thanks codebuff

Makefile

@@ -320,11 +320,11 @@ release-rc: _check-uv-publish-token release-plan
 	: confirming release plan
 	@read -p "Press Enter to continue..."
 	@echo "Releasing..."
-	# cd humanlayer-ts && npm run build && npm publish
+	cd humanlayer-ts && npm run build && npm publish
-	# @$(MAKE) update-examples-ts-versions VERSION=$(current-ts-version)
+	@$(MAKE) update-examples-ts-versions VERSION=$(current-ts-version)
-	# @$(MAKE) smoke-test-examples-ts
+	@$(MAKE) smoke-test-examples-ts
-	# @$(MAKE) build-and-publish
+	@$(MAKE) build-and-publish
-	# @$(MAKE) update-examples-versions VERSION=$(current-py-version)
+	@$(MAKE) update-examples-versions VERSION=$(current-py-version)
 	@$(MAKE) smoke-test-examples-py
 
 	@echo "RC tested against staging, to proceed, update env vars to point at production"
@@ -340,20 +340,20 @@ release-rc: _check-uv-publish-token release-plan
 release-and-test-prod: _release-plan-versions _release-branch-check _production-env-check
 	@echo "Releasing..."
 	@echo "Publish TypeScript:"
-	#sed -i '' 's/$(current-ts-version)/$(new-version)/' humanlayer-ts/package.json
+	sed -i '' 's/$(current-ts-version)/$(new-version)/' humanlayer-ts/package.json
-	#cat humanlayer-ts/package.json | grep version
+	cat humanlayer-ts/package.json | grep version
-	#@read -p "Press Enter to continue..."
+	@read -p "Press Enter to continue..."
-	#cd humanlayer-ts && npm publish
+	cd humanlayer-ts && npm publish
-	#@$(MAKE) update-examples-ts-versions VERSION=$(new-version)
+	@$(MAKE) update-examples-ts-versions VERSION=$(new-version)
-	#@$(MAKE) smoke-test-examples-ts
+	@$(MAKE) smoke-test-examples-ts
 
-	#@echo "Publish Python:"
+	@echo "Publish Python:"
-	#sed -i '' 's/$(current-py-version)/$(new-version)/' pyproject.toml
+	sed -i '' 's/$(current-py-version)/$(new-version)/' pyproject.toml
-	#cat pyproject.toml | grep version
+	cat pyproject.toml | grep version
-	#@read -p "Press Enter to continue..."
+	@read -p "Press Enter to continue..."
-	#@$(MAKE) build-and-publish
+	@$(MAKE) build-and-publish
-	#@$(MAKE) update-examples-versions VERSION=$(new-version)
+	@$(MAKE) update-examples-versions VERSION=$(new-version)
-	#@$(MAKE) smoke-test-examples-py
+	@$(MAKE) smoke-test-examples-py
 
 	@echo "Finalize:"
 	#git commit -am 'release: v$(current-ts-version)' && git push upstream release-$(new-version)
@@ -367,3 +367,10 @@ release-and-test-prod: _release-plan-versions _release-branch-check _production-
 	git push upstream release-$(new-version)
 
 	hub compare
+
+.PHONY: check-local
+check-local:
+	@if [[ $$(git rev-parse --abbrev-ref HEAD) == local/* ]]; then \
+		echo "blocking push of local branch"; \
+		exit 1; \
+	fi

examples/langchain/00a-math_example_cli.py

@@ -56,6 +56,7 @@ prompt = ChatPromptTemplate.from_messages(
 agent = create_tool_calling_agent(llm, tools, prompt)
 agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
 
