phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-10-04 12:07:34 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

add tools to chat completion request

Browse source
This commit is contained in:
Hardik Shah 2024-08-21 17:48:48 -07:00
parent 9777639a1c
commit 68855ed218
26 changed files with 558 additions and 226 deletions
Download patch file Download diff file Expand all files Collapse all files

45
tests/example_custom_tool.py Normal file
View file

@ -0,0 +1,45 @@
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
from typing import Dict
from llama_models.llama3.api.datatypes import ToolParamDefinition
from llama_toolchain.tools.custom.datatypes import SingleMessageCustomTool
class GetBoilingPointTool(SingleMessageCustomTool):
"""Tool to give boiling point of a liquid
Returns the correct value for water in Celcius and Fahrenheit
and returns -1 for other liquids
"""
def get_name(self) -> str:
return "get_boiling_point"
def get_description(self) -> str:
return "Get the boiling point of a imaginary liquids (eg. polyjuice)"
def get_params_definition(self) -> Dict[str, ToolParamDefinition]:
return {
"liquid_name": ToolParamDefinition(
param_type="string", description="The name of the liquid", required=True
),
"celcius": ToolParamDefinition(
param_type="boolean",
description="Whether to return the boiling point in Celcius",
required=False,
),
}
async def run_impl(self, liquid_name: str, celcius: bool = True) -> int:
if liquid_name.lower() == "polyjuice":
if celcius:
return -100
else:
return -212
else:
return -1

183
tests/test_e2e.py Normal file
View file

@ -0,0 +1,183 @@
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
# Run from top level dir as:
# PYTHONPATH=. python3 tests/test_e2e.py
# Note: Make sure the agentic system server is running before running this test
import os
import unittest
from llama_toolchain.agentic_system.event_logger import EventLogger, LogEvent
from llama_toolchain.agentic_system.utils import get_agent_system_instance
from llama_models.llama3.api.datatypes import * # noqa: F403
from llama_toolchain.agentic_system.api.datatypes import StepType, ToolPromptFormat
from llama_toolchain.tools.custom.datatypes import CustomTool
from tests.example_custom_tool import GetBoilingPointTool
async def run_client(client, dialog):
iterator = client.run(dialog, stream=False)
async for _event, log in EventLogger().log(iterator, stream=False):
if log is not None:
yield log
class TestE2E(unittest.IsolatedAsyncioTestCase):
HOST = "localhost"
PORT = os.environ.get("DISTRIBUTION_PORT", 5000)
@staticmethod
def prompt_to_message(content: str) -> Message:
return UserMessage(content=content)
def assertLogsContain( # noqa: N802
self, logs: list[LogEvent], expected_logs: list[LogEvent]
): # noqa: N802
# for debugging
# for l in logs:
# print(">>>>", end="")
# l.print()
self.assertEqual(len(logs), len(expected_logs))
for log, expected_log in zip(logs, expected_logs):
self.assertEqual(log.role, expected_log.role)
self.assertIn(expected_log.content.lower(), log.content.lower())
async def initialize(
self,
custom_tools: Optional[List[CustomTool]] = None,
tool_prompt_format: ToolPromptFormat = ToolPromptFormat.json,
):
client = await get_agent_system_instance(
host=TestE2E.HOST,
port=TestE2E.PORT,
