Mark Sze ed47ec9d4b
Transform to add an agent's name into the message content (#3334)
* Initial commit with ability to add name into content with a transform

* Transforms documentation

* Fix transform links in documentation

---------

Co-authored-by: Li Jiang <bnujli@gmail.com>
2024-08-26 07:35:27 +00:00

495 lines
20 KiB
Python

import copy
from typing import Any, Dict, List
from unittest.mock import MagicMock, patch
import pytest
from autogen.agentchat.contrib.capabilities.text_compressors import TextCompressor
from autogen.agentchat.contrib.capabilities.transforms import (
MessageHistoryLimiter,
MessageTokenLimiter,
TextMessageCompressor,
TextMessageContentName,
)
from autogen.agentchat.contrib.capabilities.transforms_util import count_text_tokens
class _MockTextCompressor:
def compress_text(self, text: str, **compression_params) -> Dict[str, Any]:
return {"compressed_prompt": ""}
def get_long_messages() -> List[Dict]:
return [
{"role": "assistant", "content": [{"type": "text", "text": "are you doing?"}]},
{"role": "user", "content": "very very very very very very long string"},
{"role": "user", "content": "hello"},
{"role": "assistant", "content": [{"type": "text", "text": "there"}]},
{"role": "user", "content": "how"},
]
def get_short_messages() -> List[Dict]:
return [
{"role": "user", "content": "hello"},
{"role": "assistant", "content": [{"type": "text", "text": "there"}]},
{"role": "user", "content": "how are you"},
]
def get_no_content_messages() -> List[Dict]:
return [{"role": "user", "function_call": "example"}, {"role": "assistant", "content": None}]
def get_tool_messages() -> List[Dict]:
return [
{"role": "user", "content": "hello"},
{"role": "tool_calls", "content": "calling_tool"},
{"role": "tool", "content": "tool_response"},
{"role": "user", "content": "how are you"},
{"role": "assistant", "content": [{"type": "text", "text": "are you doing?"}]},
]
def get_tool_messages_kept() -> List[Dict]:
return [
{"role": "user", "content": "hello"},
{"role": "tool_calls", "content": "calling_tool"},
{"role": "tool", "content": "tool_response"},
{"role": "tool_calls", "content": "calling_tool"},
{"role": "tool", "content": "tool_response"},
]
def get_messages_with_names() -> List[Dict]:
return [
{"role": "system", "content": "I am the system."},
{"role": "user", "name": "charlie", "content": "I think the sky is blue."},
{"role": "user", "name": "mary", "content": "The sky is red."},
{"role": "user", "name": "bob", "content": "The sky is crimson."},
]
def get_messages_with_names_post_start() -> List[Dict]:
return [
{"role": "system", "content": "I am the system."},
{"role": "user", "name": "charlie", "content": "'charlie' said:\nI think the sky is blue."},
{"role": "user", "name": "mary", "content": "'mary' said:\nThe sky is red."},
{"role": "user", "name": "bob", "content": "'bob' said:\nThe sky is crimson."},
]
def get_messages_with_names_post_end() -> List[Dict]:
return [
{"role": "system", "content": "I am the system."},
{"role": "user", "name": "charlie", "content": "I think the sky is blue.\n(said 'charlie')"},
{"role": "user", "name": "mary", "content": "The sky is red.\n(said 'mary')"},
{"role": "user", "name": "bob", "content": "The sky is crimson.\n(said 'bob')"},
]
def get_messages_with_names_post_filtered() -> List[Dict]:
return [
{"role": "system", "content": "I am the system."},
{"role": "user", "name": "charlie", "content": "I think the sky is blue."},
{"role": "user", "name": "mary", "content": "'mary' said:\nThe sky is red."},
{"role": "user", "name": "bob", "content": "'bob' said:\nThe sky is crimson."},
]
def get_text_compressors() -> List[TextCompressor]:
compressors: List[TextCompressor] = [_MockTextCompressor()]
try:
from autogen.agentchat.contrib.capabilities.text_compressors import LLMLingua
compressors.append(LLMLingua())
except ImportError:
pass
return compressors
@pytest.fixture
def message_history_limiter() -> MessageHistoryLimiter:
return MessageHistoryLimiter(max_messages=3)
@pytest.fixture
def message_history_limiter_keep_first() -> MessageHistoryLimiter:
return MessageHistoryLimiter(max_messages=3, keep_first_message=True)
@pytest.fixture
def message_token_limiter() -> MessageTokenLimiter:
return MessageTokenLimiter(max_tokens_per_message=3)
@pytest.fixture
def message_token_limiter_with_threshold() -> MessageTokenLimiter:
return MessageTokenLimiter(max_tokens_per_message=1, min_tokens=10)
def _filter_dict_test(
post_transformed_message: Dict, pre_transformed_messages: Dict, roles: List[str], exclude_filter: bool
) -> bool:
is_role = post_transformed_message["role"] in roles
if exclude_filter:
