feat: add audio model classes and clean up unused function (#46)

* feat: enhance LLM response handling and add client function

- Updated `get_litellm_response` to include a default method for OpenAI style requests.
- Introduced a new function `get_llm_client` for LLM client initialization, currently a placeholder.

* feat: add audio model classes and clean up unused function

- Added `AudioModel` and `SpeechModel` to the `__init__.py` for better model organization.
- Removed the unused `get_llm_client` function from `utils.py` to streamline the codebase.

* chore: code format

* docs: update README with environment setup instructions and remove setup_guide.md

- Added instructions for setting environment variables `OPENAI_API_KEY`, `NERIF_DEFAULT_LLM_MODEL`, and `NERIF_DEFAULT_EMBEDDING_MODEL` in README.md.
- Provided an example of how to export these variables in a bash environment.
- Removed the outdated setup_guide.md file to streamline documentation.

README.md

@@ -19,6 +19,16 @@ Our goal is to make LLMs easier to use for developers, turning complex AI capabi
 ### Pre-requisite
 
 - Python >= 3.9
+- Environment variable `OPENAI_API_KEY` or other LLM API keys, for more details, see [here](https://nerif-ai.com/docs/nerif-environment)
+- Set default model and embedding model with `NERIF_DEFAULT_LLM_MODEL` and `NERIF_DEFAULT_EMBEDDING_MODEL`, for more details, see [here](https://nerif-ai.com/docs/nerif-environment)
+
+Example:
+
+```bash
+export OPENAI_API_KEY=`your_api_key`
+export NERIF_DEFAULT_LLM_MODEL=gpt-4o
+export NERIF_DEFAULT_EMBEDDING_MODEL=text-embedding-3-small
+```
 
 ### Install
 
@@ -33,6 +43,7 @@ from nerif.core import nerif
 from nerif.model import SimpleChatModel
 
 model = SimpleChatModel()
+# Default model is `gpt-4o`
 
 # Use nerif judge "natural language statement"
 if nerif("the sky is blue"):

nerif/model/__init__.py

@@ -1,8 +1,11 @@
+from .audio_model import AudioModel, SpeechModel
 from .model import LogitsChatModel, SimpleChatModel, SimpleEmbeddingModel, VisionModel
 
 __all__ = [
     "LogitsChatModel",
     "SimpleChatModel",
     "SimpleEmbeddingModel",
     "VisionModel",
+    "AudioModel",
+    "SpeechModel",
 ]

nerif/model/audio_model.py

@@ -0,0 +1,52 @@
+import base64
+from pathlib import Path
+from typing import Any, List, Optional
+
+from openai import OpenAI
+
+from ..utils import (
+    LOGGER,
+    NERIF_DEFAULT_LLM_MODEL,
+    OPENAI_MODEL,
+    MessageType,
+    NerifTokenCounter,
+    get_litellm_embedding,
+    get_litellm_response,
+    get_ollama_response,
+    get_sllm_response,
+    get_vllm_response,
+)
+
+
+class AudioModel:
+    """
+    A simple agent for audio tasks. (audio transcription, speech to text)
+    """
+
+    def __init__(self):
+        self.client = OpenAI()
+
+    def transcribe(self, file: Path):
+        with open(file, "rb") as audio_file:
+            transcription = self.client.audio.transcriptions.create(
+                model="whisper-1",
+                file=audio_file,
+            )
+        return transcription
+
+
+class SpeechModel:
+    """
+    A simple agent for speech tasks. (speech model, text to speech)
+    """
+
+    def __init__(self):
+        self.client = OpenAI()
+
+    def text_to_speech(self, text: str, voice: str = "alloy"):
+        response = self.client.audio.speech.create(
+            model="tts-1",
+            input=text,
+            voice=voice,
+        )
+        return response

nerif/utils/utils.py

@@ -185,7 +185,13 @@ def get_litellm_response(
             "messages": messages,
         }
     else:
-        raise ValueError(f"Model {model} not supported")
+        # default method: use openai style
+        kargs = {
+            "model": model,
+            "messages": messages,
+            "api_key": api_key,
+            "base_url": base_url,
+        }
 
     kargs["stream"] = stream
     kargs["temperature"] = temperature