dao-ai 0.0.20__py3-none-any.whl → 0.0.21__py3-none-any.whl

dao_ai/config.py

@@ -427,7 +427,8 @@ class GenieRoomModel(BaseModel, IsDatabricksResource):
     def as_resources(self) -> Sequence[DatabricksResource]:
         return [
             DatabricksGenieSpace(
-                genie_space_id=self.space_id, on_behalf_of_user=self.on_behalf_of_user
+                genie_space_id=value_of(self.space_id),
+                on_behalf_of_user=self.on_behalf_of_user,
             )
         ]
 
@@ -437,7 +438,7 @@ class GenieRoomModel(BaseModel, IsDatabricksResource):
         return self
 
 
-class VolumeModel(BaseModel, HasFullName):
+class VolumeModel(BaseModel, HasFullName, IsDatabricksResource):
     model_config = ConfigDict(use_enum_values=True, extra="forbid")
     schema_model: Optional[SchemaModel] = Field(default=None, alias="schema")
     name: str
@@ -455,6 +456,13 @@ class VolumeModel(BaseModel, HasFullName):
         provider: ServiceProvider = DatabricksProvider(w=w)
         provider.create_volume(self)
 
+    @property
+    def api_scopes(self) -> Sequence[str]:
+        return ["files.files", "catalog.volumes"]
+
+    def as_resources(self) -> Sequence[DatabricksResource]:
+        return []
+
 
 class VolumePathModel(BaseModel, HasFullName):
     model_config = ConfigDict(use_enum_values=True, extra="forbid")
@@ -683,7 +691,8 @@ class WarehouseModel(BaseModel, IsDatabricksResource):
     def as_resources(self) -> Sequence[DatabricksResource]:
         return [
             DatabricksSQLWarehouse(
-                warehouse_id=self.warehouse_id, on_behalf_of_user=self.on_behalf_of_user
+                warehouse_id=value_of(self.warehouse_id),
+                on_behalf_of_user=self.on_behalf_of_user,
             )
         ]
 
@@ -867,7 +876,7 @@ class PythonFunctionModel(BaseFunctionModel, HasFullName):
 
 class FactoryFunctionModel(BaseFunctionModel, HasFullName):
     model_config = ConfigDict(use_enum_values=True, extra="forbid")
-    args: Optional[dict[str, AnyVariable]] = Field(default_factory=dict)
+    args: Optional[dict[str, Any]] = Field(default_factory=dict)
     type: Literal[FunctionType.FACTORY] = FunctionType.FACTORY
 
     @property
@@ -879,6 +888,12 @@ class FactoryFunctionModel(BaseFunctionModel, HasFullName):
 
         return [create_factory_tool(self, **kwargs)]
 
+    @model_validator(mode="after")
+    def update_args(self):
+        for key, value in self.args.items():
+            self.args[key] = value_of(value)
+        return self
+
 
 class TransportType(str, Enum):
     STREAMABLE_HTTP = "streamable_http"
@@ -1262,12 +1277,12 @@ class AppModel(BaseModel):
             if len(self.agents) > 1:
                 default_agent: AgentModel = self.agents[0]
                 self.orchestration = OrchestrationModel(
-                    swarm=SupervisorModel(model=default_agent.model)
+                    supervisor=SupervisorModel(model=default_agent.model)
                 )
             elif len(self.agents) == 1:
                 default_agent: AgentModel = self.agents[0]
                 self.orchestration = OrchestrationModel(
-                    supervisor=SwarmModel(
+                    swarm=SwarmModel(
                         model=default_agent.model, default_agent=default_agent
                     )
                 )

dao_ai/memory/postgres.py

@@ -74,8 +74,17 @@ class AsyncPostgresPoolManager:
         async with cls._lock:
             for connection_key, pool in cls._pools.items():
                 try:
-                    await pool.close()
+                    # Use a short timeout to avoid blocking on pool closure
+                    await asyncio.wait_for(pool.close(), timeout=2.0)
                     logger.debug(f"Closed PostgreSQL pool: {connection_key}")
+                except asyncio.TimeoutError:
+                    logger.warning(
+                        f"Timeout closing pool {connection_key}, forcing closure"
+                    )
+                except asyncio.CancelledError:
+                    logger.warning(
+                        f"Pool closure cancelled for {connection_key} (shutdown in progress)"
+                    )
                 except Exception as e:
                     logger.error(f"Error closing pool {connection_key}: {e}")
             cls._pools.clear()
@@ -369,8 +378,27 @@ def _shutdown_pools():
 
