Merge branch 'main' of https://github.com/continuedev/continue into anthropic

1 files changed, 10 insertions, 4 deletions

diff --git a/continuedev/src/continuedev/libs/util/count_tokens.py b/continuedev/src/continuedev/libs/util/count_tokens.py
index 73be0717..1ca98fe6 100644
--- a/continuedev/src/continuedev/libs/util/count_tokens.py
+++ b/continuedev/src/continuedev/libs/util/count_tokens.py
@@ -1,15 +1,19 @@
 import json
 from typing import Dict, List, Union
 from ...core.main import ChatMessage
+from .templating import render_system_message
 import tiktoken
 
-aliases = {}
+aliases = {
+    "ggml": "gpt-3.5-turbo",
+}
 DEFAULT_MAX_TOKENS = 2048
 MAX_TOKENS_FOR_MODEL = {
     "gpt-3.5-turbo": 4096,
     "gpt-3.5-turbo-0613": 4096,
     "gpt-3.5-turbo-16k": 16384,
-    "gpt-4": 8192
+    "gpt-4": 8192,
+    "ggml": 2048
 }
 CHAT_MODELS = {
     "gpt-3.5-turbo", "gpt-3.5-turbo-16k", "gpt-4", "gpt-3.5-turbo-0613"
@@ -82,13 +86,15 @@ def compile_chat_messages(model: str, msgs: List[ChatMessage], max_tokens: int,
         for function in functions:
             prompt_tokens += count_tokens(model, json.dumps(function))
 
+    rendered_system_message = render_system_message(system_message)
+
     msgs = prune_chat_history(model,
-                              msgs, MAX_TOKENS_FOR_MODEL[model], prompt_tokens + max_tokens + count_tokens(model, system_message))
+                              msgs, MAX_TOKENS_FOR_MODEL[model], prompt_tokens + max_tokens + count_tokens(model, rendered_system_message))
     history = []
     if system_message:
         history.append({
             "role": "system",
-            "content": system_message
+            "content": rendered_system_message
         })
     history += [msg.to_dict(with_functions=functions is not None)
                 for msg in msgs]