feat: hitl models support

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "uipath-llamaindex"
-version = "0.0.19"
+version = "0.0.20"
 description = "UiPath LlamaIndex SDK"
 readme = { file = "README.md", content-type = "text/markdown" }
 requires-python = ">=3.10"

samples/action-center-hitl-agent/.env.example

@@ -0,0 +1 @@
+OPENAI_API_KEY=xxx

samples/action-center-hitl-agent/agent.mermaid

@@ -0,0 +1,35 @@
+flowchart TD
+    step__done["_done"]:::stepStyle
+    step_aggregate_tool_results["aggregate_tool_results"]:::stepStyle
+    step_call_tool["call_tool"]:::stepStyle
+    step_init_run["init_run"]:::stepStyle
+    step_parse_agent_output["parse_agent_output"]:::stepStyle
+    step_run_agent_step["run_agent_step"]:::stepStyle
+    step_setup_agent["setup_agent"]:::stepStyle
+    event_StopEvent([<p>StopEvent</p>]):::stopEventStyle
+    event_ToolCallResult([<p>ToolCallResult</p>]):::defaultEventStyle
+    event_AgentInput([<p>AgentInput</p>]):::defaultEventStyle
+    event_ToolCall([<p>ToolCall</p>]):::defaultEventStyle
+    event_AgentWorkflowStartEvent([<p>AgentWorkflowStartEvent</p>]):::defaultEventStyle
+    event_AgentOutput([<p>AgentOutput</p>]):::defaultEventStyle
+    event_AgentSetup([<p>AgentSetup</p>]):::defaultEventStyle
+    event_StopEvent --> step__done
+    step_aggregate_tool_results --> event_AgentInput
+    step_aggregate_tool_results --> event_StopEvent
+    event_ToolCallResult --> step_aggregate_tool_results
+    step_call_tool --> event_ToolCallResult
+    event_ToolCall --> step_call_tool
+    step_init_run --> event_AgentInput
+    event_AgentWorkflowStartEvent --> step_init_run
+    step_parse_agent_output --> event_StopEvent
+    step_parse_agent_output --> event_ToolCall
+    event_AgentOutput --> step_parse_agent_output
+    step_run_agent_step --> event_AgentOutput
+    event_AgentSetup --> step_run_agent_step
+    step_setup_agent --> event_AgentSetup
+    event_AgentInput --> step_setup_agent
+    classDef stepStyle fill:#f2f0ff,line-height:1.2
+    classDef externalStyle fill:#f2f0ff,line-height:1.2
+    classDef defaultEventStyle fill-opacity:0
+    classDef stopEventStyle fill:#bfb6fc
+    classDef inputRequiredStyle fill:#f2f0ff,line-height:1.2

samples/action-center-hitl-agent/llama_index.json

@@ -0,0 +1,7 @@
+{
+  "dependencies": ["."],
+  "workflows": {
+    "agent": "main.py:workflow"
+  },
+  "env": ".env"
+}

samples/action-center-hitl-agent/main.py

@@ -0,0 +1,58 @@
+import json
+
+from dotenv import load_dotenv
+from llama_index.core.agent.workflow import AgentWorkflow
+from llama_index.core.workflow import (
+    Context,
+    HumanResponseEvent,
+)
+from llama_index.llms.openai import OpenAI
+
+from uipath_llamaindex.models import CreateActionEvent
+
+load_dotenv()
+
+llm = OpenAI(model="gpt-4o-mini")
+
+
+async def may_research_company(ctx: Context, company_name: str) -> bool:
+    """Find whether a company may be researched.
+    Args:
+        ctx (Context): The context in which this function is called (autopopulated).
+        company_name (str): Name of the company to be researched.
+    Returns:
+        bool: True if the company can be researched, False otherwise.
+    """
+    # emit an event to the external stream to be captured
+    agent_name = "Company researcher"
+    ctx.write_event_to_stream(
+        CreateActionEvent(
+            prefix="hitl escalation to research company",
+            app_name="generic_escalation_app",
+            title=f"Action required for {agent_name}",
+            data={
+                "AgentOutput": (f"May I perform a research on company {company_name}?"),
+                "AgentName": agent_name,
+            },
+            app_version=1,
+            app_folder_path="Shared",
+            # assignee="(optional)<assignee email>"
+        )
+    )
+
+    # wait until we see a HumanResponseEvent
+    hitl_response = await ctx.wait_for_event(HumanResponseEvent)
+    feedback = json.loads(hitl_response.response)
+    # act on the input from the event
+    if isinstance(feedback["Answer"], bool) and feedback["Answer"] is True:
+        return True
+    else:
+        return False
+
+
+# example user input {"user_msg": "research Uipath company"}
+workflow = AgentWorkflow.from_tools_or_functions(
+    [may_research_company()],
+    llm=llm,
+    system_prompt="You are a helpful assistant that can use tools to perform actions requested by user",
+)