 def _shutdown_async_pools():
     try:
-        asyncio.run(AsyncPostgresPoolManager.close_all_pools())
-        logger.debug("Successfully closed all asynchronous PostgreSQL pools")
+        # Try to get the current event loop first
+        try:
+            loop = asyncio.get_running_loop()
+            # If we're already in an event loop, create a task
+            loop.create_task(AsyncPostgresPoolManager.close_all_pools())
+            logger.debug("Scheduled async pool closure in running event loop")
+        except RuntimeError:
+            # No running loop, try to get or create one
+            try:
+                loop = asyncio.get_event_loop()
+                if loop.is_closed():
+                    # Loop is closed, create a new one
+                    loop = asyncio.new_event_loop()
+                    asyncio.set_event_loop(loop)
+                loop.run_until_complete(AsyncPostgresPoolManager.close_all_pools())
+                logger.debug("Successfully closed all asynchronous PostgreSQL pools")
+            except Exception as inner_e:
+                # If all else fails, just log the error
+                logger.warning(
+                    f"Could not close async pools cleanly during shutdown: {inner_e}"
+                )
     except Exception as e:
         logger.error(
             f"Error closing asynchronous PostgreSQL pools during shutdown: {e}"

dao_ai/models.py

@@ -5,6 +5,7 @@ from typing import Any, Generator, Optional, Sequence, Union
 
 from langchain_core.messages import AIMessage, AIMessageChunk, BaseMessage
 from langgraph.graph.state import CompiledStateGraph
+from langgraph.types import StateSnapshot
 from loguru import logger
 from mlflow import MlflowClient
 from mlflow.pyfunc import ChatAgent, ChatModel, ResponsesAgent
@@ -59,6 +60,113 @@ def get_latest_model_version(model_name: str) -> int:
     return latest_version
 
 
+async def get_state_snapshot_async(
+    graph: CompiledStateGraph, thread_id: str
+) -> Optional[StateSnapshot]:
+    """
+    Retrieve the state snapshot from the graph for a given thread_id asynchronously.
+
+    This utility function accesses the graph's checkpointer to retrieve the current
+    state snapshot, which contains the full state values and metadata.
+
+    Args:
+        graph: The compiled LangGraph state machine
+        thread_id: The thread/conversation ID to retrieve state for
+
+    Returns:
+        StateSnapshot if found, None otherwise
+    """
+    logger.debug(f"Retrieving state snapshot for thread_id: {thread_id}")
+    try:
+        # Check if graph has a checkpointer
+        if graph.checkpointer is None:
+            logger.debug("No checkpointer available in graph")
+            return None
+
+        # Get the current state from the checkpointer (use async version)
+        config: dict[str, Any] = {"configurable": {"thread_id": thread_id}}
+        state_snapshot: Optional[StateSnapshot] = await graph.aget_state(config)
+
+        if state_snapshot is None:
+            logger.debug(f"No state found for thread_id: {thread_id}")
+            return None
+
+        return state_snapshot
+
+    except Exception as e:
+        logger.warning(f"Error retrieving state snapshot for thread {thread_id}: {e}")
+        return None
+
+
+def get_state_snapshot(
+    graph: CompiledStateGraph, thread_id: str
+) -> Optional[StateSnapshot]:
+    """
+    Retrieve the state snapshot from the graph for a given thread_id.
+
+    This is a synchronous wrapper around get_state_snapshot_async.
+    Use this for backward compatibility in synchronous contexts.
+
+    Args:
+        graph: The compiled LangGraph state machine
+        thread_id: The thread/conversation ID to retrieve state for
+
+    Returns:
+        StateSnapshot if found, None otherwise
+    """
+    import asyncio
+
+    try:
+        loop = asyncio.get_event_loop()
+    except RuntimeError:
+        loop = asyncio.new_event_loop()
+        asyncio.set_event_loop(loop)
+
+    try:
+        return loop.run_until_complete(get_state_snapshot_async(graph, thread_id))
+    except Exception as e:
+        logger.warning(f"Error in synchronous state snapshot retrieval: {e}")
+        return None
+
+
+def get_genie_conversation_ids_from_state(
+    state_snapshot: Optional[StateSnapshot],
+) -> dict[str, str]:
+    """
+    Extract genie_conversation_ids from a state snapshot.
+
+    This function extracts the genie_conversation_ids dictionary from the state
+    snapshot values if present.
+
+    Args:
+        state_snapshot: The state snapshot to extract conversation IDs from
+
+    Returns:
+        A dictionary mapping genie space_id to conversation_id, or empty dict if not found
+    """
+    if state_snapshot is None:
+        return {}
+
+    try:
+        # Extract state values - these contain the actual state data
+        state_values: dict[str, Any] = state_snapshot.values
+
+        # Extract genie_conversation_ids from state values
+        genie_conversation_ids: dict[str, str] = state_values.get(
+            "genie_conversation_ids", {}
+        )
+
+        if genie_conversation_ids:
+            logger.debug(f"Retrieved genie_conversation_ids: {genie_conversation_ids}")
+            return genie_conversation_ids
+
+        return {}
+
+    except Exception as e:
+        logger.warning(f"Error extracting genie_conversation_ids from state: {e}")
+        return {}
+
+
 class LanggraphChatModel(ChatModel):
     """
     ChatModel that delegates requests to a LangGraph CompiledStateGraph.
@@ -257,7 +365,19 @@ class LanggraphResponsesAgent(ResponsesAgent):
             text=last_message.content, id=f"msg_{uuid.uuid4().hex[:8]}"
         )
 