+
 def main() -> None:
     result = agent_executor.invoke({"input": "multiply 2 and 5, then add 32 to the result"})
     print("\n\n----------Result----------\n\n")

examples/langchain/01-math_example_allowed_repliers.py

@@ -1,7 +1,8 @@
 from dotenv import load_dotenv
-from langchain.agents import AgentType, initialize_agent
+from langchain_core.prompts import ChatPromptTemplate
 from langchain_openai import ChatOpenAI
-from langchain.tools import StructuredTool
+import langchain.tools
+from langchain.agents import AgentExecutor, create_tool_calling_agent
 
 from humanlayer import ContactChannel, HumanLayer, SlackContactChannel
 
@@ -43,19 +44,30 @@ def multiply(a: float, b: float) -> float:
 
 
 tools = [
-    StructuredTool.from_function(add),
+    langchain.tools.StructuredTool.from_function(add),
-    StructuredTool.from_function(multiply),
+    langchain.tools.StructuredTool.from_function(multiply),
 ]
 
 llm = ChatOpenAI(model="gpt-4", temperature=0)
-agent = initialize_agent(
+
-    tools=tools,
+# Prompt for creating Tool Calling Agent
-    llm=llm,
+prompt = ChatPromptTemplate.from_messages(
-    agent=AgentType.OPENAI_FUNCTIONS,
+    [
-    verbose=True,
+        (
+            "system",
+            "You are a helpful assistant.",
+        ),
+        ("placeholder", "{chat_history}"),
+        ("human", "{input}"),
+        ("placeholder", "{agent_scratchpad}"),
+    ]
 )
 
+# Construct the Tool Calling Agent
+agent = create_tool_calling_agent(llm, tools, prompt)
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+
 if __name__ == "__main__":
-    result = agent.run("What is (4 + 5) * 3?")
+    result = agent_executor.invoke({"input": "What is (4 + 5) * 3?"})
     print("\n\n----------Result----------\n\n")
     print(result)

examples/langchain/02-customer_email.py

@@ -1,11 +1,11 @@
 from humanlayer import ContactChannel, SlackContactChannel
-from langchain.agents import AgentType, initialize_agent
+from langchain_core.prompts import ChatPromptTemplate
 import langchain_core.tools as langchain_tools
 from langchain_openai import ChatOpenAI
+from langchain.agents import AgentExecutor, create_tool_calling_agent
 
 from dotenv import load_dotenv
 
-
 from humanlayer.core.approval import HumanLayer
 
 load_dotenv()
@@ -23,14 +23,12 @@ hl = HumanLayer(
 )
 
 task_prompt = """
-
 You are the email onboarding assistant. You check on the progress customers
 are making and then based on that info, you
 send friendly and encouraging emails to customers to help them fully onboard
 into the product.
 
 Your task is to send an email to the customer danny@example.com
-
 """
 
 
@@ -56,15 +54,25 @@ tools = [
 ]
 
 llm = ChatOpenAI(model="gpt-4o", temperature=0)
-agent = initialize_agent(
+
-    tools=tools,
+# Prompt for creating Tool Calling Agent
-    llm=llm,
+prompt = ChatPromptTemplate.from_messages(
-    agent=AgentType.OPENAI_FUNCTIONS,
+    [
-    verbose=True,
+        (
-    handle_parsing_errors=True,
+            "system",
+            "You are a helpful assistant.",
+        ),
+        ("placeholder", "{chat_history}"),
+        ("human", "{input}"),
+        ("placeholder", "{agent_scratchpad}"),
+    ]
 )
 
+# Construct the Tool Calling Agent
+agent = create_tool_calling_agent(llm, tools, prompt)
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+
 if __name__ == "__main__":
-    result = agent.run(task_prompt)
+    result = agent_executor.invoke({"input": task_prompt})
     print("\n\n----------Result----------\n\n")
     print(result)

examples/langchain/04-human_as_tool_linkedin.py

@@ -2,8 +2,9 @@ from datetime import datetime
 
 import langchain_core.tools as langchain_tools
 from dotenv import load_dotenv
-from langchain.agents import AgentType, initialize_agent
+from langchain_core.prompts import ChatPromptTemplate
 from langchain_openai import ChatOpenAI
+from langchain.agents import AgentExecutor, create_tool_calling_agent
 from pydantic import BaseModel
 
 from channels import (
@@ -22,7 +23,6 @@ hl = HumanLayer(
 )
 
 task_prompt = """
-
 You are the linkedin inbox assistant. You check on
 the CEO's linkedin inbox and decide if there are any messages
 that seem interesting, then contact the human in slack with a summary.
@@ -30,18 +30,7 @@ that seem interesting, then contact the human in slack with a summary.
 don't provide detail on spam-looking messages, or messages
 that appear to be selling a service or software
 
