Input Tracing for LangChain in Python

An example Python app 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.

In this case, our topic context consists of articles about authentication from our Secure by Design Hub included in langchain_input_tracing/data.

Prerequisites

• Python v3.12 or greater.
• pip v24.2 or uv v0.4.29.
• A Pangea account with Secure Audit Log enabled with the AI Audit Log Schema Config.
• An OpenAI API key.
• libmagic

Setup

git clone https://github.com/pangeacyber/langchain-python-input-tracing.git
cd langchain-python-input-tracing

Install libmagic

This is included in Windows via the python-magic-bin package

On macOS, you can install via this shell command:

brew install libmagic

If using pip:

python -m venv .venv
source .venv/bin/activate
pip install .

Or, if using uv:

uv sync
source .venv/bin/activate

The sample can then be executed with:

python -m langchain_input_tracing "What do you know about OAuth?"

Note: Because our context is limited to the authentication articles mentioned above, if you ask a question outside that context, you will get some variation of "I don't know."

Usage

Usage: python -m langchain_input_tracing [OPTIONS] PROMPT

Options:
  --model TEXT             OpenAI model.  [default: gpt-4o-mini; required]
  --audit-token SECRET     Pangea Secure Audit Log API token. May also be set
                           via the `PANGEA_AUDIT_TOKEN` environment variable.
                           [required]
  --audit-config-id TEXT   Pangea Secure Audit Log configuration ID.
  --pangea-domain TEXT     Pangea API domain. May also be set via the
                           `PANGEA_DOMAIN` environment variable.  [default:
                           aws.us.pangea.cloud; required]
  --openai-api-key SECRET  OpenAI API key. May also be set via the
                           `OPENAI_API_KEY` environment variable.  [required]
  --help                   Show this message and exit.

Example Input

python -m langchain_input_tracing "What do you know about OAuth?"

Sample Output

To secure your web app, embrace transparency in your security efforts and adopt Secure by Design principles, which focus on integrating security throughout the software development lifecycle. Implement application security hardening, robust user authentication, and secure default settings to enhance overall security. Additionally, consider using multi-factor authentication (MFA) and regularly educate users on best practices to protect their accounts.