agent: distinguish between client and gateway spans (#92)

Signed-off-by: Adrian Cole <adrian@tetrate.io>

Author: codefromthecrypt

inference-platforms/agent.py

@@ -49,10 +49,10 @@ async def run_agent(tools: list[Tool]):
 
 
 async def main():
-    mcp_url = os.getenv("MCP_URL", "https://mcp.kiwi.com")
     async with MCPServerStreamableHttp(
         {
-            "url": mcp_url,
+            "url": os.getenv("MCP_URL", "https://mcp.kiwi.com"),
+            "headers": dict(h.split("=", 1) for h in os.getenv("MCP_HEADERS", "").split(",") if h),
             "timeout": 30.0,
         },
         client_session_timeout_seconds=60.0,

inference-platforms/aigw/README.md

@@ -31,8 +31,7 @@ Start Ollama and your OpenTelemetry Collector via this repository's [README](../
 ## Run Envoy AI Gateway
 
 ```bash
-docker compose pull
-docker compose up --force-recreate --remove-orphans
+docker compose up --force-recreate --pull always --remove-orphans --wait -d
 ```
 
 Clean up when finished, like this:

inference-platforms/aigw/docker-compose.yml

@@ -34,6 +34,7 @@ services:
     env_file:
       - env.local
     environment:
+      - OTEL_SERVICE_NAME=aigw
       - OPENAI_BASE_URL=http://host.docker.internal:11434/v1
       - OTEL_EXPORTER_OTLP_ENDPOINT=http://host.docker.internal:4318
     ports:

inference-platforms/aigw/env.local

@@ -4,8 +4,12 @@ OPENAI_API_KEY=unused
 CHAT_MODEL=qwen3:0.6b
 AGENT_MODEL=qwen3:1.7b
 
+# Override as necessary for MCP proxying
+MCP_URL=https://mcp.kiwi.com
+MCP_HEADERS=
+
 # OpenTelemetry configuration
-OTEL_SERVICE_NAME=aigw
+OTEL_SERVICE_NAME=openai-agent
 OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318
 OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
 OTEL_TRACES_EXPORTER=otlp