Baseten integration

.env.example

@@ -25,3 +25,7 @@ CARTESIA_API_KEY=your_cartesia_api_key_here
 
 # Anthropic API credentials
 ANTHROPIC_API_KEY=your_anthropic_api_key_here
+
+# Baseten API credentials
+BASETEN_API_KEY=your_baseten_api_key_here
+BASETEN_BASE_URL=your_baseten_base_url_here

agents-core/vision_agents/core/agents/agents.py

@@ -551,6 +551,8 @@ async def join(self, call: Call) -> "AgentSessionContextManager":
 
         # wait for conversation creation coro at the very end of the join flow
         self.conversation = await create_conversation_coro
+        # Provide conversation to the LLM so it can access the chat history.
+        self.llm.set_conversation(self.conversation)
         return AgentSessionContextManager(self, self._connection)
 
     async def finish(self):

agents-core/vision_agents/core/llm/llm.py

@@ -26,7 +26,7 @@
 from getstream.video.rtc.pb.stream.video.sfu.models.models_pb2 import Participant
 from getstream.video.rtc import AudioStreamTrack, PcmData
 from vision_agents.core.processors import Processor
-from vision_agents.core.utils.utils import parse_instructions
+from vision_agents.core.utils.utils import Instructions, parse_instructions
 from vision_agents.core.events.manager import EventManager
 from .function_registry import FunctionRegistry
 from .llm_types import ToolSchema, NormalizedToolCallItem
@@ -50,7 +50,6 @@ class LLM(abc.ABC):
     before_response_listener: BeforeCb
     after_response_listener: AfterCb
     agent: Optional["Agent"]
-    _conversation: Optional["Conversation"]
     function_registry: FunctionRegistry
 
     def __init__(self):
@@ -59,6 +58,9 @@ def __init__(self):
         self.events = EventManager()
         self.events.register_events_from_module(events)
         self.function_registry = FunctionRegistry()
+        self.instructions: Optional[str] = None
+        self.parsed_instructions: Optional[Instructions] = None
+        self._conversation: Optional[Conversation] = None
 
     async def warmup(self) -> None:
         """
@@ -187,9 +189,20 @@ def _attach_agent(self, agent: Agent):
         Attach agent to the llm
         """
         self.agent = agent
-        self._conversation = agent.conversation
         self._set_instructions(agent.instructions)
 
+    def set_conversation(self, conversation: Conversation):
+        """
+        Provide the Conversation object to the LLM to access the chat history.
+        To be called by the Agent after it joins the call.
+
+        Args:
+            conversation: a Conversation object
+
+        Returns:
+        """
+        self._conversation = conversation
+
     def _set_instructions(self, instructions: str):
         self.instructions = instructions
 

plugins/anthropic/tests/test_anthropic_llm.py

@@ -18,7 +18,7 @@ class TestClaudeLLM:
     async def llm(self) -> ClaudeLLM:
         """Test ClaudeLLM initialization with a provided client."""
         llm = ClaudeLLM(model="claude-sonnet-4-20250514")
-        llm._conversation = InMemoryConversation("be friendly", [])
+        llm.set_conversation(InMemoryConversation("be friendly", []))
         return llm
 
     @pytest.mark.asyncio
@@ -58,7 +58,7 @@ async def test_native_api(self, llm: ClaudeLLM):
     @pytest.mark.integration
     async def test_stream(self, llm: ClaudeLLM):
         streamingWorks = False
-        
+
         @llm.events.subscribe
         async def passed(event: LLMResponseChunkEvent):
             nonlocal streamingWorks
@@ -70,7 +70,6 @@ async def passed(event: LLMResponseChunkEvent):
 
         assert streamingWorks
 
-
     @pytest.mark.integration
     async def test_memory(self, llm: ClaudeLLM):
         await llm.simple_response(

plugins/aws/tests/test_aws.py

@@ -35,7 +35,7 @@ def assert_response_successful(self, response):
     async def llm(self) -> BedrockLLM:
         """Test BedrockLLM initialization with a provided client."""
         llm = BedrockLLM(model="qwen.qwen3-32b-v1:0", region_name="us-east-1")
-        llm._conversation = InMemoryConversation("be friendly", [])
+        llm.set_conversation(InMemoryConversation("be friendly", []))
         return llm
 
     @pytest.mark.asyncio

plugins/baseten/README.md

@@ -0,0 +1,26 @@
+# Baseten Plugin for Vision Agents
+
+LLM integrations for the models hosted on Baseten for Vision Agents framework.
+
+TODO
+
+## Installation
+
+```bash
+pip install vision-agents-plugins-baseten
+```
+
+## Usage
+
+```python
+
+```
+
+
+## Requirements
+- Python 3.10+
+- `openai`
+- GetStream SDK
+
+## License
+MIT

plugins/baseten/pyproject.toml

@@ -0,0 +1,36 @@
+[build-system]
+requires = ["hatchling", "hatch-vcs"]
+build-backend = "hatchling.build"
+
+[project]
+name = "vision-agents-plugins-baseten"
+dynamic = ["version"]
+description = "Baseten plugin for vision agents"
+readme = "README.md"
+requires-python = ">=3.10"
+license = "MIT"
+dependencies = [
+    "vision-agents",
+    "openai>=2.5.0",
+]
+
+[project.urls]
+Documentation = "https://visionagents.ai/"
+Website = "https://visionagents.ai/"
+Source = "https://github.com/GetStream/Vision-Agents"
+
+[tool.hatch.version]
+source = "vcs"
+raw-options = { root = "..", search_parent_directories = true, fallback_version = "0.0.0" }
+
+[tool.hatch.build.targets.wheel]
+packages = ["."]
+
+[tool.uv.sources]
+vision-agents = { workspace = true }
+
+[dependency-groups]
+dev = [
+    "pytest>=8.4.1",
+    "pytest-asyncio>=1.0.0",
+]

plugins/baseten/vision_agents/plugins/baseten/__init__.py

@@ -0,0 +1,4 @@
+from .baseten_vlm import BasetenVLM as VLM
+
+
+__all__ = ["VLM"]