-        custom_outputs = custom_inputs
+        # Retrieve genie_conversation_ids from state if available
+        custom_outputs: dict[str, Any] = custom_inputs.copy()
+        thread_id: Optional[str] = context.thread_id
+        if thread_id:
+            state_snapshot: Optional[StateSnapshot] = loop.run_until_complete(
+                get_state_snapshot_async(self.graph, thread_id)
+            )
+            genie_conversation_ids: dict[str, str] = (
+                get_genie_conversation_ids_from_state(state_snapshot)
+            )
+            if genie_conversation_ids:
+                custom_outputs["genie_conversation_ids"] = genie_conversation_ids
+
         return ResponsesAgentResponse(
             output=[output_item], custom_outputs=custom_outputs
         )
@@ -318,7 +438,22 @@ class LanggraphResponsesAgent(ResponsesAgent):
                                 **self.create_text_delta(delta=content, item_id=item_id)
                             )
 
-                custom_outputs = custom_inputs
+                # Retrieve genie_conversation_ids from state if available
+                custom_outputs: dict[str, Any] = custom_inputs.copy()
+                thread_id: Optional[str] = context.thread_id
+
+                if thread_id:
+                    state_snapshot: Optional[
+                        StateSnapshot
+                    ] = await get_state_snapshot_async(self.graph, thread_id)
+                    genie_conversation_ids: dict[str, str] = (
+                        get_genie_conversation_ids_from_state(state_snapshot)
+                    )
+                    if genie_conversation_ids:
+                        custom_outputs["genie_conversation_ids"] = (
+                            genie_conversation_ids
+                        )
+
                 # Yield final output item
                 yield ResponsesAgentStreamEvent(
                     type="response.output_item.done",

dao_ai/providers/databricks.py

@@ -226,6 +226,7 @@ class DatabricksProvider(ServiceProvider):
             config.resources.connections.values()
         )
         databases: Sequence[DatabaseModel] = list(config.resources.databases.values())
+        volumes: Sequence[VolumeModel] = list(config.resources.volumes.values())
 
         resources: Sequence[IsDatabricksResource] = (
             llms
@@ -236,6 +237,7 @@ class DatabricksProvider(ServiceProvider):
             + tables
             + connections
             + databases
+            + volumes
         )
 
         # Flatten all resources from all models into a single list

dao_ai/state.py

@@ -31,6 +31,9 @@ class SharedState(MessagesState):
     is_valid: bool  # message validation node
     message_error: str
 
+    # A mapping of genie space_id to conversation_id
+    genie_conversation_ids: dict[str, str]  # Genie
+
 
 class Context(BaseModel):
     user_id: str | None = None

dao_ai/tools/genie.py

@@ -1,20 +1,284 @@
+import bisect
+import json
+import logging
 import os
+import time
+from dataclasses import asdict, dataclass
+from datetime import datetime
 from textwrap import dedent
-from typing import Any, Callable, Optional
+from typing import Annotated, Any, Callable, Optional, Union
 
-from databricks_ai_bridge.genie import GenieResponse
-from databricks_langchain.genie import Genie
-from langchain_core.tools import StructuredTool
+import mlflow
+import pandas as pd
+from databricks.sdk import WorkspaceClient
+from langchain_core.messages import ToolMessage
+from langchain_core.tools import InjectedToolCallId, tool
+from langgraph.prebuilt import InjectedState
+from langgraph.types import Command
+from loguru import logger
+from pydantic import BaseModel, Field
 