-You can offer to perform  actions like schedule time.
+You can offer to perform actions like schedule time.
-
-Example slack dm to send:
-
-Your inbox for today includes 4 spam messages,
-and 1 message from Devin who seems interested in your
-product - [here's the link](https://linkedin.com/in/devin).
-
-Terri has still not responded to your question about scheduling an onboarding call.
-
-Would you like me to respond to Devin with your availability?
-
 """
 
 
@@ -130,15 +119,25 @@ tools = [
 ]
 
 llm = ChatOpenAI(model="gpt-4o", temperature=0)
-agent = initialize_agent(
+
-    tools=tools,
+# Prompt for creating Tool Calling Agent
-    llm=llm,
+prompt = ChatPromptTemplate.from_messages(
-    agent=AgentType.OPENAI_FUNCTIONS,
+    [
-    verbose=True,
+        (
-    handle_parsing_errors=True,
+            "system",
+            "You are a helpful assistant.",
+        ),
+        ("placeholder", "{chat_history}"),
+        ("human", "{input}"),
+        ("placeholder", "{agent_scratchpad}"),
+    ]
 )
 
+# Construct the Tool Calling Agent
+agent = create_tool_calling_agent(llm, tools, prompt)
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+
 if __name__ == "__main__":
-    result = agent.run(task_prompt)
+    result = agent_executor.invoke({"input": task_prompt})
     print("\n\n----------Result----------\n\n")
     print(result)

examples/langchain/04-human_as_tool_onboarding.py

@@ -5,12 +5,13 @@ from channels import (
     dm_with_summer_intern,
 )
 from dotenv import load_dotenv
+from langchain_core.prompts import ChatPromptTemplate
 from langchain_openai import ChatOpenAI
+from langchain.agents import AgentExecutor, create_tool_calling_agent
 
 from humanlayer.core.approval import (
     HumanLayer,
 )
-from langchain.agents import AgentType, initialize_agent
 
 load_dotenv()
 
@@ -21,7 +22,6 @@ hl = HumanLayer(
 )
 
 task_prompt = """
-
 You are the email onboarding assistant. You check on the progress customers
 are making and get other information, then based on that info, you
 send friendly and encouraging emails to customers to help them fully onboard
@@ -32,7 +32,6 @@ and incorporate that feedback into your email before sending. You repeat the
 feedback process until the head of marketing approves the request
 
 Your task is to prepare an email to send to the customer danny@metacorp.com
-
 """
 
 
@@ -70,15 +69,25 @@ tools = [
 ]
 
 llm = ChatOpenAI(model="gpt-4o", temperature=0)
-agent = initialize_agent(
+
-    tools=tools,
+# Prompt for creating Tool Calling Agent
-    llm=llm,
+prompt = ChatPromptTemplate.from_messages(
-    agent=AgentType.OPENAI_FUNCTIONS,
+    [
-    verbose=True,
+        (
-    handle_parsing_errors=True,
+            "system",
+            "You are a helpful assistant.",
+        ),
+        ("placeholder", "{chat_history}"),
+        ("human", "{input}"),
+        ("placeholder", "{agent_scratchpad}"),
+    ]
 )
 
