Input Tracing for LangChain in Cloudflare Workers

An example Cloudflare Worker that demonstrates integrating Pangea's Secure Audit Log service into a LangChain app to maintain an audit log of context and prompts being sent to LLMs.

Prerequisites

• Node.js v22.
• A Pangea account with Secure Audit Log enabled with the AI Audit Log Schema Config.
• A Cloudflare account.

Setup

git clone https://github.com/pangeacyber/langchain-js-cloudflare-input-tracing.git
cd langchain-js-cloudflare-input-tracing
npm install
cp .dev.vars.example .dev.vars

Fill out the following environment variables in .dev.vars:

• CLOUDFLARE_ACCOUNT_ID: Cloudflare account ID.
• CLOUDFLARE_API_TOKEN: Cloudflare API token with access to Workers AI.
• PANGEA_AUDIT_TOKEN: Pangea Secure Audit Log API token.
• PANGEA_AUDIT_CONFIG_ID: Pangea Secure Audit Log configuration ID.

Context is fetched from a Vectorize index, which can be created like so:

npx wrangler vectorize create langchain-js-cloudflare-input-tracing --dimensions=768 --metric=cosine

Adding documents to the index is elided from this sample and is not strictly necessary to run the app.

Usage

A local version of the Worker can be started with:

npm start

Then prompts can be sent to the worker via an HTTP POST request like so:

curl -X POST http://localhost:8787 \
  -H 'Content-Type: application/json' \
  -d '"What is MFA?"'

This does not modify the input or output so it's transparent to the LLM and end user.

Audit logs can be viewed at the Secure Audit Log Viewer.