-from dao_ai.config import (
-    GenieRoomModel,
-)
+from dao_ai.config import AnyVariable, CompositeVariableModel, GenieRoomModel, value_of
+
+MAX_TOKENS_OF_DATA: int = 20000
+MAX_ITERATIONS: int = 50
+DEFAULT_POLLING_INTERVAL_SECS: int = 2
+
+
+def _count_tokens(text):
+    import tiktoken
+
+    encoding = tiktoken.encoding_for_model("gpt-4o")
+    return len(encoding.encode(text))
+
+
+@dataclass
+class GenieResponse:
+    conversation_id: str
+    result: Union[str, pd.DataFrame]
+    query: Optional[str] = ""
+    description: Optional[str] = ""
+
+    def to_json(self):
+        return json.dumps(asdict(self))
+
+
+class GenieToolInput(BaseModel):
+    """Input schema for the Genie tool."""
+
+    question: str = Field(
+        description="The question to ask Genie about your data. Ask simple, clear questions about your tabular data. For complex analysis, ask multiple simple questions rather than one complex question."
+    )
+
+
+def _truncate_result(dataframe: pd.DataFrame) -> str:
+    query_result = dataframe.to_markdown()
+    tokens_used = _count_tokens(query_result)
+
+    # If the full result fits, return it
+    if tokens_used <= MAX_TOKENS_OF_DATA:
+        return query_result.strip()
+
+    def is_too_big(n):
+        return _count_tokens(dataframe.iloc[:n].to_markdown()) > MAX_TOKENS_OF_DATA
+
+    # Use bisect_left to find the cutoff point of rows within the max token data limit in a O(log n) complexity
+    # Passing True, as this is the target value we are looking for when _is_too_big returns
+    cutoff = bisect.bisect_left(range(len(dataframe) + 1), True, key=is_too_big)
+
+    # Slice to the found limit
+    truncated_df = dataframe.iloc[:cutoff]
+
+    # Edge case: Cannot return any rows because of tokens so return an empty string
+    if len(truncated_df) == 0:
+        return ""
+
+    truncated_result = truncated_df.to_markdown()
+
+    # Double-check edge case if we overshot by one
+    if _count_tokens(truncated_result) > MAX_TOKENS_OF_DATA:
+        truncated_result = truncated_df.iloc[:-1].to_markdown()
+    return truncated_result
+
+
+@mlflow.trace(span_type="PARSER")
+def _parse_query_result(resp, truncate_results) -> Union[str, pd.DataFrame]:
+    output = resp["result"]
+    if not output:
+        return "EMPTY"
+
+    columns = resp["manifest"]["schema"]["columns"]
+    header = [str(col["name"]) for col in columns]
+    rows = []
+
+    for item in output["data_array"]:
+        row = []
+        for column, value in zip(columns, item):
+            type_name = column["type_name"]
+            if value is None:
+                row.append(None)
+                continue
+
+            if type_name in ["INT", "LONG", "SHORT", "BYTE"]:
+                row.append(int(value))
+            elif type_name in ["FLOAT", "DOUBLE", "DECIMAL"]:
+                row.append(float(value))
+            elif type_name == "BOOLEAN":
+                row.append(value.lower() == "true")
+            elif type_name == "DATE" or type_name == "TIMESTAMP":
+                row.append(datetime.strptime(value[:10], "%Y-%m-%d").date())
+            elif type_name == "BINARY":
+                row.append(bytes(value, "utf-8"))
+            else:
+                row.append(value)
+
+        rows.append(row)
+
+    dataframe = pd.DataFrame(rows, columns=header)
+
+    if truncate_results:
+        query_result = _truncate_result(dataframe)
+    else:
+        query_result = dataframe.to_markdown()
+
+    return query_result.strip()
+
+
+class Genie:
+    def __init__(
+        self,
+        space_id,
+        client: WorkspaceClient | None = None,
+        truncate_results: bool = False,
+        polling_interval: int = DEFAULT_POLLING_INTERVAL_SECS,
+    ):
+        self.space_id = space_id
+        workspace_client = client or WorkspaceClient()
+        self.genie = workspace_client.genie
+        self.description = self.genie.get_space(space_id).description
+        self.headers = {
+            "Accept": "application/json",
+            "Content-Type": "application/json",
+        }
+        self.truncate_results = truncate_results
+        if polling_interval < 1 or polling_interval > 30:
+            raise ValueError("poll_interval must be between 1 and 30 seconds")
+        self.poll_interval = polling_interval
+
+    @mlflow.trace()
+    def start_conversation(self, content):
+        resp = self.genie._api.do(
+            "POST",
+            f"/api/2.0/genie/spaces/{self.space_id}/start-conversation",
+            body={"content": content},
+            headers=self.headers,
+        )
+        return resp
+
+    @mlflow.trace()
+    def create_message(self, conversation_id, content):
+        resp = self.genie._api.do(
+            "POST",
+            f"/api/2.0/genie/spaces/{self.space_id}/conversations/{conversation_id}/messages",
+            body={"content": content},
+            headers=self.headers,
+        )
+        return resp
+
+    @mlflow.trace()
+    def poll_for_result(self, conversation_id, message_id):
+        @mlflow.trace()