+# Construct the Tool Calling Agent
+agent = create_tool_calling_agent(llm, tools, prompt)
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+
 if __name__ == "__main__":
-    result = agent.run(task_prompt)
+    result = agent_executor.invoke({"input": task_prompt})
     print("\n\n----------Result----------\n\n")
     print(result)

examples/langchain/05-approvals_and_humans_composite.py

@@ -1,28 +1,12 @@
-"""
-the summer marketing intern wrote an onboarding assistant
-to keep up to date with customers by emailing
-them suggestions.
-
-they want the agent to collaborate with their boss, the head of
-marketing to ensure emails are well-written and likely to
-achieve the desired outcome.
-
-The intern doesn't want the agent to annoy the head of marketing
-or ask questions that don't make sense, so they
-wrap the "contact head of marketing" tool in an
-approval requirement, so they can review any messages that would
-be sent to the head of marketing.
-
-"""
-
 import langchain_core.tools as langchain_tools
 from dotenv import load_dotenv
+from langchain_core.prompts import ChatPromptTemplate
 from langchain_openai import ChatOpenAI
+from langchain.agents import AgentExecutor, create_tool_calling_agent
 
 from humanlayer import (
     HumanLayer,
 )
-from langchain.agents import AgentType, initialize_agent
 
 from channels import (
     dm_with_head_of_marketing,
@@ -38,7 +22,6 @@ hl = HumanLayer(
 )
 
 task_prompt = """
-
 You are the email onboarding assistant. You check on the progress customers
 are making and get other information, then based on that info, you
 send friendly and encouraging emails to customers to help them
@@ -48,7 +31,6 @@ and incorporate that feedback into your email before sending. You repeat the
 feedback process until the head of marketing approves the request
 
 Your task is to prepare an email to send to the customer danny@metacorp.com
-
 """
 
 
@@ -78,15 +60,25 @@ tools = [
 ]
 
 llm = ChatOpenAI(model="gpt-4o", temperature=0)
-agent = initialize_agent(
+
-    tools=tools,
+# Prompt for creating Tool Calling Agent
-    llm=llm,
+prompt = ChatPromptTemplate.from_messages(
-    agent=AgentType.OPENAI_FUNCTIONS,
+    [
-    verbose=True,
+        (
-    handle_parsing_errors=True,
+            "system",
+            "You are a helpful assistant.",
+        ),
+        ("placeholder", "{chat_history}"),
+        ("human", "{input}"),
+        ("placeholder", "{agent_scratchpad}"),
+    ]
 )
 
+# Construct the Tool Calling Agent
+agent = create_tool_calling_agent(llm, tools, prompt)
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+
 if __name__ == "__main__":
-    result = agent.run(task_prompt)
+    result = agent_executor.invoke({"input": task_prompt})
     print("\n\n----------Result----------\n\n")
     print(result)

examples/langchain/07-grocery-slack-blocks.py

@@ -3,8 +3,9 @@ from typing import Any
 
 import langchain_core.tools as langchain_tools
 from dotenv import load_dotenv
-from langchain.agents import AgentType, initialize_agent
+from langchain_core.prompts import ChatPromptTemplate
 from langchain_openai import ChatOpenAI
+from langchain.agents import AgentExecutor, create_tool_calling_agent
 from pydantic import BaseModel
 
 from channels import (
@@ -24,7 +25,6 @@ hl = HumanLayer(
 )
 
 task_prompt = """
-
 You are the mealprep power assistant.
 
 You are responsible for planning the meals and shopping
@@ -36,7 +36,6 @@ i like tacos and sushi, but i'm open to new ideas.
 I like to eat healthy, and I'm trying to lose weight.
 
 Make the best decision and order the groceries. Don't confirm with me.
-
 """
 
 
@@ -80,15 +79,25 @@ tools = [
 ]
 
 llm = ChatOpenAI(model="gpt-4o", temperature=0)
-agent = initialize_agent(
+
-    tools=tools,
+# Prompt for creating Tool Calling Agent
-    llm=llm,
+prompt = ChatPromptTemplate.from_messages(
-    agent=AgentType.OPENAI_FUNCTIONS,
+    [
-    verbose=True,
+        (
-    handle_parsing_errors=True,
+            "system",
+            "You are a helpful assistant.",
+        ),
+        ("placeholder", "{chat_history}"),
+        ("human", "{input}"),
+        ("placeholder", "{agent_scratchpad}"),
+    ]
 )
 