is_role = not is_role
if isinstance(post_transformed_message["content"], list):
condition = (
len(post_transformed_message["content"][0]["text"]) < len(pre_transformed_messages["content"][0]["text"])
if is_role
else len(post_transformed_message["content"][0]["text"])
== len(pre_transformed_messages["content"][0]["text"])
)
else:
condition = (
len(post_transformed_message["content"]) < len(pre_transformed_messages["content"])
if is_role
else len(post_transformed_message["content"]) == len(pre_transformed_messages["content"])
)
return condition
# MessageHistoryLimiter
@pytest.mark.parametrize(
"messages, expected_messages_len",
[
(get_long_messages(), 3),
(get_short_messages(), 3),
(get_no_content_messages(), 2),
(get_tool_messages(), 2),
(get_tool_messages_kept(), 2),
],
)
def test_message_history_limiter_apply_transform(message_history_limiter, messages, expected_messages_len):
transformed_messages = message_history_limiter.apply_transform(messages)
assert len(transformed_messages) == expected_messages_len
if messages == get_tool_messages_kept():
assert transformed_messages[0]["role"] == "tool_calls"
assert transformed_messages[1]["role"] == "tool"
@pytest.mark.parametrize(
"messages, expected_messages_len",
[
(get_long_messages(), 3),
(get_short_messages(), 3),
(get_no_content_messages(), 2),
(get_tool_messages(), 3),
(get_tool_messages_kept(), 3),
],
)
def test_message_history_limiter_apply_transform_keep_first(
message_history_limiter_keep_first, messages, expected_messages_len
):
transformed_messages = message_history_limiter_keep_first.apply_transform(messages)
assert len(transformed_messages) == expected_messages_len
if messages == get_tool_messages_kept():
assert transformed_messages[1]["role"] == "tool_calls"
assert transformed_messages[2]["role"] == "tool"
@pytest.mark.parametrize(
"messages, expected_logs, expected_effect",
[
(get_long_messages(), "Removed 2 messages. Number of messages reduced from 5 to 3.", True),
(get_short_messages(), "No messages were removed.", False),
(get_no_content_messages(), "No messages were removed.", False),
(get_tool_messages(), "Removed 3 messages. Number of messages reduced from 5 to 2.", True),
(get_tool_messages_kept(), "Removed 3 messages. Number of messages reduced from 5 to 2.", True),
],
)
def test_message_history_limiter_get_logs(message_history_limiter, messages, expected_logs, expected_effect):
pre_transform_messages = copy.deepcopy(messages)
transformed_messages = message_history_limiter.apply_transform(messages)
logs_str, had_effect = message_history_limiter.get_logs(pre_transform_messages, transformed_messages)
assert had_effect == expected_effect
assert logs_str == expected_logs
# MessageTokenLimiter tests
@pytest.mark.parametrize(
"messages, expected_token_count, expected_messages_len",
[(get_long_messages(), 9, 5), (get_short_messages(), 5, 3), (get_no_content_messages(), 0, 2)],
)
def test_message_token_limiter_apply_transform(
message_token_limiter, messages, expected_token_count, expected_messages_len
):
transformed_messages = message_token_limiter.apply_transform(copy.deepcopy(messages))
assert (
sum(count_text_tokens(msg["content"]) for msg in transformed_messages if "content" in msg)
== expected_token_count
)
assert len(transformed_messages) == expected_messages_len
@pytest.mark.parametrize("messages", [get_long_messages(), get_short_messages()])
def test_message_token_limiter_with_filter(messages):
# Test truncating all messages except for user
message_token_limiter = MessageTokenLimiter(max_tokens_per_message=0, filter_dict={"role": "user"})
transformed_messages = message_token_limiter.apply_transform(copy.deepcopy(messages))
pre_post_messages = zip(messages, transformed_messages)
for pre_transform, post_transform in pre_post_messages:
assert _filter_dict_test(post_transform, pre_transform, ["user"], exclude_filter=True)
# Test truncating all user messages only
message_token_limiter = MessageTokenLimiter(
max_tokens_per_message=0, filter_dict={"role": "user"}, exclude_filter=False
)
transformed_messages = message_token_limiter.apply_transform(copy.deepcopy(messages))
pre_post_messages = zip(messages, transformed_messages)
for pre_transform, post_transform in pre_post_messages:
assert _filter_dict_test(post_transform, pre_transform, ["user"], exclude_filter=False)
@pytest.mark.parametrize(
"messages, expected_token_count, expected_messages_len",
[(get_long_messages(), 5, 5), (get_short_messages(), 5, 3), (get_no_content_messages(), 0, 2)],
)
def test_message_token_limiter_with_threshold_apply_transform(
message_token_limiter_with_threshold, messages, expected_token_count, expected_messages_len