+        def poll_query_results(attachment_id, query_str, description):
+            iteration_count = 0
+            while iteration_count < MAX_ITERATIONS:
+                iteration_count += 1
+                resp = self.genie._api.do(
+                    "GET",
+                    f"/api/2.0/genie/spaces/{self.space_id}/conversations/{conversation_id}/messages/{message_id}/attachments/{attachment_id}/query-result",
+                    headers=self.headers,
+                )["statement_response"]
+                state = resp["status"]["state"]
+                if state == "SUCCEEDED":
+                    result = _parse_query_result(resp, self.truncate_results)
+                    return GenieResponse(
+                        conversation_id, result, query_str, description
+                    )
+                elif state in ["RUNNING", "PENDING"]:
+                    logging.debug("Waiting for query result...")
+                    time.sleep(self.poll_interval)
+                else:
+                    return GenieResponse(
+                        conversation_id,
+                        f"No query result: {resp['state']}",
+                        query_str,
+                        description,
+                    )
+            return GenieResponse(
+                conversation_id,
+                f"Genie query for result timed out after {MAX_ITERATIONS} iterations of {self.poll_interval} seconds",
+                query_str,
+                description,
+            )
+
+        @mlflow.trace()
+        def poll_result():
+            iteration_count = 0
+            while iteration_count < MAX_ITERATIONS:
+                iteration_count += 1
+                resp = self.genie._api.do(
+                    "GET",
+                    f"/api/2.0/genie/spaces/{self.space_id}/conversations/{conversation_id}/messages/{message_id}",
+                    headers=self.headers,
+                )
+                if resp["status"] == "COMPLETED":
+                    # Check if attachments key exists in response
+                    attachments = resp.get("attachments", [])
+                    if not attachments:
+                        # Handle case where response has no attachments
+                        return GenieResponse(
+                            conversation_id,
+                            result=f"Genie query completed but no attachments found. Response: {resp}",
+                        )
+
+                    attachment = next((r for r in attachments if "query" in r), None)
+                    if attachment:
+                        query_obj = attachment["query"]
+                        description = query_obj.get("description", "")
+                        query_str = query_obj.get("query", "")
+                        attachment_id = attachment["attachment_id"]
+                        return poll_query_results(attachment_id, query_str, description)
+                    if resp["status"] == "COMPLETED":
+                        text_content = next(
+                            (r for r in attachments if "text" in r), None
+                        )
+                        if text_content:
+                            return GenieResponse(
+                                conversation_id, result=text_content["text"]["content"]
+                            )
+                        return GenieResponse(
+                            conversation_id,
+                            result="Genie query completed but no text content found in attachments.",
+                        )
+                elif resp["status"] in {"CANCELLED", "QUERY_RESULT_EXPIRED"}:
+                    return GenieResponse(
+                        conversation_id, result=f"Genie query {resp['status'].lower()}."
+                    )
+                elif resp["status"] == "FAILED":
+                    return GenieResponse(
+                        conversation_id,
+                        result=f"Genie query failed with error: {resp.get('error', 'Unknown error')}",
+                    )
+                # includes EXECUTING_QUERY, Genie can retry after this status
+                else:
+                    logging.debug(f"Waiting...: {resp['status']}")
+                    time.sleep(self.poll_interval)
+            return GenieResponse(
+                conversation_id,
+                f"Genie query timed out after {MAX_ITERATIONS} iterations of {self.poll_interval} seconds",
+            )
+
+        return poll_result()
+
+    @mlflow.trace()
+    def ask_question(self, question: str, conversation_id: str | None = None):
+        logger.debug(
+            f"ask_question called with question: {question}, conversation_id: {conversation_id}"
+        )
+        if conversation_id:
+            resp = self.create_message(conversation_id, question)
+        else:
+            resp = self.start_conversation(question)
+        logger.debug(f"ask_question response: {resp}")
+        return self.poll_for_result(resp["conversation_id"], resp["message_id"])
 
 
 def create_genie_tool(
     genie_room: GenieRoomModel | dict[str, Any],
     name: Optional[str] = None,
     description: Optional[str] = None,
+    persist_conversation: bool = True,
+    truncate_results: bool = False,
+    poll_interval: int = DEFAULT_POLLING_INTERVAL_SECS,
 ) -> Callable[[str], GenieResponse]:
     """
     Create a tool for interacting with Databricks Genie for natural language queries to databases.
@@ -24,22 +288,33 @@ def create_genie_tool(
     answering questions about inventory, sales, and other structured retail data.
 
     Args:
-        space_id: Databricks workspace ID where Genie is configured. If None, tries to
-                get it from DATABRICKS_GENIE_SPACE_ID environment variable.
+        genie_room: GenieRoomModel or dict containing Genie configuration
+        name: Optional custom name for the tool. If None, uses default "genie_tool"
+        description: Optional custom description for the tool. If None, uses default description
 
     Returns:
-        A callable tool function that processes natural language queries through Genie
+        A LangGraph tool that processes natural language queries through Genie
     """
 
     if isinstance(genie_room, dict):
         genie_room = GenieRoomModel(**genie_room)
 
-    space_id: str = genie_room.space_id or os.environ.get("DATABRICKS_GENIE_SPACE_ID")
-
-    genie: Genie = Genie(
-        space_id=space_id,
-        client=genie_room.workspace_client,
+    space_id: AnyVariable = genie_room.space_id or os.environ.get(
+        "DATABRICKS_GENIE_SPACE_ID"
+    )
+    space_id: AnyVariable = genie_room.space_id or os.environ.get(
+        "DATABRICKS_GENIE_SPACE_ID"
     )
+    if isinstance(space_id, dict):
+        space_id = CompositeVariableModel(**space_id)
+    space_id = value_of(space_id)
+
+    # genie: Genie = Genie(
+    #     space_id=space_id,
+    #     client=genie_room.workspace_client,
+    #     truncate_results=truncate_results,
+    #     polling_interval=poll_interval,
+    # )
 
     default_description: str = dedent("""
     This tool lets you have a conversation and chat with tabular data about <topic>. You should ask
@@ -49,29 +324,66 @@ def create_genie_tool(
     Prefer to call this tool multiple times rather than asking a complex question.
     """)
 
-    if description is None:
-        description = default_description
+    tool_description: str = (
+        description if description is not None else default_description
+    )
+    tool_name: str = name if name is not None else "genie_tool"
 
-    doc_signature: str = dedent("""
-    Args:
-        question (str): The question to ask to ask Genie
+    function_docs = """
 
-    Returns:
-        response (GenieResponse): An object containing the Genie response
-    """)
+Args:
+question (str): The question to ask to ask Genie about your data. Ask simple, clear questions about your tabular data. For complex analysis, ask multiple simple questions rather than one complex question.
 
-    doc: str = description + "\n" + doc_signature
+Returns:
+GenieResponse: A response object containing the conversation ID and result from Genie."""
+    tool_description = tool_description + function_docs
 
-    async def genie_tool(question: str) -> GenieResponse:
-        # Use sync API for now since Genie doesn't support async yet
-        # Can be easily updated to await when Genie gets async support
-        response: GenieResponse = genie.ask_question(question)
-        return response
+    @tool(
+        name_or_callable=tool_name,
+        description=tool_description,
+    )
+    def genie_tool(
+        question: Annotated[str, "The question to ask Genie about your data"],
+        state: Annotated[dict, InjectedState],
+        tool_call_id: Annotated[str, InjectedToolCallId],
+    ) -> Command:
+        genie: Genie = Genie(
+            space_id=space_id,
+            client=genie_room.workspace_client,
+            truncate_results=truncate_results,
+            polling_interval=poll_interval,
+        )
 
-    name: str = name if name else genie_tool.__name__
+        """Process a natural language question through Databricks Genie."""
+        # Get existing conversation mapping and retrieve conversation ID for this space
+        conversation_ids: dict[str, str] = state.get("genie_conversation_ids", {})
+        existing_conversation_id: str | None = conversation_ids.get(space_id)
+        logger.debug(
+            f"Existing conversation ID for space {space_id}: {existing_conversation_id}"
+        )
 
