Add support for Ollama, Palm, Claude-2, Cohere, Replicate Llama2, CodeLlama, Hugging Face (100+LLMs) - using LiteLLM

OAI_CONFIG_LIST_sample

@@ -17,5 +17,22 @@
         "api_base": "<your Azure OpenAI API base here>",
         "api_type": "azure",
         "api_version": "2023-07-01-preview"
+    },
+    {
+        "model": "command-nightly",
+        "api_key": "<your Cohere API key here>"
+    },
+    {
+        "model": "palm/chat-bison",
+        "api_key": "<your PaLM API key here>"
+    },
+    {
+        "model": "huggingface/glaiveai/glaive-coder-7b",
+        "api_key": "<your hugging face API key here>"
+    },
+    {
+        "model": "huggingface/glaiveai/glaive-coder-7b",
+        "api_key": "<your hugging face API key here>",
+        "api_base": "<your hugging face inference endpoint here>"
     }
 ]

autogen/oai/completion.py

@@ -12,6 +12,7 @@
 
 try:
     import openai
+    import litellm
     from openai.error import (
         ServiceUnavailableError,
         RateLimitError,
@@ -102,6 +103,28 @@ class Completion(openai_Completion):
         "prompt": "{prompt}",
     }
 
+    llm_lite_mapping = {
+        'code-davinci-002': 'davinci-002',
+        'gpt-3.5-turbo': 'gpt-3.5-turbo',
+        'gpt-3.5-turbo-0301': 'gpt-3.5-turbo-0301',
+        'gpt-3.5-turbo-0613': 'gpt-3.5-turbo-0613',
+        'gpt-3.5-turbo-16k': 'gpt-3.5-turbo-16k',
+        'gpt-3.5-turbo-16k-0613': 'gpt-3.5-turbo-16k-0613',
+        'gpt-35-turbo': 'gpt-3.5-turbo-instruct',
+        'gpt-4': 'gpt-4',
+        'gpt-4-0314': 'gpt-4-0314',
+        'gpt-4-0613': 'gpt-4-0613',
+        'gpt-4-32k': 'gpt-4-32k',
+        'gpt-4-32k-0314': 'gpt-4-32k-0314',
+        'gpt-4-32k-0613': 'gpt-4-32k-0613',
+        'text-ada-001': 'text-ada-001',
+        'text-babbage-001': 'text-babbage-001',
+        'text-curie-001': 'text-curie-001',
+        'text-davinci-002': 'davinci-002',
+        'text-davinci-003': 'text-davinci-003'
+    }
+
+
     seed = 41
     cache_path = f".cache/{seed}"
     # retry after this many seconds
@@ -185,28 +208,31 @@ def _get_response(cls, config: Dict, raise_on_ratelimit_or_timeout=False, use_ca
         """
         config = config.copy()
         openai.api_key_path = config.pop("api_key_path", openai.api_key_path)
+
+        model_name = config.get("model")
+        mapped_model_name = cls.llm_lite_mapping.get(model_name, model_name)  # Fallback to original if not found
+        config["model"] = mapped_model_name
+
         key = get_key(config)
         if use_cache:
-            response = cls._cache.get(key, None)
-            if response is not None and (response != -1 or not raise_on_ratelimit_or_timeout):
-                # print("using cached response")
-                cls._book_keeping(config, response)
-                return response
-        openai_completion = (
-            openai.ChatCompletion
-            if config["model"] in cls.chat_models or issubclass(cls, ChatCompletion)
-            else openai.Completion
-        )
+            try:
+                response = cls._cache.get(key, None)
+                if response is not None and (response != -1 or not raise_on_ratelimit_or_timeout):
+                    # print("using cached response")
+                    cls._book_keeping(config, response)
+                    return response
+            except:
+                pass
         start_time = time.time()
         request_timeout = cls.request_timeout
         max_retry_period = config.pop("max_retry_period", cls.max_retry_period)
         retry_wait_time = config.pop("retry_wait_time", cls.retry_wait_time)
         while True:
             try:
                 if "request_timeout" in config:
-                    response = openai_completion.create(**config)
+                    response = litellm.completion(**config)
                 else:
-                    response = openai_completion.create(request_timeout=request_timeout, **config)
+                    response = litellm.completion(request_timeout=request_timeout, **config)
             except (
                 ServiceUnavailableError,
                 APIConnectionError,
@@ -1121,4 +1147,4 @@ class ChatCompletion(Completion):
 
     default_search_space = Completion.default_search_space.copy()
     default_search_space["model"] = tune.choice(["gpt-3.5-turbo", "gpt-4"])
-    openai_completion_class = not ERROR and openai.ChatCompletion
+    openai_completion_class = not ERROR and openai.ChatCompletion

autogen/oai/openai_utils.py

@@ -313,6 +313,7 @@ def config_list_from_dotenv(
         FileNotFoundError: If the specified .env file does not exist.
         TypeError: If an unsupported type of configuration is provided in model_api_key_map.
     """
+
     if dotenv_file_path:
         dotenv_path = Path(dotenv_file_path)
         if dotenv_path.exists():
@@ -328,14 +329,14 @@ def config_list_from_dotenv(
     # Ensure the model_api_key_map is not None to prevent TypeErrors during key assignment.
     model_api_key_map = model_api_key_map or {}
 
-    # Ensure default models are always considered
-    default_models = ["gpt-4", "gpt-3.5-turbo"]
-
-    for model in default_models:
-        # Only assign default API key if the model is not present in the map.
-        # If model is present but set to invalid/empty, do not overwrite.
-        if model not in model_api_key_map:
-            model_api_key_map[model] = "OPENAI_API_KEY"
+    if not model_api_key_map:
+        # Ensure default models are always considered
+        default_models = ["gpt-4", "gpt-3.5-turbo"]
+        for model in default_models:
+            # Only assign default API key if the model is not present in the map.
+            # If model is present but set to invalid/empty, do not overwrite.
+            if model not in model_api_key_map:
+                model_api_key_map[model] = "OPENAI_API_KEY"
 
     env_var = []
     # Loop over the models and create configuration dictionaries
@@ -344,7 +345,9 @@ def config_list_from_dotenv(
             api_key_env_var = config
             config_dict = get_config(api_key=os.getenv(api_key_env_var))
         elif isinstance(config, dict):
-            api_key = os.getenv(config.get("api_key_env_var", "OPENAI_API_KEY"))
+            api_key = os.getenv(config.get("api_key_env_var", None))
+            if api_key is None:
+                api_key = config.get("api_key_default", "")  # Use the default API key if not found in environment
             config_without_key_var = {k: v for k, v in config.items() if k != "api_key_env_var"}
             config_dict = get_config(api_key=api_key, **config_without_key_var)
         else: