Added AWS Bedrock support as a provider

.env.example

@@ -24,6 +24,10 @@ ALIBABA_API_KEY=
 MOONSHOT_ENDPOINT=https://api.moonshot.cn/v1
 MOONSHOT_API_KEY=
 
+# AWS only needs the region and the default aws credentials. TODO: Add support for AWS profiles
+AWS_BEDROCK_REGION=us-west-2
+BEDROCK_API_KEY=
+
 # Set to false to disable anonymized telemetry
 ANONYMIZED_TELEMETRY=true
 

docker-compose.yml

@@ -35,6 +35,7 @@ services:
       - CHROME_DEBUGGING_HOST=localhost
     volumes:
       - /tmp/.X11-unix:/tmp/.X11-unix
+      - ~/.aws/credentials:/root/.aws/credentials
     restart: unless-stopped
     shm_size: '2gb'
     cap_add:

requirements.txt

@@ -3,3 +3,4 @@ pyperclip==1.9.0
 gradio==5.10.0
 json-repair
 langchain-mistralai==0.2.4
+langchain-aws

src/agent/custom_agent.py

@@ -89,9 +89,13 @@ def __init__(
             tool_calling_method: Optional[str] = 'auto',
             page_extraction_llm: Optional[BaseChatModel] = None,
             planner_llm: Optional[BaseChatModel] = None,
-            planner_interval: int = 1,  # Run planner every N steps
+            planner_interval: int = 1,  # Run planner every N steps,
+            placeholders: Optional[str] = None,
     ):
 
+        # make placeholders available to the class
+        self.placeholders = placeholders
+
         # Load sensitive data from environment variables
         env_sensitive_data = {}
         for key, value in os.environ.items():
@@ -239,6 +243,14 @@ async def get_next_action(self, input_messages: list[BaseMessage]) -> AgentOutpu
 
         ai_content = ai_content.replace("```json", "").replace("```", "")
         ai_content = repair_json(ai_content)
+
+
+        # Replace placeholders in ai_content with values from self.placeholders
+        for key, value in self.placeholders.items():
+            print(key, value)
+            ai_content = ai_content.replace(key, value)
+
+
         parsed_json = json.loads(ai_content)
         parsed: AgentOutput = self.AgentOutput(**parsed_json)
 

src/utils/default_config_settings.py

@@ -12,8 +12,26 @@ def default_config():
         "max_actions_per_step": 10,
         "use_vision": True,
         "tool_calling_method": "auto",
-        "llm_provider": "openai",
-        "llm_model_name": "gpt-4o",
+        "llm_provider": "bedrock",
+        "llm_model_name": "anthropic.claude-3-5-sonnet-20241022-v2:0",
+        "prerequisite": """
+
+import boto3
+
+session = boto3.Session(region_name="us-west-2")
+sagemaker_client = session.client("sagemaker")
+
+response = sagemaker_client.create_presigned_domain_url(
+    DomainId="d-8aldpksok8tq",
+    UserProfileName="arkaprav-ssh-test"
+    )
+
+PLACEHOLDERS={}
+PLACEHOLDERS["PLACEHOLDER_URL"] = response["AuthorizedUrl"]
+
+
+
+""",
         "llm_num_ctx": 32000,
         "llm_temperature": 1.0,
         "llm_base_url": "",
@@ -28,7 +46,7 @@ def default_config():
         "save_recording_path": "./tmp/record_videos",
         "save_trace_path": "./tmp/traces",
         "save_agent_history_path": "./tmp/agent_history",
-        "task": "go to google.com and type 'OpenAI' click search and give me the first url",
+        "task": "open PLACEHOLDER_URL and open space names test1234 else create it",
     }
 
 

src/utils/utils.py

@@ -4,12 +4,14 @@
 from pathlib import Path
 from typing import Dict, Optional
 import requests
+import boto3
 
 from langchain_anthropic import ChatAnthropic
 from langchain_mistralai import ChatMistralAI
 from langchain_google_genai import ChatGoogleGenerativeAI
 from langchain_ollama import ChatOllama
 from langchain_openai import AzureChatOpenAI, ChatOpenAI
+from langchain_aws import ChatBedrock
 import gradio as gr
 
 from .llm import DeepSeekR1ChatOpenAI, DeepSeekR1ChatOllama
@@ -21,7 +23,8 @@
     "deepseek": "DeepSeek",
     "google": "Google",
     "alibaba": "Alibaba",
-    "moonshot": "MoonShot"
+    "moonshot": "MoonShot",
+    "bedrock": "AWS Bedrock"
 }
 
 def get_llm_model(provider: str, **kwargs):
@@ -158,6 +161,21 @@ def get_llm_model(provider: str, **kwargs):
             base_url=os.getenv("MOONSHOT_ENDPOINT"),
             api_key=os.getenv("MOONSHOT_API_KEY"),
         )
+    elif provider == "bedrock":
+        region = kwargs.get("region", "") or os.getenv("AWS_BEDROCK_REGION", "us-west-2")
+
+        session = boto3.Session(region_name=region)
+        bedrock_runtime = session.client(
+            service_name="bedrock-runtime",
+            region_name=region,
+        )
+
+        model_id = kwargs.get("model_name", "anthropic.claude-3-5-sonnet-20241022-v2:0")
+
+        return ChatBedrock(
+            client=bedrock_runtime,
+            model_id=model_id,
+        )
     else:
         raise ValueError(f"Unsupported provider: {provider}")
 
@@ -172,6 +190,7 @@ def get_llm_model(provider: str, **kwargs):
     "mistral": ["pixtral-large-latest", "mistral-large-latest", "mistral-small-latest", "ministral-8b-latest"],
     "alibaba": ["qwen-plus", "qwen-max", "qwen-turbo", "qwen-long"],
     "moonshot": ["moonshot-v1-32k-vision-preview", "moonshot-v1-8k-vision-preview"],
+    "bedrock": ["anthropic.claude-3-5-sonnet-20241022-v2:0"]
 }
 
 # Callback to update the model name dropdown based on the selected provider