-    structured_tool: StructuredTool = StructuredTool.from_function(
-        coroutine=genie_tool, name=name, description=doc, parse_docstring=False
-    )
+        response: GenieResponse = genie.ask_question(
+            question, conversation_id=existing_conversation_id
+        )
+
+        current_conversation_id: str = response.conversation_id
+        logger.debug(
+            f"Current conversation ID for space {space_id}: {current_conversation_id}"
+        )
+
+        # Update the conversation mapping with the new conversation ID for this space
+
+        update: dict[str, Any] = {
+            "messages": [ToolMessage(response.to_json(), tool_call_id=tool_call_id)],
+        }
+
+        if persist_conversation:
+            updated_conversation_ids: dict[str, str] = conversation_ids.copy()
+            updated_conversation_ids[space_id] = current_conversation_id
+            update["genie_conversation_ids"] = updated_conversation_ids
+
+        logger.debug(f"State update: {update}")
+
+        return Command(update=update)
 
-    return structured_tool
+    return genie_tool

{dao_ai-0.0.20.dist-info → dao_ai-0.0.21.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dao-ai
-Version: 0.0.20
+Version: 0.0.21
 Summary: DAO AI: A modular, multi-agent orchestration framework for complex AI workflows. Supports agent handoff, tool integration, and dynamic configuration via YAML.
 Project-URL: Homepage, https://github.com/natefleming/dao-ai
 Project-URL: Documentation, https://natefleming.github.io/dao-ai

