Skip to main content

Documentation Index

Fetch the complete documentation index at: https://narev.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

Use your production traces as test datasets. Instead of guessing which model works better, test against the prompts your users are actually sending.

​
Prerequisites

  • An active LangSmith account with production traces
  • LangSmith API key (starts with lsv2*pt*<token>)
  • Production traces in your LangSmith project

​
Setup

  1. Create a new A/B test
  2. Navigate to Setup tab
  3. In the Data Source section select Import from traces
  4. Select LangSmith as provider
  5. Enter your credentials:
    • Project Name: Your LangSmith project identifier
    • API Key: Your LangSmith API key (starts with lsv2*pt*<token>)
  6. Select your Date Range (default: last 7 days)
  7. Click Save Project to import traces
Import traces from LangSmith interface

​
Common questions

No. Narev reads from your existing LangSmith project without any changes to your instrumentation.
Yes. Narev imports prompts and usage patterns from LangSmith, then lets you test any supported model provider on those same prompts.
No. Narev imports traces to run experiments but doesn’t write back to LangSmith.
Yes. Narev imports all types of LangSmith runs including LangChain chains, LangGraph agents, and individual LLM calls.
For most use cases, 50-500 traces provide statistically significant results. Start with 7 days of data by default. For high-variability workloads, import 30 days to capture edge cases.