From 3ced372ca2d8cf4571f16f75967794401ddfb9be Mon Sep 17 00:00:00 2001
From: Krish Dholakia <krrishdholakia@gmail.com>
Date: Mon, 14 Aug 2023 19:32:02 -0700
Subject: [PATCH 1/2] adding support for Palm, Cohere, HF Inference Endpoints,
 Replicate

---
 realtime_ai_character/llm/openai_llm.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/realtime_ai_character/llm/openai_llm.py b/realtime_ai_character/llm/openai_llm.py
index ef736345b..7798e45b8 100644
--- a/realtime_ai_character/llm/openai_llm.py
+++ b/realtime_ai_character/llm/openai_llm.py
@@ -5,7 +5,7 @@
 if os.getenv('OPENAI_API_TYPE') == 'azure':
     from langchain.chat_models import AzureChatOpenAI
 else:
-    from langchain.chat_models import ChatOpenAI
+    from langchain.chat_models import ChatLiteLLM
 from langchain.schema import BaseMessage, HumanMessage
 
 from realtime_ai_character.database.chroma import get_chroma
@@ -28,7 +28,7 @@ def __init__(self, model):
                 streaming=True
             )
         else:
-            self.chat_open_ai = ChatOpenAI(
+            self.chat_open_ai = ChatLiteLLM(
                 model=model,
                 temperature=0.5,
                 streaming=True

From 789a091c7404fe5aa23a88f0cf07efae1b0884f8 Mon Sep 17 00:00:00 2001
From: Krrish Dholakia <krrishdholakia@gmail.com>
Date: Sat, 2 Sep 2023 14:37:30 -0700
Subject: [PATCH 2/2] Update Readme with usage instructions

---
 README.md                               | 12 +++++++++++-
 realtime_ai_character/llm/__init__.py   |  4 ++--
 realtime_ai_character/llm/openai_llm.py |  7 ++++---
 requirements.txt                        |  1 +
 4 files changed, 18 insertions(+), 6 deletions(-)

diff --git a/README.md b/README.md
index 617ed0ee5..ed437e378 100644
--- a/README.md
+++ b/README.md
@@ -245,7 +245,7 @@ Note if you want to remotely connect to a RealChar server, SSL set up is require
 
 <br/>
 
-## 🆕! Anyscale and LangSmith integration
+## 🆕! Anyscale, LangSmith, LiteLLM integrations
 <details><summary>👇click me</summary>
 
 ### Anyscale
@@ -265,6 +265,16 @@ LANGCHAIN_PROJECT=YOUR_LANGCHAIN_PROJECT
 ```
 And it should work out of the box.
 
+
+### LiteLLM
+Use any model from Replicate, Cohere, AI21, Huggingface, TogetherAI, Aleph Alpha, and more - https://docs.litellm.ai/docs/completion/supported
+
+in your .env, just set the `LITELLM_API_KEY` to be the one for any of the providers you want to use
+```
+LITELLM_API_KEY="" # this is the api key for any of the providers supported by litellm
+```
+
+and this will be passed to litellm to use! 
 </details>
 
 <br/>
diff --git a/realtime_ai_character/llm/__init__.py b/realtime_ai_character/llm/__init__.py
index 8c557b302..436083304 100644
--- a/realtime_ai_character/llm/__init__.py
+++ b/realtime_ai_character/llm/__init__.py
@@ -1,8 +1,8 @@
 from realtime_ai_character.llm.base import AsyncCallbackAudioHandler, AsyncCallbackTextHandler, LLM
-
+import litellm
 
 def get_llm(model='gpt-3.5-turbo-16k') -> LLM:
-    if model.startswith('gpt'):
+    if model.startswith('gpt') or model in litellm.model_list or model.split("/")[0] in litellm.provider_list:
         from realtime_ai_character.llm.openai_llm import OpenaiLlm
         return OpenaiLlm(model=model)
     elif model.startswith('claude'):
diff --git a/realtime_ai_character/llm/openai_llm.py b/realtime_ai_character/llm/openai_llm.py
index ef736345b..7b658b225 100644
--- a/realtime_ai_character/llm/openai_llm.py
+++ b/realtime_ai_character/llm/openai_llm.py
@@ -5,7 +5,7 @@
 if os.getenv('OPENAI_API_TYPE') == 'azure':
     from langchain.chat_models import AzureChatOpenAI
 else:
-    from langchain.chat_models import ChatOpenAI
+    from langchain.chat_models import ChatLiteLLM
 from langchain.schema import BaseMessage, HumanMessage
 
 from realtime_ai_character.database.chroma import get_chroma
@@ -28,10 +28,11 @@ def __init__(self, model):
                 streaming=True
             )
         else:
-            self.chat_open_ai = ChatOpenAI(
+            self.chat_open_ai = ChatLiteLLM(
                 model=model,
                 temperature=0.5,
-                streaming=True
+                streaming=True,
+                api_key=os.getenv("LITELLM_API_KEY")
             )
         self.config = {
             "model": model,
diff --git a/requirements.txt b/requirements.txt
index 2802037d9..2a9a2ee03 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -36,3 +36,4 @@ typing
 uvicorn
 websockets
 google-auth
+litellm
\ No newline at end of file