Run the EveryAPI gateway inside LangSmith
LangSmith is an observability and evaluation platform for LangChain applications and instrumented clients. Trace an OpenAI client that points at EveryAPI, and separately control which application inputs and outputs are sent to LangSmith.
One base URL line points LangSmith at EveryAPI
Default global Base URL / China-first Base URL: https://api.everyapi.ai/v1 / https://api-cn.everyapi.ai/v1
From zero to your first call
- Sign up for LangSmith
Create a LangSmith project or use an approved self-hosted deployment, then obtain credentials according to the current LangSmith documentation.
smith.langchain.com
- Install the SDK
Install a LangSmith SDK version compatible with your OpenAI or LangChain packages.
- Set the env vars
Configure LangSmith tracing and project settings through the current environment variables or client options. Store both LangSmith and EveryAPI keys as secrets.
- Point the OpenAI client at EveryAPI + wrap_openai
Instrument the OpenAI client whose base URL is https://api.everyapi.ai/v1 using the wrapper or LangChain integration supported by your installed SDK.
- View traces
Send a test request and confirm the intended telemetry is captured. Use the EveryAPI dashboard as the source of truth for EveryAPI usage and billing.
LangSmith × EveryAPI
The cost figures in LangSmith don't match the EveryAPI dashboard?
LangSmith estimates cost from upstream models' public pricing; EveryAPI uses unified pricing + matching, and your actual charge as a buyer is whatever the EveryAPI dashboard shows. It's the same situation as with Langfuse / Helicone.
What's the difference between wrap_openai and the traceable decorator?
wrap_openai adds tracing to a single OpenAI client instance; @traceable adds a span to any function (not just LLMs). You can use both together: wrap_openai captures the LLM call, and @traceable wraps your business function around it to form a trace tree.