFIG_FILE, "r") as file:
@@ -73,7 +73,6 @@ def validate_and_render_schema():
 
     print("updated clusters", inferred_clusters)
 
-    config_yaml = add_secret_key_to_llm_providers(config_yaml)
     arch_llm_providers = config_yaml["llm_providers"]
     arch_tracing = config_yaml.get("tracing", {})
     arch_config_string = yaml.dump(config_yaml)

arch/tools/cli/main.py

@@ -58,6 +58,8 @@ def main(ctx, version):
         click.echo(f"archgw cli version: {get_version()}")
         ctx.exit()
 
+    log.info(f"Starting archgw cli version: {get_version()}")
+
     if ctx.invoked_subcommand is None:
         click.echo("""Arch (The Intelligent Prompt Gateway) CLI""")
         click.echo(logo)
@@ -68,7 +70,7 @@ def main(ctx, version):
 @click.option(
     "--service",
     default=SERVICE_ALL,
-    help="Optioanl parameter to specify which service to build. Options are model_server, archgw",
+    help="Optional parameter to specify which service to build. Options are model_server, archgw",
 )
 def build(service):
     """Build Arch from source. Must be in root of cloned repo."""
@@ -168,7 +170,7 @@ def up(file, path, service):
             arch_config_schema_file=arch_schema_config,
         )
     except Exception as e:
-        log.info(f"Exiting archgw up: {e}")
+        log.info(f"Exiting archgw up: validation failed")
         sys.exit(1)
 
     log.info("Starging arch model server and arch gateway")
@@ -178,6 +180,12 @@ def up(file, path, service):
     env = os.environ.copy()
     # check if access_keys are preesnt in the config file
     access_keys = get_llm_provider_access_keys(arch_config_file=arch_config_file)
+
+    # remove duplicates
+    access_keys = set(access_keys)
+    # remove the $ from the access_keys
+    access_keys = [item[1:] if item.startswith("$") else item for item in access_keys]
+
     if access_keys:
         if file:
             app_env_file = os.path.join(
@@ -186,6 +194,7 @@ def up(file, path, service):
         else:
             app_env_file = os.path.abspath(os.path.join(path, ".env"))
 
+        print(f"app_env_file: {app_env_file}")
         if not os.path.exists(
             app_env_file
         ):  # check to see if the environment variables in the current environment or not
@@ -205,7 +214,7 @@ def up(file, path, service):
                     env_stage[access_key] = env_file_dict[access_key]
 
     with open(
-        pkg_resources.resource_filename(__name__, "../config/stage.env"), "w"
+        pkg_resources.resource_filename(__name__, "../config/env.list"), "w"
     ) as file:
         for key, value in env_stage.items():
             file.write(f"{key}={value}\n")