+# Construct the Tool Calling Agent
+agent = create_tool_calling_agent(llm, tools, prompt)
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+
 if __name__ == "__main__":
-    result = agent.run(task_prompt)
+    result = agent_executor.invoke({"input": task_prompt})
     print("\n\n----------Result----------\n\n")
     print(result)

examples/langchain/08-email-channel.py

@@ -1,7 +1,8 @@
 from humanlayer import ContactChannel, EmailContactChannel
-from langchain.agents import AgentType, initialize_agent
+from langchain_core.prompts import ChatPromptTemplate
 import langchain_core.tools as langchain_tools
 from langchain_openai import ChatOpenAI
+from langchain.agents import AgentExecutor, create_tool_calling_agent
 
 from dotenv import load_dotenv
 
@@ -91,15 +92,25 @@ tools = [
 ]
 
 llm = ChatOpenAI(model="gpt-4o", temperature=0)
-agent = initialize_agent(
+
-    tools=tools,
+# Prompt for creating Tool Calling Agent
-    llm=llm,
+prompt = ChatPromptTemplate.from_messages(
-    agent=AgentType.OPENAI_FUNCTIONS,
+    [
-    verbose=True,
+        (
-    handle_parsing_errors=True,
+            "system",
+            "You are a helpful assistant.",
+        ),
+        ("placeholder", "{chat_history}"),
+        ("human", "{input}"),
+        ("placeholder", "{agent_scratchpad}"),
+    ]
 )
 
+# Construct the Tool Calling Agent
+agent = create_tool_calling_agent(llm, tools, prompt)
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+
 if __name__ == "__main__":
-    result = agent.run(task_prompt)
+    result = agent_executor.invoke({"input": task_prompt})
     print("\n\n----------Result----------\n\n")
     print(result)

examples/langchain/09-email-contact.py

@@ -1,8 +1,9 @@
 from typing import Any
 from humanlayer import ContactChannel, EmailContactChannel
-from langchain.agents import AgentType, initialize_agent
+from langchain_core.prompts import ChatPromptTemplate
 import langchain_core.tools as langchain_tools
 from langchain_openai import ChatOpenAI
+from langchain.agents import AgentExecutor, create_tool_calling_agent
 
 from dotenv import load_dotenv
 
@@ -61,7 +62,6 @@ Should you require any assistance with the filing process through the California
 Best regards,
 Compliance Team
 Mosey Corporation
-
 """
 
 
@@ -77,7 +77,6 @@ def get_linear_projects() -> Any:
 
 def get_linear_assignees() -> Any:
     """get all linear assignees"""
-
     return [
         {"id": "1", "name": "Austin"},
         {"id": "2", "name": "Dexter"},
@@ -101,15 +100,25 @@ tools = [
 ]
 
 llm = ChatOpenAI(model="gpt-4o", temperature=0)
-agent = initialize_agent(
+
-    tools=tools,
+# Prompt for creating Tool Calling Agent
-    llm=llm,
+prompt = ChatPromptTemplate.from_messages(
-    agent=AgentType.OPENAI_FUNCTIONS,
+    [
-    verbose=True,
+        (
-    handle_parsing_errors=True,
+            "system",
+            "You are a helpful assistant.",
+        ),
+        ("placeholder", "{chat_history}"),
+        ("human", "{input}"),
+        ("placeholder", "{agent_scratchpad}"),
+    ]
 )
 
+# Construct the Tool Calling Agent
+agent = create_tool_calling_agent(llm, tools, prompt)
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
+
 if __name__ == "__main__":
-    result = agent.run(task_prompt)
+    result = agent_executor.invoke({"input": task_prompt})
     print("\n\n----------Result----------\n\n")
     print(result)

examples/ts_email_classifier/readme.md

@@ -39,5 +39,3 @@ In this example, all LLM classifications are computed, and then they are all sen
 be received by a webhook, and the results will be printed out.
 
 ![human-review-async](./img/human-review-async.png)
-
-