daviddwlee84 / PromptFlow

Playing some Prompt Flow stuff

Prompt Flow

Playing some Prompt Flow stuff

Getting Started

Streamlit

pip install -r requirements.txt

# Optional
cp example.env .env
# Fill-in your keys
# ...

streamlit run Overview.py

Prompt Flow

Chat Example

pf flow init --flow ./flows/simple_chat --type chat

# Azure OpenAI
# https://stackoverflow.com/questions/72236557/how-do-i-read-a-env-file-from-a-ps1-script
# https://learn.microsoft.com/en-us/powershell/module/microsoft.powershell.core/about/about_environment_variables?view=powershell-7.4
# https://stackoverflow.com/questions/7760013/why-does-continue-behave-like-break-in-a-foreach-object
# Load .env and set environment variable in Powershell
Get-Content .env | foreach {
    $name, $value = $_.split('=')
    if ([string]::IsNullOrWhiteSpace($name) -or $name.Contains('#')) {
        return
    }
    Set-Content env:\$name $value
}

pf connection create --file ./flows/simple_chat/azure_openai.yaml --set api_key=$env:AZURE_OPENAI_KEY api_base=$env:AZURE_OPENAI_ENDPOINT/openai/deployments/$env:AZURE_OPENAI_DEPLOYMENT_NAME/chat/completions?api-version=$env:AZURE_OPENAI_VERSION --set api_version=$env:AZURE_OPENAI_VERSION --name open_ai_connection

# Interactive test
pf flow test --flow ./flows/simple_chat --interactive

# or

# Single run with default input
pf flow test --flow ./flows/simple_chat

• Deploy a flow using development server — Prompt flow documentation

# Serving API (serve a simple chat web UI)
pf flow serve --source ./flows/simple_chat --port 28080 --host localhost

$response = Invoke-WebRequest -URI http://localhost:28080/score -Body '{"question":"How are you?"}' -Method POST  -ContentType "application/json"
echo $response.Content

curl http://localhost:28080/score --data '{"question":"How are you?"}' -X POST  -H "Content-Type: application/json"

Todo

• Create flow
• Call Prompt Flow scoring API
• Clean history button
• Docker Compose for running Prompt Flow & Streamlit at the same time (specify flow)
• Try build index locally
  • Faiss
    • Faiss Index Lookup — Prompt flow documentation
    • Faiss | 🦜️🔗 Langchain
    • Faiss: A library for efficient similarity search - Engineering at Meta
    • Welcome to Faiss Documentation — Faiss documentation
    • facebookresearch/faiss: A library for efficient similarity search and clustering of dense vectors.
    • Faiss Index Lookup tool in Azure Machine Learning prompt flow - Azure Machine Learning | Microsoft Learn

Resources

• microsoft/promptflow: Build high-quality LLM apps - from prototyping, testing to production deployment and monitoring.
• What is Azure Machine Learning prompt flow - Azure Machine Learning | Microsoft Learn
• Prompt flow for VS Code - Visual Studio Marketplace
• Prompt flow — Prompt flow documentation
  • Quick Start — Prompt flow documentation

Trouble Shooting

promptflow 0.1.0b7.post1 requires openai<0.28.0,>=0.27.8, but you have openai 1.13.3 which is incompatible. / AttributeError: module 'openai' has no attribute 'BadRequestError'

• v1.0.0 Migration Guide · openai/openai-python · Discussion #742
• Prompt flow supports OpenAI 1.x since v1.1.0. This may introduce breaking change.
• Frequency asked questions (FAQ) — Prompt flow documentation

pf.flow.test failed with UserErrorException: Exception: OpenAI API hits NotFoundError: Error code: 404 - {'error': {'code': 'DeploymentNotFound', 'message': 'The API deployment for this resource does not exist. If you created the deployment within the last 5 minutes, please wait a moment and try again.'}} [Error reference: https://platform.openai.com/docs/guides/error-codes/api-errors]

Basically your connection config are not set correctly

Seems the azure_openai.yaml is just a template, you don't need to change the file.