samples/action-center-hitl-agent/pyproject.toml

@@ -0,0 +1,11 @@
+[project]
+name = "llama-action-center"
+version = "0.0.1"
+description = "UiPath LlamaIndex Simple HITL Agent"
+authors = [{ name = "John Doe", email = "john.doe@myemail.com" }]
+readme = { file = "README.md", content-type = "text/markdown" }
+requires-python = ">=3.10"
+dependencies = [
+    "uipath-llamaindex==0.0.20",
+    "llama-index-llms-openai>=0.2.2"
+]

samples/action-center-hitl-agent/uipath.json

@@ -0,0 +1,27 @@
+{
+  "entryPoints": [
+    {
+      "filePath": "agent",
+      "uniqueId": "ca9074cb-d758-483c-a88d-4346b3214fb7",
+      "type": "agent",
+      "input": {
+        "type": "object",
+        "properties": {
+          "hack": {
+            "type": "object"
+          }
+        },
+        "required": []
+      },
+      "output": {
+        "type": "object",
+        "properties": {},
+        "required": []
+      }
+    }
+  ],
+  "bindings": {
+    "version": "2.0",
+    "resources": []
+  }
+}

samples/multi-agent/.env.example

@@ -0,0 +1 @@
+OPENAI_API_KEY=xxx

samples/multi-agent/agent.mermaid

@@ -0,0 +1,35 @@
+flowchart TD
+    step__done["_done"]:::stepStyle
+    step_aggregate_tool_results["aggregate_tool_results"]:::stepStyle
+    step_call_tool["call_tool"]:::stepStyle
+    step_init_run["init_run"]:::stepStyle
+    step_parse_agent_output["parse_agent_output"]:::stepStyle
+    step_run_agent_step["run_agent_step"]:::stepStyle
+    step_setup_agent["setup_agent"]:::stepStyle
+    event_StopEvent([<p>StopEvent</p>]):::stopEventStyle
+    event_ToolCallResult([<p>ToolCallResult</p>]):::defaultEventStyle
+    event_AgentInput([<p>AgentInput</p>]):::defaultEventStyle
+    event_ToolCall([<p>ToolCall</p>]):::defaultEventStyle
+    event_AgentWorkflowStartEvent([<p>AgentWorkflowStartEvent</p>]):::defaultEventStyle
+    event_AgentOutput([<p>AgentOutput</p>]):::defaultEventStyle
+    event_AgentSetup([<p>AgentSetup</p>]):::defaultEventStyle
+    event_StopEvent --> step__done
+    step_aggregate_tool_results --> event_AgentInput
+    step_aggregate_tool_results --> event_StopEvent
+    event_ToolCallResult --> step_aggregate_tool_results
+    step_call_tool --> event_ToolCallResult
+    event_ToolCall --> step_call_tool
+    step_init_run --> event_AgentInput
+    event_AgentWorkflowStartEvent --> step_init_run
+    step_parse_agent_output --> event_StopEvent
+    step_parse_agent_output --> event_ToolCall
+    event_AgentOutput --> step_parse_agent_output
+    step_run_agent_step --> event_AgentOutput
+    event_AgentSetup --> step_run_agent_step
+    step_setup_agent --> event_AgentSetup
+    event_AgentInput --> step_setup_agent
+    classDef stepStyle fill:#f2f0ff,line-height:1.2
+    classDef externalStyle fill:#f2f0ff,line-height:1.2
+    classDef defaultEventStyle fill-opacity:0
+    classDef stopEventStyle fill:#bfb6fc
+    classDef inputRequiredStyle fill:#f2f0ff,line-height:1.2

samples/multi-agent/llama_index.json

@@ -0,0 +1,7 @@
+{
+  "dependencies": ["."],
+  "workflows": {
+    "agent": "main.py:workflow"
+  },
+  "env": ".env"
+}