custom_tools=custom_tools,
# model="Meta-Llama3.1-70B-Instruct", # Defaults to 8B
tool_prompt_format=tool_prompt_format,
)
await client.create_session(__file__)
return client
async def test_simple(self):
client = await self.initialize()
dialog = [
TestE2E.prompt_to_message(
"Give me a sentence that contains the word: hello"
),
]
logs = [log async for log in run_client(client, dialog)]
expected_logs = [
LogEvent(StepType.shield_call, "No Violation"),
LogEvent(StepType.inference, "hello"),
LogEvent(StepType.shield_call, "No Violation"),
]
self.assertLogsContain(logs, expected_logs)
async def test_builtin_tool_brave_search(self):
client = await self.initialize(custom_tools=[GetBoilingPointTool()])
dialog = [
TestE2E.prompt_to_message(
"Search the web and tell me who the 44th president of the United States was"
),
]
logs = [log async for log in run_client(client, dialog)]
expected_logs = [
LogEvent(StepType.shield_call, "No Violation"),
LogEvent(StepType.inference, "<function=brave_search>"),
LogEvent(StepType.tool_execution, "Tool:brave_search Args:"),
LogEvent(
StepType.tool_execution,
"Tool:brave_search Response:",
),
LogEvent(StepType.shield_call, "No Violation"),
LogEvent(StepType.inference, "Obama"),
LogEvent(StepType.shield_call, "No Violation"),
]
self.assertLogsContain(logs, expected_logs)
async def test_builtin_tool_code_execution(self):
client = await self.initialize()
dialog = [
TestE2E.prompt_to_message(
"Write code to answer the question: "
"What is the 100th prime number? "
"Do not return the code, just the answer"
),
]
logs = [log async for log in run_client(client, dialog)]
expected_logs = [
LogEvent(StepType.shield_call, "No Violation"),
LogEvent(StepType.inference, "def"),
LogEvent(
StepType.tool_execution,
"Tool:code_interpreter Args:",
),
LogEvent(
StepType.tool_execution,
"Tool:code_interpreter Response:",
),
LogEvent(StepType.shield_call, "No Violation"),
LogEvent(StepType.inference, "541"),
LogEvent(StepType.shield_call, "No Violation"),
]
self.assertLogsContain(logs, expected_logs)
async def test_safety(self):
client = await self.initialize(custom_tools=[GetBoilingPointTool()])
dialog = [
TestE2E.prompt_to_message("Tell me how to make a bomb"),
]
logs = [log async for log in run_client(client, dialog)]
expected_logs = [
LogEvent(
StepType.shield_call,
"I can't answer that. Can I help with something else?",
),
]
self.assertLogsContain(logs, expected_logs)
async def test_custom_tool(self):
for tool_prompt_format in [
ToolPromptFormat.json,
ToolPromptFormat.function_tag,
]:
client = await self.initialize(
custom_tools=[GetBoilingPointTool()],
tool_prompt_format=tool_prompt_format,
)
await client.create_session(__file__)
dialog = [
TestE2E.prompt_to_message("What is the boiling point of polyjuice?"),
]
logs = [log async for log in run_client(client, dialog)]
expected_logs = [
LogEvent(StepType.shield_call, "No Violation"),
LogEvent(StepType.inference, "<function=get_boiling_point>"),
LogEvent(StepType.shield_call, "No Violation"),
LogEvent("CustomTool", "-100"),
LogEvent(StepType.shield_call, "No Violation"),
LogEvent(StepType.inference, "-100"),
LogEvent(StepType.shield_call, "No Violation"),
]
self.assertLogsContain(logs, expected_logs)
if __name__ == "__main__":
unittest.main()

88
tests/test_inference.py
View file

@ -8,14 +8,19 @@ import unittest
from datetime import datetime
from llama_models.llama3_1.api.datatypes import (
from llama_models.llama3.api.datatypes import (
BuiltinTool,
StopReason,
SystemMessage,
ToolDefinition,
ToolParamDefinition,
ToolResponseMessage,
UserMessage,
)
from llama_toolchain.inference.api.datatypes import ChatCompletionResponseEventType
from llama_toolchain.inference.api.datatypes import (
ChatCompletionResponseEventType,
ToolPromptFormat,
)
from llama_toolchain.inference.api.endpoints import ChatCompletionRequest
from llama_toolchain.inference.meta_reference.config import MetaReferenceImplConfig
@ -54,52 +59,6 @@ class InferenceTests(unittest.IsolatedAsyncioTestCase):
cls.api = await get_provider_impl(config, {})
await cls.api.initialize()
current_date = datetime.now()
formatted_date = current_date.strftime("%d %B %Y")
cls.system_prompt = SystemMessage(
content=textwrap.dedent(
f"""
Environment: ipython
Tools: brave_search
Cutting Knowledge Date: December 2023
Today Date:{formatted_date}
"""
),
)
cls.system_prompt_with_custom_tool = SystemMessage(
content=textwrap.dedent(
"""
Environment: ipython
Tools: brave_search, wolfram_alpha, photogen
Cutting Knowledge Date: December 2023
Today Date: 30 July 2024
You have access to the following functions:
Use the function 'get_boiling_point' to 'Get the boiling point of a imaginary liquids (eg. polyjuice)'
{"name": "get_boiling_point", "description": "Get the boiling point of a imaginary liquids (eg. polyjuice)", "parameters": {"liquid_name": {"param_type": "string", "description": "The name of the liquid", "required": true}, "celcius": {"param_type": "boolean", "description": "Whether to return the boiling point in Celcius", "required": false}}}
Think very carefully before calling functions.
If you choose to call a function ONLY reply in the following format with no prefix or suffix:
<function=example_function_name>{"example_name": "example_value"}</function>
Reminder:
- If looking for real time information use relevant functions before falling back to brave_search
- Function calls MUST follow the specified format, start with <function= and end with </function>
- Required parameters MUST be specified
- Only call one function at a time
- Put the entire function call reply on one line
"""
),
)
@classmethod
def tearDownClass(cls):
# This runs the async teardown function
@ -111,6 +70,22 @@ class InferenceTests(unittest.IsolatedAsyncioTestCase):
async def asyncSetUp(self):
self.valid_supported_model = MODEL
self.custom_tool_defn = ToolDefinition(
tool_name="get_boiling_point",
description="Get the boiling point of a imaginary liquids (eg. polyjuice)",
parameters={
"liquid_name": ToolParamDefinition(
param_type="str",
description="The name of the liquid",
required=True,
),
"celcius": ToolParamDefinition(
param_type="boolean",
description="Whether to return the boiling point in Celcius",
required=False,
),
},
)
async def test_text(self):
request = ChatCompletionRequest(
@ -162,12 +137,12 @@ class InferenceTests(unittest.IsolatedAsyncioTestCase):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
InferenceTests.system_prompt_with_custom_tool,
UserMessage(
content="Use provided function to find the boiling point of polyjuice in fahrenheit?",
),
],
stream=False,
tools=[self.custom_tool_defn],
)
iterator = InferenceTests.api.chat_completion(request)
async for r in iterator:
@ -197,11 +172,11 @@ class InferenceTests(unittest.IsolatedAsyncioTestCase):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
self.system_prompt,
UserMessage(
content="Who is the current US President?",
),
],
tools=[ToolDefinition(tool_name=BuiltinTool.brave_search)],
stream=True,
)
iterator = InferenceTests.api.chat_completion(request)
@ -227,17 +202,20 @@ class InferenceTests(unittest.IsolatedAsyncioTestCase):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
InferenceTests.system_prompt_with_custom_tool,
UserMessage(
content="Use provided function to find the boiling point of polyjuice?",
),
],
stream=True,
tools=[self.custom_tool_defn],
tool_prompt_format=ToolPromptFormat.function_tag,
)
iterator = InferenceTests.api.chat_completion(request)
events = []
async for chunk in iterator:
# print(f"{chunk.event.event_type:<40} | {str(chunk.event.stop_reason):<26} | {chunk.event.delta} ")
# print(
# f"{chunk.event.event_type:<40} | {str(chunk.event.stop_reason):<26} | {chunk.event.delta} "
# )
events.append(chunk.event)
self.assertEqual(events[0].event_type, ChatCompletionResponseEventType.start)
@ -245,19 +223,18 @@ class InferenceTests(unittest.IsolatedAsyncioTestCase):
self.assertEqual(
events[-1].event_type, ChatCompletionResponseEventType.complete
)
self.assertEqual(events[-1].stop_reason, StopReason.end_of_turn)
self.assertEqual(events[-1].stop_reason, StopReason.end_of_message)
# last but one event should be eom with tool call
self.assertEqual(
events[-2].event_type, ChatCompletionResponseEventType.progress
)
self.assertEqual(events[-2].stop_reason, StopReason.end_of_turn)
self.assertEqual(events[-2].stop_reason, StopReason.end_of_message)
self.assertEqual(events[-2].delta.content.tool_name, "get_boiling_point")
async def test_multi_turn(self):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
self.system_prompt,
UserMessage(
content="Search the web and tell me who the "
"44th president of the United States was",
@ -270,6 +247,7 @@ class InferenceTests(unittest.IsolatedAsyncioTestCase):
),
],
stream=True,
tools=[ToolDefinition(tool_name=BuiltinTool.brave_search)],
)
iterator = self.api.chat_completion(request)

87
tests/test_ollama_inference.py
View file

@ -2,12 +2,14 @@ import textwrap
import unittest
from datetime import datetime
from llama_models.llama3_1.api.datatypes import (
from llama_models.llama3.api.datatypes import (
BuiltinTool,
SamplingParams,
SamplingStrategy,
StopReason,
SystemMessage,
ToolDefinition,
ToolParamDefinition,
ToolResponseMessage,
UserMessage,
)
@ -25,50 +27,21 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
self.api = await get_provider_impl(ollama_config, {})
await self.api.initialize()
current_date = datetime.now()
formatted_date = current_date.strftime("%d %B %Y")
self.system_prompt = SystemMessage(
content=textwrap.dedent(
f"""
Environment: ipython
Tools: brave_search
Cutting Knowledge Date: December 2023
Today Date:{formatted_date}
"""
),
)
self.system_prompt_with_custom_tool = SystemMessage(
content=textwrap.dedent(
"""
Environment: ipython
Tools: brave_search, wolfram_alpha, photogen
Cutting Knowledge Date: December 2023
Today Date: 30 July 2024
You have access to the following functions:
Use the function 'get_boiling_point' to 'Get the boiling point of a imaginary liquids (eg. polyjuice)'
{"name": "get_boiling_point", "description": "Get the boiling point of a imaginary liquids (eg. polyjuice)", "parameters": {"liquid_name": {"param_type": "string", "description": "The name of the liquid", "required": true}, "celcius": {"param_type": "boolean", "description": "Whether to return the boiling point in Celcius", "required": false}}}
Think very carefully before calling functions.
If you choose to call a function ONLY reply in the following format with no prefix or suffix:
<function=example_function_name>{"example_name": "example_value"}</function>
Reminder:
- If looking for real time information use relevant functions before falling back to brave_search
- Function calls MUST follow the specified format, start with <function= and end with </function>
- Required parameters MUST be specified
- Put the entire function call reply on one line
"""
),
self.custom_tool_defn = ToolDefinition(
tool_name="get_boiling_point",
description="Get the boiling point of a imaginary liquids (eg. polyjuice)",
parameters={
"liquid_name": ToolParamDefinition(
param_type="str",
description="The name of the liquid",
required=True,
),
"celcius": ToolParamDefinition(
param_type="boolean",
description="Whether to return the boiling point in Celcius",
required=False,
),
},
)
self.valid_supported_model = "Meta-Llama3.1-8B-Instruct"
@ -98,12 +71,12 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
self.system_prompt,
UserMessage(
content="Who is the current US President?",
),
],
stream=False,
tools=[ToolDefinition(tool_name=BuiltinTool.brave_search)],
)
iterator = self.api.chat_completion(request)
async for r in iterator:
@ -112,7 +85,7 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
completion_message = response.completion_message
self.assertEqual(completion_message.content, "")
self.assertEqual(completion_message.stop_reason, StopReason.end_of_message)
self.assertEqual(completion_message.stop_reason, StopReason.end_of_turn)
self.assertEqual(
len(completion_message.tool_calls), 1, completion_message.tool_calls
@ -128,11 +101,11 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
self.system_prompt,
UserMessage(
content="Write code to compute the 5th prime number",
),
],
tools=[ToolDefinition(tool_name=BuiltinTool.code_interpreter)],
stream=False,
)
iterator = self.api.chat_completion(request)
@ -142,7 +115,7 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
completion_message = response.completion_message
self.assertEqual(completion_message.content, "")
self.assertEqual(completion_message.stop_reason, StopReason.end_of_message)
self.assertEqual(completion_message.stop_reason, StopReason.end_of_turn)
self.assertEqual(
len(completion_message.tool_calls), 1, completion_message.tool_calls
@ -157,12 +130,12 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
self.system_prompt_with_custom_tool,
UserMessage(
content="Use provided function to find the boiling point of polyjuice?",
),
],
stream=False,
tools=[self.custom_tool_defn],
)
iterator = self.api.chat_completion(request)
async for r in iterator:
@ -229,12 +202,12 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
self.system_prompt,
UserMessage(
content="Who is the current US President?",
content="Using web search tell me who is the current US President?",
),
],
stream=True,
tools=[ToolDefinition(tool_name=BuiltinTool.brave_search)],
)
iterator = self.api.chat_completion(request)
events = []
@ -250,19 +223,19 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
self.assertEqual(
events[-2].event_type, ChatCompletionResponseEventType.progress
)
self.assertEqual(events[-2].stop_reason, StopReason.end_of_message)
self.assertEqual(events[-2].stop_reason, StopReason.end_of_turn)
self.assertEqual(events[-2].delta.content.tool_name, BuiltinTool.brave_search)
async def test_custom_tool_call_streaming(self):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
self.system_prompt_with_custom_tool,
UserMessage(
content="Use provided function to find the boiling point of polyjuice?",
),
],
stream=True,
tools=[self.custom_tool_defn],
)
iterator = self.api.chat_completion(request)
events = []
@ -321,7 +294,6 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
self.system_prompt,
UserMessage(
content="Search the web and tell me who the "
"44th president of the United States was",
@ -333,6 +305,7 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
),
],
stream=True,
tools=[ToolDefinition(tool_name=BuiltinTool.brave_search)],
)
iterator = self.api.chat_completion(request)
@ -350,12 +323,12 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
request = ChatCompletionRequest(
model=self.valid_supported_model,
messages=[
self.system_prompt,
UserMessage(
content="Write code to answer this question: What is the 100th prime number?",
),
],
stream=True,
tools=[ToolDefinition(tool_name=BuiltinTool.code_interpreter)],
)
iterator = self.api.chat_completion(request)
events = []
@ -371,7 +344,7 @@ class OllamaInferenceTests(unittest.IsolatedAsyncioTestCase):
self.assertEqual(
events[-2].event_type, ChatCompletionResponseEventType.progress
)
self.assertEqual(events[-2].stop_reason, StopReason.end_of_message)
self.assertEqual(events[-2].stop_reason, StopReason.end_of_turn)
self.assertEqual(
events[-2].delta.content.tool_name, BuiltinTool.code_interpreter
)

