dapr
/
components-contrib
mirror of https://github.com/dapr/components-contrib.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
components-contrib/tests/config/conversation/README.md

5.4 KiB
Raw Blame History

Conversation Component Conformance Tests

This directory contains conformance tests for all conversation components, including all langchaingo-based models.

Available Components

  • echo - Simple echo component for testing (no configuration needed)
  • openai - OpenAI GPT models (also supports Azure OpenAI)
  • anthropic - Anthropic Claude models
  • googleai - Google Gemini models
  • mistral - Mistral AI models
  • huggingface - HuggingFace models (using OpenAI compatibility layer)
  • ollama - Local Ollama models
  • bedrock - AWS Bedrock models

Running the Tests

To run the conformance tests:

# Run all conformance tests (will skip components without API keys)
go test -tags conftests ./tests/conformance -run TestConversationConformance -timeout 30s

# Run specific component test
go test -tags conftests ./tests/conformance -run TestConversationConformance/openai -timeout 30s

# Run with verbose output
go test -tags conftests ./tests/conformance -run TestConversationConformance -v -timeout 30s

Environment Variables

The tests will automatically skip components for which the required environment variables are not set. You can either set environment variables directly or use a .env file.

Using a .env file (Recommended)

  1. Copy the template file:
cp env.template .env
  1. Edit the .env file and add your API keys
  2. Run the tests - they will automatically load the environment variables from the .env file

Setting Environment Variables Directly

Alternatively, you can set the following environment variables to run the respective tests:

OpenAI

export OPENAI_API_KEY="your_openai_api_key"

Get your API key from: https://platform.openai.com/api-keys

Azure OpenAI

export AZURE_OPENAI_API_KEY="your_openai_api_key"
export AZURE_OPENAI_ENDPOINT="your_azureopenai_endpoint_here"
export AZURE_OPENAI_API_VERSION="your_azreopenai_api_version_here"

Get your configuration values from: https://ai.azure.com/

Anthropic

export ANTHROPIC_API_KEY="your_anthropic_api_key"

Get your API key from: https://console.anthropic.com/

Google AI

export GOOGLE_AI_API_KEY="your_google_ai_api_key"

Get your API key from: https://aistudio.google.com/app/apikey

Mistral

export MISTRAL_API_KEY="your_mistral_api_key"

Get your API key from: https://console.mistral.ai/

HuggingFace

export HUGGINGFACE_API_KEY="your_huggingface_api_key"

Get your API key from: https://huggingface.co/settings/tokens

AWS Bedrock

export AWS_ACCESS_KEY_ID="your_aws_access_key"
export AWS_SECRET_ACCESS_KEY="your_aws_secret_key"
export AWS_REGION="us-east-1"  # Optional, defaults to us-east-1

Get your credentials from AWS Console

Ollama

export OLLAMA_ENABLED="1"

Requires a local Ollama server running with the llama3.2:latest model available.

Test Configuration

Each component has its own configuration file in this directory:

  • echo/echo.yml - Echo component configuration
  • openai/openai.yml - OpenAI configuration with gpt-4o-mini model
  • anthropic/anthropic.yml - Anthropic configuration with Claude 3 Haiku
  • googleai/googleai.yml - Google AI configuration with Gemini 1.5 Flash
  • mistral/mistral.yml - Mistral configuration with open-mistral-7b
  • huggingface/huggingface.yml - HuggingFace configuration (uses OpenAI compatibility layer)
  • ollama/ollama.yml - Ollama configuration with llama3.2:latest
  • bedrock/bedrock.yml - AWS Bedrock configuration with Claude 3 Haiku

The configurations use cost-effective models where possible to minimize testing costs.

HuggingFace OpenAI Compatibility Layer

The HuggingFace component uses a workaround due to issues with the native HuggingFace implementation in langchaingo. Instead of using the HuggingFace SDK directly, it uses the OpenAI SDK with HuggingFace's OpenAI-compatible API endpoints.

How it works:

  • Model Selection: Any HuggingFace model can be used by specifying its full name (e.g., deepseek-ai/DeepSeek-R1-Distill-Qwen-32B)
  • Dynamic Endpoints: The endpoint URL is automatically generated based on the model name using the template: https://router.huggingface.co/hf-inference/models/{{model}}/v1
  • Custom Endpoints: You can override the endpoint by specifying a custom endpoint parameter
  • Authentication: Uses the same HuggingFace API key authentication

Example Configuration:

apiVersion: dapr.io/v1alpha1
kind: Component
metadata:
  name: huggingface
spec:
  type: conversation.huggingface
  version: v1
  metadata:
    - name: key
      value: "hf_your_api_key_here"
    - name: model
      value: "microsoft/DialoGPT-medium"
    # Optional: custom endpoint (auto-generated if not specified)
    - name: endpoint
      value: "https://router.huggingface.co/hf-inference/models/microsoft/DialoGPT-medium/v1"

This approach provides better reliability and compatibility while maintaining access to the full range of HuggingFace models.

Notes

  • The tests will automatically skip components when required environment variables are not set
  • Cost-effective models are used by default to minimize API costs
  • HuggingFace uses the OpenAI compatibility layer as a workaround due to langchaingo API issues
  • Ollama requires a local server and must be explicitly enabled
  • OpenAI component is tested for OpenAI and Azure
  • All tests include proper initialization and basic conversation functionality testing