weflayr 0.12.0

Drop-in instrumented wrappers for AI clients with zero-overhead telemetry

Weflayr SDK

Drop-in instrumented wrappers for AI clients. Add telemetry to your LLM calls in one line - no changes to your existing code structure required.

How it works

Weflayr wraps the official provider SDKs and automatically fires telemetry events to your Weflayr intake API before, after, and on error for every LLM call. Your application code stays identical.

your code  →  weflayr.sdk.openai.OpenAI  →  openai.OpenAI  →  OpenAI API
                       ↓
              Weflayr Intake API
              (before / after / error events)

Installation

pip install weflayr

Quickstart

Configuration

Set the following environment variables:

Variable	Description	Default
WEFLAYR_INTAKE_URL	Your Weflayr intake API base URL	https://api.weflayr.com
WEFLAYR_CLIENT_ID	Your client identifier	unknown_client
WEFLAYR_CLIENT_SECRET	Your bearer token	(empty)

export WEFLAYR_INTAKE_URL="https://api.weflayr.com"
export WEFLAYR_CLIENT_ID="my-app"
export WEFLAYR_CLIENT_SECRET="your-secret-token"

---

OpenAI

Drop-in replacement for openai.OpenAI and openai.AsyncOpenAI.

Sync

from weflayr.sdk.openai.client import OpenAI

client = OpenAI(api_key="sk-...")

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}],
)

print(response.choices[0].message.content)

Async

import asyncio
from weflayr.sdk.openai.client import AsyncOpenAI

client = AsyncOpenAI(api_key="sk-...")

async def main():
    response = await client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": "Hello!"}],
    )
    print(response.choices[0].message.content)

asyncio.run(main())

Tagging calls

Pass a tags dict to attach arbitrary metadata to your telemetry events:

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Summarize this."}],
    tags={"feature": "summarization", "user_id": "u_123"},
)

---

Mistral AI

Drop-in replacement for mistralai.client.Mistral.

Sync

from weflayr.sdk.mistralai.client import Mistral

client = Mistral(api_key="sk-...")

response = client.chat.complete(
    model="mistral-small-latest",
    messages=[{"role": "user", "content": "Hello!"}],
)

print(response.choices[0].message.content)

Async

import asyncio
from weflayr.sdk.mistralai.client import Mistral

client = Mistral(api_key="sk-...")

async def main():
    response = await client.chat.complete_async(
        model="mistral-small-latest",
        messages=[{"role": "user", "content": "Hello!"}],
    )
    print(response.choices[0].message.content)

asyncio.run(main())

Tagging calls

response = client.chat.complete(
    model="mistral-small-latest",
    messages=[{"role": "user", "content": "Translate this."}],
    tags={"feature": "translation", "env": "production"},
)

---

Telemetry events

For every LLM call, Weflayr sends up to three events to your intake API:

Event	When	Payload includes
<call>.before	Before the provider call	model, message_count, tags
<call>.after	On success	model, elapsed_ms, prompt_tokens, completion_tokens
<call>.error	On failure	model, elapsed_ms, error_type, error_message, status_code

Events are sent fire-and-forget in background threads — they never block your application or raise exceptions.

---

Advanced: per-client configuration

Override the intake URL, client ID, and bearer token directly on the client instead of using environment variables:

from weflayr.sdk.openai.client import OpenAI

client = OpenAI(
    api_key="sk-...",
    intake_url="https://api.weflayr.com",
    client_id="my-service",
    bearer_token="my-secret",
)

---

License

Elastic License 2.0 — free to use, modifications and redistribution not permitted.