Kong AI Gateway → OpenObserve

Automatically capture token usage, latency, and model metadata for every request routed through Kong AI Gateway. Kong acts as a transparent OpenAI-compatible proxy, so the standard OpenAI instrumentor captures spans with no additional configuration.

Prerequisites

Python 3.8+
Kong Gateway running in DB-less mode with an OpenAI proxy route (Docker recommended)
An OpenObserve account (cloud or self-hosted)
Your OpenObserve organisation ID and Base64-encoded auth token
An OpenAI API key

Installation

pip install openobserve openinference-instrumentation-openai openai python-dotenv

Configuration

Start Kong in DB-less mode with a declarative config that proxies to OpenAI:

# kong.yml
_format_version: "3.0"

services:
  - name: openai-service
    url: https://api.openai.com
    routes:
      - name: openai-route
        paths:
          - /openai
        strip_path: true

docker run -d --name kong \
  -e KONG_DATABASE=off \
  -e KONG_DECLARATIVE_CONFIG=/kong/declarative/kong.yml \
  -e KONG_PROXY_LISTEN=0.0.0.0:8000 \
  -v $(pwd)/kong.yml:/kong/declarative/kong.yml \
  -p 8000:8000 \
  kong:latest

Create a .env file in your project root:

OPENOBSERVE_URL=https://api.openobserve.ai/
OPENOBSERVE_ORG=your_org_id
OPENOBSERVE_AUTH_TOKEN=Basic <your_base64_token>
OPENAI_API_KEY=your-openai-api-key
KONG_GATEWAY_URL=http://localhost:8000/openai/v1

Instrumentation

Call OpenAIInstrumentor().instrument() before creating the OpenAI client. Point the client at the Kong proxy URL and pass your OpenAI API key directly.

from dotenv import load_dotenv
load_dotenv()

from openinference.instrumentation.openai import OpenAIInstrumentor
from openobserve import openobserve_init

OpenAIInstrumentor().instrument()
openobserve_init()

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["OPENAI_API_KEY"],
    base_url=os.environ.get("KONG_GATEWAY_URL", "http://localhost:8000/openai/v1"),
)

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Explain distributed tracing in one sentence."}],
    max_tokens=20,
)
print(response.choices[0].message.content)

What Gets Captured

Attribute	Description
`llm_system`	`openai`
`llm_model_name`	Resolved model returned by the API (e.g. `gpt-4o-mini-2024-07-18`)
`llm_request_parameters_model`	Model name sent in the request (e.g. `gpt-4o-mini`)
`llm_request_parameters_max_tokens`	`max_tokens` value from the request
`gen_ai_response_model`	Same as `llm_model_name`
`llm_observation_type`	`GENERATION`
`llm_token_count_prompt`	Prompt tokens consumed
`llm_token_count_completion`	Completion tokens returned
`llm_token_count_total`	Total tokens consumed
`llm_token_count_prompt_details_cache_read`	Cached prompt tokens
`llm_token_count_prompt_details_audio`	Audio prompt tokens
`llm_token_count_completion_details_reasoning`	Reasoning tokens
`llm_token_count_completion_details_audio`	Audio completion tokens
`llm_usage_tokens_input`	Input tokens (mirrors prompt count)
`llm_usage_tokens_output`	Output tokens (mirrors completion count)
`llm_usage_tokens_total`	Total tokens
`llm_usage_cost_input`	Estimated input cost in USD
`llm_usage_cost_output`	Estimated output cost in USD
`llm_invocation_parameters`	JSON string of model config sent with the request
`openinference_span_kind`	`LLM`
`operation_name`	`ChatCompletion`
`input_mime_type`	`application/json`
`output_mime_type`	`application/json`
`duration`	End-to-end latency including Kong proxy overhead
`span_status`	`OK` on success, `ERROR` on failure

Viewing Traces

Log in to OpenObserve and navigate to Traces
Spans appear with operation_name: ChatCompletion and llm_system: openai
Since Kong is a transparent proxy, spans look identical to direct OpenAI calls. Use service_name or a custom span attribute to tag Kong traffic separately
Use duration to measure Kong proxy overhead versus direct API latency

Kong Gateway trace in OpenObserve

Next Steps

With Kong AI Gateway instrumented, every proxied request is recorded in OpenObserve. From here you can monitor latency per route, track token usage across consumers, and set alerts on error rates.