Skip to content

README.md

Latest commit

 

History

History
113 lines (80 loc) · 6.14 KB

README.md

File metadata and controls

113 lines (80 loc) · 6.14 KB
Raw

LaunchDarkly sample Python application

We've built a simple console application that demonstrates how LaunchDarkly's SDK works.

Below, you'll find the build procedure. For more comprehensive instructions, you can visit your Quickstart page or the Python reference guide.

This demo requires Python 3.10 or higher.

Build Instructions

This repository includes examples for OpenAI, Bedrock, and LangChain for multi-provider support. Depending on your preferred provider, you may have to take some additional steps.

General setup

  1. Create an AI Config using the key specified in each example, or copy the key of existing AI Config in your LaunchDarkly project that you want to evaluate.

  2. Set the environment variable LAUNCHDARKLY_SDK_KEY to your LaunchDarkly SDK key and LAUNCHDARKLY_AI_CONFIG_KEY to the AI Config key; otherwise, an AI Config of sample-ai-config or sample-ai-agent-config will be assumed for most examples.

    export LAUNCHDARKLY_SDK_KEY="1234567890abcdef"
export LAUNCHDARKLY_AI_CONFIG_KEY="sample-ai-config"

  3. Ensure you have Poetry installed.

Provider-Specific Setup

OpenAI setup (single provider)

  1. Install the required dependencies with poetry install -E openai or poetry install --all-extras.
  2. Set the environment variable OPENAI_API_KEY to your OpenAI key.
  3. On the command line, run poetry run openai-example.

Chat with observability (observability plugin example)

This example demonstrates how to use the LaunchDarkly observability SDK plugin to monitor AI chat operations. For more details, see the Python SDK observability reference.

  1. Install the required dependencies with poetry install -E observability or poetry install --all-extras.
  2. Set the environment variable for your AI provider (e.g., OPENAI_API_KEY), or configure your AI Config to use a different provider.
  3. Optionally, set service identification: 
    export SERVICE_NAME="my-ai-service"
export SERVICE_VERSION="1.0.0"
  4. On the command line, run poetry run chat-observability-example.

The observability plugin automatically captures and sends data to LaunchDarkly:

  • Observability tab: SDK operations, flag evaluations, error monitoring, logging, and distributed tracing
  • AI Config Monitoring tab: Token usage, duration, success/error rates, and custom metadata for filtering and analysis

View your data in the LaunchDarkly dashboard under Observability tabs.

Bedrock setup (single provider)

  1. Install the required dependencies with poetry install -E bedrock or poetry install --all-extras.
  2. Ensure the required AWS credentials can be auto-detected by the boto3 library. Examples might include environment variables, role providers, or shared credential files.
  3. On the command line, run poetry run bedrock-example.

Gemini setup (single provider)

  1. Install the required dependencies with poetry install -E gemini or poetry install --all-extras.
  2. Set the environment variable GOOGLE_API_KEY to your Google API key.
  3. On the command line, run poetry run gemini-example.

LangChain setup (multiple providers)

This example uses OpenAI, Bedrock, and Gemini LangChain provider packages. You can add additional LangChain providers using the poetry add command.

  1. Install all dependencies with poetry install -E langchain or poetry install --all-extras.
  2. Set up API keys for the providers you want to use.
  3. On the command line, run poetry run langchain-example

LangGraph setup (multiple providers, single agent)

  1. Install all dependencies with poetry install -E langgraph or poetry install --all-extras.
  2. Set up API keys for the providers you want to use.
  3. Optionally set this environment variable to use a different agent config: 
    export LAUNCHDARKLY_AGENT_CONFIG_KEY="sample-ai-agent-config"
  4. On the command line, run poetry run langgraph-agent-example.

LangGraph setup (multiple providers, multiple agents)

  1. Install all dependencies with poetry install -E langgraph or poetry install --all-extras.
  2. Set up API keys for the providers you want to use.
  3. Create an AI Config (Agent-based) using the keys below. Write a goal for each config and enable it with targeting rules.
  4. Optionally set these environment variables to use different agent configs: 
    export LAUNCHDARKLY_ANALYZER_CONFIG_KEY="code-review-analyzer"
export LAUNCHDARKLY_DOCUMENTATION_CONFIG_KEY="code-review-documentation"
  5. On the command line, run poetry run langgraph-multi-agent-example.

Judge setup (judge evaluation)

These examples demonstrate how to use LaunchDarkly's judge functionality to evaluate AI responses for accuracy, relevance, and other metrics.

  1. Install dependencies with poetry install -E langchain or poetry install --all-extras.
  2. Set up API keys for the provider you want to use (OpenAI, Bedrock, or Gemini).
  3. Create an AI Config for chat functionality.
  4. Create a Judge Config for evaluation.
  5. Set the required environment variables: 
    export LAUNCHDARKLY_SDK_KEY="your-sdk-key"
export LAUNCHDARKLY_AI_CONFIG_KEY="sample-ai-config"
export LAUNCHDARKLY_AI_JUDGE_KEY="sample-ai-judge-accuracy"
    Note: The default values are sample-ai-config for AI Config and sample-ai-judge-accuracy for Judge Config if not specified.
Available judge examples:
  • Chat with automatic judge evaluation (poetry run chat-judge-example): Uses the chat functionality which automatically evaluates responses with any judges defined in the AI config.
  • Direct judge evaluation (poetry run direct-judge-example): Evaluates specific input/output pairs using a judge configuration directly.