):
transformed_messages = message_token_limiter_with_threshold.apply_transform(messages)
assert (
sum(count_text_tokens(msg["content"]) for msg in transformed_messages if "content" in msg)
== expected_token_count
)
assert len(transformed_messages) == expected_messages_len
@pytest.mark.parametrize(
"messages, expected_logs, expected_effect",
[
(get_long_messages(), "Truncated 6 tokens. Number of tokens reduced from 15 to 9", True),
(get_short_messages(), "No tokens were truncated.", False),
(get_no_content_messages(), "No tokens were truncated.", False),
],
)
def test_message_token_limiter_get_logs(message_token_limiter, messages, expected_logs, expected_effect):
pre_transform_messages = copy.deepcopy(messages)
transformed_messages = message_token_limiter.apply_transform(messages)
logs_str, had_effect = message_token_limiter.get_logs(pre_transform_messages, transformed_messages)
assert had_effect == expected_effect
assert logs_str == expected_logs
# TextMessageCompressor tests
@pytest.mark.parametrize("text_compressor", get_text_compressors())
def test_text_compression(text_compressor):
"""Test the TextMessageCompressor transform."""
compressor = TextMessageCompressor(text_compressor=text_compressor)
text = "Run this test with a long string. "
messages = [
{"role": "assistant", "content": [{"type": "text", "text": "".join([text] * 3)}]},
{"role": "role", "content": [{"type": "text", "text": "".join([text] * 3)}]},
{"role": "assistant", "content": [{"type": "text", "text": "".join([text] * 3)}]},
{"role": "assistant", "content": [{"type": "text", "text": "".join([text] * 3)}]},
]
transformed_messages = compressor.apply_transform([{"content": text}])
assert len(transformed_messages[0]["content"]) < len(text)
# Test compressing all messages
compressor = TextMessageCompressor(text_compressor=text_compressor)
transformed_messages = compressor.apply_transform(copy.deepcopy(messages))
pre_post_messages = zip(messages, transformed_messages)
for pre_transform, post_transform in pre_post_messages:
assert len(post_transform["content"][0]["text"]) < len(pre_transform["content"][0]["text"])
@pytest.mark.parametrize("messages", [get_long_messages(), get_short_messages()])
@pytest.mark.parametrize("text_compressor", get_text_compressors())
def test_text_compression_with_filter(messages, text_compressor):
# Test truncating all messages except for user
compressor = TextMessageCompressor(text_compressor=text_compressor, filter_dict={"role": "user"})
transformed_messages = compressor.apply_transform(copy.deepcopy(messages))
pre_post_messages = zip(messages, transformed_messages)
for pre_transform, post_transform in pre_post_messages:
assert _filter_dict_test(post_transform, pre_transform, ["user"], exclude_filter=True)
# Test truncating all user messages only
compressor = TextMessageCompressor(
text_compressor=text_compressor, filter_dict={"role": "user"}, exclude_filter=False
)
transformed_messages = compressor.apply_transform(copy.deepcopy(messages))
pre_post_messages = zip(messages, transformed_messages)
for pre_transform, post_transform in pre_post_messages:
assert _filter_dict_test(post_transform, pre_transform, ["user"], exclude_filter=False)
@pytest.mark.parametrize("messages", [get_messages_with_names()])
def test_message_content_name(messages):
# Test including content name in messages
# Add name at the start with format: "'{name}' said:\n"
content_transform = TextMessageContentName(position="start", format_string="'{name}' said:\n")
transformed_messages = content_transform.apply_transform(messages=messages)
assert transformed_messages == get_messages_with_names_post_start()
# Add name at the end with format: "\n(said '{name}')"
content_transform = TextMessageContentName(position="end", format_string="\n(said '{name}')")
transformed_messages_end = content_transform.apply_transform(messages=messages)
assert transformed_messages_end == get_messages_with_names_post_end()
# Test filtering out exclusion
content_transform = TextMessageContentName(
position="start",
format_string="'{name}' said:\n",
filter_dict={"name": ["charlie"]},
exclude_filter=True, # Exclude
)
transformed_messages_end = content_transform.apply_transform(messages=messages)
assert transformed_messages_end == get_messages_with_names_post_filtered()
# Test filtering (inclusion)
content_transform = TextMessageContentName(
position="start",
format_string="'{name}' said:\n",
filter_dict={"name": ["mary", "bob"]},
exclude_filter=False, # Include
)
transformed_messages_end = content_transform.apply_transform(messages=messages)
