Add support for Open Router API in Coder class

This commit introduces the ability to use the Open Router API in the Coder class if an `openrouter_api_key` is provided in the configuration. This allows for access to models with larger token limits. If the Open Router API key is not provided, the code falls back to using the OpenAI API key. The commit also refactors the `get_llm_model_name` method to support model selection based on the provided API key. 🚀🔑

Author: TechNickAI

aicodebot/coder.py

@@ -63,50 +63,71 @@ def get_llm(
         callbacks=None,
     ):
         config = read_config()
+        if "openrouter_api_key" in config:
+            # If the openrouter_api_key is set, use the Open Router API
+            # OpenRouter allows for access to many models that have larger token limits
+            openai.api_key = config["openrouter_api_key"]
+            openai.api_base = "https://openrouter.ai/api/v1"
+            headers = {"HTTP-Referer": "https://aicodebot.dev", "X-Title": "AICodeBot"}
+            tiktoken_model_name = model_name.replace("openai/", "")
+        else:
+            openai.api_key = config["openai_api_key"]
+            headers = None
+            tiktoken_model_name = model_name
 
         return ChatOpenAI(
-            openai_api_key=config["openai_api_key"],
+            openai_api_key=openai.api_key,
             model=model_name,
             max_tokens=response_token_size,
             verbose=verbose,
             temperature=temperature,
             streaming=streaming,
             callbacks=callbacks,
+            tiktoken_model_name=tiktoken_model_name,
+            model_kwargs={"headers": headers},
         )
 
     @staticmethod
-    def get_llm_model_name(token_size=0):
-        model_options = {
-            "gpt-4": 8192,
-            "gpt-4-32k": 32768,
-            "gpt-3.5-turbo": 4096,
-            "gpt-3.5-turbo-16k": 16384,
-        }
+    def get_llm_headers():
+        config = read_config()
+        if "openrouter_api_key" in config:
+            return {"HTTP-Referer": "https://aicodebot.dev", "X-Title": "AICodeBot"}
+        else:
+            return None
 
-        engines = Coder.get_openai_supported_engines()
+    @staticmethod
+    def get_llm_model_name(token_size=0):
+        config = read_config()
+        if "openrouter_api_key" in config:
+            model_options = {
+                "openai/gpt-4": 8192,
+                "openai/gpt-4-32k": 32768,
+                # Not working yet "anthropic/claude-2": 100_000,
+            }
+            supported_engines = model_options.keys()
+        else:
+            model_options = {
+                "gpt-4": 8192,
+                "gpt-4-32k": 32768,
+                "gpt-3.5-turbo": 4096,
+                "gpt-3.5-turbo-16k": 16384,
+            }
+            # Pull the list of supported engines from the OpenAI API for this key
+            supported_engines = Coder.get_openai_supported_engines()
 
         # For some unknown reason, tiktoken often underestimates the token size by ~10%, so let's buffer
         token_size = int(token_size * 1.1)
 
-        # Try to use GPT-4 if it is supported and the token size is small enough
-        if "gpt-4" in engines and token_size <= model_options["gpt-4"]:
-            logger.info(f"Using GPT-4 for token size {token_size}")
-            return "gpt-4"
-        elif "gpt-4-32k" in engines and token_size <= model_options["gpt-4-32k"]:
-            logger.info(f"Using GPT-4-32k for token size {token_size}")
-            return "gpt-4-32k"
-        elif token_size <= model_options["gpt-3.5-turbo"]:
-            logger.info(f"Using GPT-3.5-turbo for token size {token_size}")
-            return "gpt-3.5-turbo"
-        elif token_size <= model_options["gpt-3.5-turbo-16k"]:
-            logger.info(f"Using GPT-3.5-turbo-16k for token size {token_size}")
-            return "gpt-3.5-turbo-16k"
-        else:
-            logger.critical(
-                f"🛑 The context is too large ({token_size})"
-                "for the any of the models supported by your Open AI API key. 😞"
-            )
-            return None
+        for model, max_tokens in model_options.items():
+            if model in supported_engines and token_size <= max_tokens:
+                logger.info(f"Using {model} for token size {token_size}")
+                return model
+
+        logger.critical(f"The context is too large ({token_size}) for any of the models supported by your API key. 😞")
+        if "openrouter_api_key" not in config:
+            logger.critical("If you provide an Open Router API key, you can access larger models, up to 32k tokens")
+
+        return None
 
     @staticmethod
     def get_token_length(text, model="gpt-3.5-turbo"):