{dao_ai-0.0.20.dist-info → dao_ai-0.0.21.dist-info}/RECORD

@@ -3,14 +3,14 @@ dao_ai/agent_as_code.py,sha256=kPSeDz2-1jRaed1TMs4LA3VECoyqe9_Ed2beRLB9gXQ,472
 dao_ai/catalog.py,sha256=sPZpHTD3lPx4EZUtIWeQV7VQM89WJ6YH__wluk1v2lE,4947
 dao_ai/chat_models.py,sha256=uhwwOTeLyHWqoTTgHrs4n5iSyTwe4EQcLKnh3jRxPWI,8626
 dao_ai/cli.py,sha256=Aez2TQW3Q8Ho1IaIkRggt0NevDxAAVPjXkePC5GPJF0,20429
-dao_ai/config.py,sha256=ZO5ei45gnhqg1BtD0R9aekJz4ClmiTw2GHhOk4Idil4,51958
+dao_ai/config.py,sha256=GeaM00wNlYecwe3HhqeG88Hprt0SvGg4HtC7g_m-v98,52386
 dao_ai/graph.py,sha256=gmD9mxODfXuvn9xWeBfewm1FiuVAWMLEdnZz7DNmSH0,7859
 dao_ai/guardrails.py,sha256=4TKArDONRy8RwHzOT1plZ1rhy3x9GF_aeGpPCRl6wYA,4016
 dao_ai/messages.py,sha256=xl_3-WcFqZKCFCiov8sZOPljTdM3gX3fCHhxq-xFg2U,7005
-dao_ai/models.py,sha256=Xb23U-lhDG8KyNRIijcJ4InluadlaGNy4rrYx7Cjgfg,26939
+dao_ai/models.py,sha256=8r8GIG3EGxtVyWsRNI56lVaBjiNrPkzh4HdwMZRq8iw,31689
 dao_ai/nodes.py,sha256=SSuFNTXOdFaKg_aX-yUkQO7fM9wvNGu14lPXKDapU1U,8461
 dao_ai/prompts.py,sha256=vpmIbWs_szXUgNNDs5Gh2LcxKZti5pHDKSfoClUcgX0,1289
-dao_ai/state.py,sha256=GwbMbd1TWZx1T5iQrEOX6_rpxOitlmyeJ8dMr2o_pag,1031
+dao_ai/state.py,sha256=_lF9krAYYjvFDMUwZzVKOn0ZnXKcOrbjWKdre0C5B54,1137
 dao_ai/types.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 dao_ai/utils.py,sha256=dkZTXNN6q0xwkrvSWdNq8937W2xGuLCRWRb6hRQM6kA,4217
 dao_ai/vector_search.py,sha256=jlaFS_iizJ55wblgzZmswMM3UOL-qOp2BGJc0JqXYSg,2839
@@ -19,22 +19,22 @@ dao_ai/hooks/core.py,sha256=ZShHctUSoauhBgdf1cecy9-D7J6-sGn-pKjuRMumW5U,6663
 dao_ai/memory/__init__.py,sha256=1kHx_p9abKYFQ6EYD05nuc1GS5HXVEpufmjBGw_7Uho,260
 dao_ai/memory/base.py,sha256=99nfr2UZJ4jmfTL_KrqUlRSCoRxzkZyWyx5WqeUoMdQ,338
 dao_ai/memory/core.py,sha256=g7chjBgVgx3iKjR2hghl0QL1j3802uIM_e7mgszur9M,4151
-dao_ai/memory/postgres.py,sha256=pxxMjGotgqjrKhx0lVR3EAjSZTQgBpiPZOB0-cyjprc,12505
+dao_ai/memory/postgres.py,sha256=aWHRLhPm-9ywjlQe2B4XSdLbeaiuVV88p4PiQJFNEWo,13924
 dao_ai/providers/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 dao_ai/providers/base.py,sha256=-fjKypCOk28h6vioPfMj9YZSw_3Kcbi2nMuAyY7vX9k,1383
-dao_ai/providers/databricks.py,sha256=fZ8mGotfA3W3t5yUej2xGmGHSybjBFYr895mOctT418,28203
+dao_ai/providers/databricks.py,sha256=PX5mBvZaIxSJIAHWVnPXsho1XvxcoR3Qs3I9UavFRsY,28306
 dao_ai/tools/__init__.py,sha256=ye6MHaJY7tUnJ8336YJiLxuZr55zDPNdOw6gm7j5jlc,1103
 dao_ai/tools/agent.py,sha256=WbQnyziiT12TLMrA7xK0VuOU029tdmUBXbUl-R1VZ0Q,1886
 dao_ai/tools/core.py,sha256=Kei33S8vrmvPOAyrFNekaWmV2jqZ-IPS1QDSvU7RZF0,1984
-dao_ai/tools/genie.py,sha256=GzV5lfDYKmzW_lSLxAsPaTwnzX6GxQOB1UcLaTDqpfY,2787
+dao_ai/tools/genie.py,sha256=1CbLViNQ3KnmDtHXuwqCPug7rEhCGvuHP1NgsY-AJZ0,15050
 dao_ai/tools/human_in_the_loop.py,sha256=yk35MO9eNETnYFH-sqlgR-G24TrEgXpJlnZUustsLkI,3681
 dao_ai/tools/mcp.py,sha256=auEt_dwv4J26fr5AgLmwmnAsI894-cyuvkvjItzAUxs,4419
 dao_ai/tools/python.py,sha256=XcQiTMshZyLUTVR5peB3vqsoUoAAy8gol9_pcrhddfI,1831
 dao_ai/tools/time.py,sha256=Y-23qdnNHzwjvnfkWvYsE7PoWS1hfeKy44tA7sCnNac,8759
 dao_ai/tools/unity_catalog.py,sha256=uX_h52BuBAr4c9UeqSMI7DNz3BPRLeai5tBVW4sJqRI,13113
 dao_ai/tools/vector_search.py,sha256=EDYQs51zIPaAP0ma1D81wJT77GQ-v-cjb2XrFVWfWdg,2621
-dao_ai-0.0.20.dist-info/METADATA,sha256=gWNRLhswz5sCe1vxbBQ6dGlgiObI9nI829Q5DQRqRRY,41380
-dao_ai-0.0.20.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-dao_ai-0.0.20.dist-info/entry_points.txt,sha256=Xa-UFyc6gWGwMqMJOt06ZOog2vAfygV_DSwg1AiP46g,43
-dao_ai-0.0.20.dist-info/licenses/LICENSE,sha256=YZt3W32LtPYruuvHE9lGk2bw6ZPMMJD8yLrjgHybyz4,1069
-dao_ai-0.0.20.dist-info/RECORD,,
+dao_ai-0.0.21.dist-info/METADATA,sha256=PG-eOltuUpaJf4lYEw-DoVy5BFT9LbMCfe8GanIV7zQ,41380
+dao_ai-0.0.21.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+dao_ai-0.0.21.dist-info/entry_points.txt,sha256=Xa-UFyc6gWGwMqMJOt06ZOog2vAfygV_DSwg1AiP46g,43
+dao_ai-0.0.21.dist-info/licenses/LICENSE,sha256=YZt3W32LtPYruuvHE9lGk2bw6ZPMMJD8yLrjgHybyz4,1069
+dao_ai-0.0.21.dist-info/RECORD,,