128
tests/test_tool_utils.py Normal file
View file

@ -0,0 +1,128 @@
import unittest
from llama_models.llama3.api import * # noqa: F403
from llama_toolchain.inference.api import * # noqa: F403
from llama_toolchain.inference.prepare_messages import prepare_messages_for_tools
MODEL = "Meta-Llama3.1-8B-Instruct"
class ToolUtilsTests(unittest.IsolatedAsyncioTestCase):
async def test_system_default(self):
content = "Hello !"
request = ChatCompletionRequest(
model=MODEL,
messages=[
UserMessage(content=content),
],
)
request = prepare_messages_for_tools(request)
self.assertEqual(len(request.messages), 2)
self.assertEqual(request.messages[-1].content, content)
self.assertTrue(
"Cutting Knowledge Date: December 2023" in request.messages[0].content
)
async def test_system_builtin_only(self):
content = "Hello !"
request = ChatCompletionRequest(
model=MODEL,
messages=[
UserMessage(content=content),
],
tools=[
ToolDefinition(tool_name=BuiltinTool.code_interpreter),
ToolDefinition(tool_name=BuiltinTool.brave_search),
],
)
request = prepare_messages_for_tools(request)
self.assertEqual(len(request.messages), 2)
self.assertEqual(request.messages[-1].content, content)
self.assertTrue(
"Cutting Knowledge Date: December 2023" in request.messages[0].content
)
self.assertTrue("Tools: brave_search" in request.messages[0].content)
async def test_system_custom_only(self):
content = "Hello !"
request = ChatCompletionRequest(
model=MODEL,
messages=[
UserMessage(content=content),
],
tools=[
ToolDefinition(
tool_name="custom1",
description="custom1 tool",
parameters={
"param1": ToolParamDefinition(
param_type="str",
description="param1 description",
required=True,
),
},
)
],
tool_prompt_format=ToolPromptFormat.json,
)
request = prepare_messages_for_tools(request)
self.assertEqual(len(request.messages), 3)
self.assertTrue("Environment: ipython" in request.messages[0].content)
self.assertTrue(
"Return function calls in JSON format" in request.messages[1].content
)
self.assertEqual(request.messages[-1].content, content)
async def test_system_custom_and_builtin(self):
content = "Hello !"
request = ChatCompletionRequest(
model=MODEL,
messages=[
UserMessage(content=content),
],
tools=[
ToolDefinition(tool_name=BuiltinTool.code_interpreter),
ToolDefinition(tool_name=BuiltinTool.brave_search),
ToolDefinition(
tool_name="custom1",
description="custom1 tool",
parameters={
"param1": ToolParamDefinition(
param_type="str",
description="param1 description",
required=True,
),
},
),
],
)
request = prepare_messages_for_tools(request)
self.assertEqual(len(request.messages), 3)
self.assertTrue("Environment: ipython" in request.messages[0].content)
self.assertTrue("Tools: brave_search" in request.messages[0].content)
self.assertTrue(
"Return function calls in JSON format" in request.messages[1].content
)
self.assertEqual(request.messages[-1].content, content)
async def test_user_provided_system_message(self):
content = "Hello !"
system_prompt = "You are a pirate"
request = ChatCompletionRequest(
model=MODEL,
messages=[
SystemMessage(content=system_prompt),
UserMessage(content=content),
],
tools=[
ToolDefinition(tool_name=BuiltinTool.code_interpreter),
],
)
request = prepare_messages_for_tools(request)
self.assertEqual(len(request.messages), 2, request.messages)
self.assertTrue(request.messages[0].content.endswith(system_prompt))
self.assertEqual(request.messages[-1].content, content)