sanitize bug

core_server/env.py

@@ -67,5 +67,5 @@
     OPENAI_MAIN_MODEL=(str, "gpt-4o"),
     OPENAI_SMALL_MODEL=(str, "gpt-4o-mini"),
     BEDROCK_MAIN_MODEL=(str, "anthropic.claude-3-5-sonnet-20240620-v1:0"),
-    BEDROCK_SMALL_MODEL=(str, "anthropic.claude-3-5-haiku-20241022-v1:0"),
+    BEDROCK_SMALL_MODEL=(str, "us.amazon.nova-micro-v1:0")
 )

nlp_scripts/model_prediction/llm/model_prediction.py

@@ -6,7 +6,7 @@
 from dataclasses import make_dataclass, field, dataclass
 from pydantic import BaseModel, Field, create_model
 from concurrent.futures import ThreadPoolExecutor
-from langchain_aws import ChatBedrock
+from langchain_aws import ChatBedrock, ChatBedrockConverse
 from langchain_core.prompts import ChatPromptTemplate
 from langchain_core.runnables import RunnableParallel
 from langchain_openai import ChatOpenAI
@@ -283,7 +283,7 @@ def __predict_entry(self, excerpt: str):
 
         def select_model_instance(model_name: str):
 
-            if self.model_family == "bedrock": model = ChatBedrock(model_id=model_name, temperature=0)
+            if self.model_family == "bedrock": model = ChatBedrockConverse(model_id=model_name, temperature=0)
             elif self.model_family == "openai": model = ChatOpenAI(model=model_name, temperature=0)
 
             return model

nlp_scripts/model_prediction/llm/utils.py

@@ -14,6 +14,7 @@ def manage_description(name, description):
 def _sanitize_keys_with_uniqueness(key, max_length: int = 64):
 
     sanitized = re.sub(r'[^a-zA-Z0-9_-]', '_', key)
+    sanitized = re.sub(r'^_+', '', sanitized) # pydantic model keys can't start with _
     sanitized = sanitized[:max_length]
 
     return sanitized