assert transformed_messages_end == get_messages_with_names_post_filtered()
# Test instantiation
with pytest.raises(AssertionError):
TextMessageContentName(position=123) # Invalid type for position
with pytest.raises(AssertionError):
TextMessageContentName(position="middle") # Invalid value for position
with pytest.raises(AssertionError):
TextMessageContentName(format_string=123) # Invalid type for format_string
with pytest.raises(AssertionError):
TextMessageContentName(format_string="Agent:\n") # Missing '{name}' in format_string
with pytest.raises(AssertionError):
TextMessageContentName(deduplicate="yes") # Invalid type for deduplicate
if __name__ == "__main__":
long_messages = get_long_messages()
short_messages = get_short_messages()
no_content_messages = get_no_content_messages()
tool_messages = get_tool_messages()
msg_history_limiter = MessageHistoryLimiter(max_messages=3)
msg_history_limiter_keep_first = MessageHistoryLimiter(max_messages=3, keep_first=True)
msg_token_limiter = MessageTokenLimiter(max_tokens_per_message=3)
msg_token_limiter_with_threshold = MessageTokenLimiter(max_tokens_per_message=1, min_tokens=10)
# Test Parameters
message_history_limiter_apply_transform_parameters = {
"messages": [long_messages, short_messages, no_content_messages, tool_messages],
"expected_messages_len": [3, 3, 2, 4],
}
message_history_limiter_get_logs_parameters = {
"messages": [long_messages, short_messages, no_content_messages, tool_messages],
"expected_logs": [
"Removed 2 messages. Number of messages reduced from 5 to 3.",
"No messages were removed.",
"No messages were removed.",
"Removed 1 messages. Number of messages reduced from 5 to 4.",
],
"expected_effect": [True, False, False, True],
}
message_token_limiter_apply_transform_parameters = {
"messages": [long_messages, short_messages, no_content_messages],
"expected_token_count": [9, 5, 0],
"expected_messages_len": [5, 3, 2],
}
message_token_limiter_with_threshold_apply_transform_parameters = {
"messages": [long_messages, short_messages, no_content_messages],
"expected_token_count": [5, 5, 0],
"expected_messages_len": [5, 3, 2],
}
message_token_limiter_get_logs_parameters = {
"messages": [long_messages, short_messages, no_content_messages],
"expected_logs": [
"Truncated 6 tokens. Number of tokens reduced from 15 to 9",
"No tokens were truncated.",
"No tokens were truncated.",
],
"expected_effect": [True, False, False],
}
# Call the MessageHistoryLimiter tests
for messages, expected_messages_len in zip(
message_history_limiter_apply_transform_parameters["messages"],
message_history_limiter_apply_transform_parameters["expected_messages_len"],
):
test_message_history_limiter_apply_transform(msg_history_limiter, messages, expected_messages_len)
for messages, expected_messages_len in zip(
message_history_limiter_apply_transform_parameters["messages"],
message_history_limiter_apply_transform_parameters["expected_messages_len"],
):
test_message_history_limiter_apply_transform_keep_first(
msg_history_limiter_keep_first, messages, expected_messages_len
)
for messages, expected_logs, expected_effect in zip(
message_history_limiter_get_logs_parameters["messages"],
message_history_limiter_get_logs_parameters["expected_logs"],
message_history_limiter_get_logs_parameters["expected_effect"],
):
test_message_history_limiter_get_logs(msg_history_limiter, messages, expected_logs, expected_effect)
# Call the MessageTokenLimiter tests
for messages, expected_token_count, expected_messages_len in zip(
message_token_limiter_apply_transform_parameters["messages"],
message_token_limiter_apply_transform_parameters["expected_token_count"],
message_token_limiter_apply_transform_parameters["expected_messages_len"],
):
test_message_token_limiter_apply_transform(
msg_token_limiter, messages, expected_token_count, expected_messages_len
)
for messages, expected_token_count, expected_messages_len in zip(
message_token_limiter_with_threshold_apply_transform_parameters["messages"],
message_token_limiter_with_threshold_apply_transform_parameters["expected_token_count"],
message_token_limiter_with_threshold_apply_transform_parameters["expected_messages_len"],
):
test_message_token_limiter_with_threshold_apply_transform(
msg_token_limiter_with_threshold, messages, expected_token_count, expected_messages_len
)
for messages, expected_logs, expected_effect in zip(
message_token_limiter_get_logs_parameters["messages"],
message_token_limiter_get_logs_parameters["expected_logs"],
message_token_limiter_get_logs_parameters["expected_effect"],
):
test_message_token_limiter_get_logs(msg_token_limiter, messages, expected_logs, expected_effect)