…nce port handling in CLI scripts

- Add additional ModelDeploy instances for Llama2-70B and Mixtral-8x7B models
- Allow specifying custom model deployments in GroqOpenaiAction constructor
- Set _model_deploys attribute based on provided or default model deployments
- Import Sequence type hint from typing module

…set to 0

The git diff shows the following changes:

1. Added several new OpenAI models to the ModelDeploy list, including various versions of GPT-3.5, GPT-4, text embeddings, moderation models, and text-to-speech models.

2. In the __init__ method, added a check to remove the "frequency_penalty" parameter from kwargs if its value is 0.0.

3. In the chat_stream method, moved the removal of the "stream" parameter from kwargs before validating the model. Also added the same check for "frequency_penalty" as in the __init__ method.

The commit comment summarizes these changes as adding support for more OpenAI models and removing the "frequency_penalty" parameter when its value is set to 0.

…tuple, including models like codellama, llama-2, mistral, mixtral, pplx, and sonar variants.

2. Modify the `__init__` method to accept an optional `model_deploys` parameter, allowing the default model list to be overridden when initializing the `PerplexityAction` class.

3. Update the `__init__` method to use the provided `model_deploys` sequence if available, otherwise fallback to the default `self.model_deploys` tuple.

4. Minor import change to include `Sequence` from the `typing` module.

Add languru-llm-pplx service for Perplexity (pplx) integration
Configure environment variables, ports, volumes, and dependencies
Associate with all and pplx profiles
Add languru-llm-groq service for Groq integration
Configure environment variables, ports, volumes, and dependencies
Associate with all and groq profiles