samples/multi-agent/main.py

@@ -0,0 +1,49 @@
+import json
+
+from dotenv import load_dotenv
+from llama_index.core.agent.workflow import AgentWorkflow
+from llama_index.core.workflow import (
+    Context,
+    HumanResponseEvent,
+)
+from llama_index.llms.openai import OpenAI
+
+from uipath_llamaindex.models import InvokeProcessEvent
+
+load_dotenv()
+
+llm = OpenAI(model="gpt-4o-mini")
+
+
+async def may_research_company(ctx: Context, company_name: str) -> str:
+    """Find whether a company may be researcher.
+    Args:
+        ctx (Context): The context in which this function is called (autopopulated).
+        company_name (str): Name of the company to be researched.
+    Returns:
+        str: company report
+    """
+    # emit an event to the external stream to be captured
+    ctx.write_event_to_stream(
+        InvokeProcessEvent(
+            prefix="invoke langgraph researcher event",
+            name="my-first-uipath-agent",
+            # process_folder_path="(optional)<process-folder-path>",
+            input_arguments={
+                "topic": company_name,
+            },
+        )
+    )
+
+    # wait until we see a HumanResponseEvent
+    hitl_response = await ctx.wait_for_event(HumanResponseEvent)
+    feedback = json.loads(hitl_response.response)
+    # act on the input from the event
+    return feedback["report"]
+
+
+workflow = AgentWorkflow.from_tools_or_functions(
+    [may_research_company],
+    llm=llm,
+    system_prompt="You are a helpful assistant that can decide whether a company can be researched or not.",
+)

samples/multi-agent/pyproject.toml

@@ -0,0 +1,11 @@
+[project]
+name = "llama-multi-agent"
+version = "0.0.1"
+description = "UiPath LlamaIndex Simple HITL Agent"
+authors = [{ name = "John Doe", email = "john.doe@myemail.com" }]
+readme = { file = "README.md", content-type = "text/markdown" }
+requires-python = ">=3.10"
+dependencies = [
+    "uipath-llamaindex==0.0.20",
+    "llama-index-llms-openai>=0.2.2"
+]

samples/multi-agent/uipath.json

@@ -0,0 +1,27 @@
+{
+  "entryPoints": [
+    {
+      "filePath": "agent",
+      "uniqueId": "ca9074cb-d758-483c-a88d-4346b3214fb8",
+      "type": "agent",
+      "input": {
+        "type": "object",
+        "properties": {
+          "hack": {
+            "type": "object"
+          }
+        },
+        "required": []
+      },
+      "output": {
+        "type": "object",
+        "properties": {},
+        "required": []
+      }
+    }
+  ],
+  "bindings": {
+    "version": "2.0",
+    "resources": []
+  }
+}

samples/simple-hitl-agent/main.py

@@ -9,13 +9,21 @@
 llm = OpenAI(model="gpt-4o-mini")
 
 
-async def research_company(ctx: Context) -> str:
-    """Research a company."""
+async def may_research_company(ctx: Context, company_name: str) -> bool:
+    """Find whether a company may be researched.
+    Args:
+        ctx (Context): The context in which this function is called (autopopulated).
+        company_name (str): Name of the company to be researched.
+    Returns:
+        bool: True if the company can be researched, False otherwise.
+    """
     print("Researching company...")
 
     # emit an event to the external stream to be captured
     ctx.write_event_to_stream(
-        InputRequiredEvent(prefix="Are you sure you want to proceed?")
+        InputRequiredEvent(
+            prefix=f"May I perform a research on company {company_name}? \n (yes/no)"
+        )
     )
 
     # wait until we see a HumanResponseEvent
@@ -24,13 +32,13 @@ async def research_company(ctx: Context) -> str:
 
     # act on the input from the event
     if response.response.strip().lower() == "yes":
-        return "Research completed successfully."
+        return True
     else:
-        return "Research task aborted."
+        return False
 
 
 workflow = AgentWorkflow.from_tools_or_functions(
-    [research_company],
+    [may_research_company],
     llm=llm,
-    system_prompt="You are a helpful assistant that can research companies.",
+    system_prompt="You are a helpful assistant that can decide whether a company can be researched or not.",
 )

samples/simple-hitl-agent/pyproject.toml

@@ -1,11 +1,11 @@
 [project]
 name = "llama-simple-hitl-agent"
-version = "0.0.8"
+version = "0.0.1"
 description = "UiPath LlamaIndex Simple HITL Agent"
 authors = [{ name = "John Doe", email = "john.doe@myemail.com" }]
 readme = { file = "README.md", content-type = "text/markdown" }
 requires-python = ">=3.10"
 dependencies = [
-    "uipath-llamaindex>=0.0.18",
+    "uipath-llamaindex==0.0.20",
     "llama-index-llms-openai>=0.2.2"
 ]

samples/simple-hitl-agent/uipath.json

@@ -2,7 +2,7 @@
   "entryPoints": [
     {
       "filePath": "agent",
-      "uniqueId": "ca9074cb-d758-483c-a88d-4346b3214fb6",
+      "uniqueId": "ca9074cb-d758-483c-a88d-4346b3214fb7",
       "type": "agent",
       "input": {
         "type": "object",