Public/kotaemon
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[AUR-385, AUR-388] Declare BaseComponent and decide LLM call interface (#2)

Browse Source 
- Use cases related to LLM call: https://cinnamon-ai.atlassian.net/browse/AUR-388?focusedCommentId=34873
- Sample usages: `test_llms_chat_models.py` and `test_llms_completion_models.py`:

```python
from kotaemon.llms.chats.openai import AzureChatOpenAI

model = AzureChatOpenAI(
    openai_api_base="https://test.openai.azure.com/",
    openai_api_key="some-key",
    openai_api_version="2023-03-15-preview",
    deployment_name="gpt35turbo",
    temperature=0,
    request_timeout=60,
)
output = model("hello world")
```

For the LLM-call component, I decide to wrap around Langchain's LLM models and Langchain's Chat models. And set the interface as follow:

- Completion LLM component:
```python
class CompletionLLM:

    def run_raw(self, text: str) -> LLMInterface:
        # Run text completion: str in -> LLMInterface out

    def run_batch_raw(self, text: list[str]) -> list[LLMInterface]:
        # Run text completion in batch: list[str] in -> list[LLMInterface] out

# run_document and run_batch_document just reuse run_raw and run_batch_raw, due to unclear use case
```

- Chat LLM component:
```python
class ChatLLM:
    def run_raw(self, text: str) -> LLMInterface:
        # Run chat completion (no chat history): str in -> LLMInterface out

    def run_batch_raw(self, text: list[str]) -> list[LLMInterface]:
        # Run chat completion in batch mode (no chat history): list[str] in -> list[LLMInterface] out

    def run_document(self, text: list[BaseMessage]) -> LLMInterface:
        # Run chat completion (with chat history): list[langchain's BaseMessage] in -> LLMInterface out

    def run_batch_document(self, text: list[list[BaseMessage]]) -> list[LLMInterface]:
        # Run chat completion in batch mode (with chat history): list[list[langchain's BaseMessage]] in -> list[LLMInterface] out
```

- The LLMInterface is as follow:

```python
@dataclass
class LLMInterface:
    text: list[str]
    completion_tokens: int = -1
    total_tokens: int = -1
    prompt_tokens: int = -1
    logits: list[list[float]] = field(default_factory=list)
```
This commit is contained in:
Nguyen Trung Duc (john)
2023-08-29 15:47:12 +07:00
committed by GitHub
parent e9d1d5c118
commit c3c25db48c
16 changed files with 472 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

78
tests/test_llms_chat_models.py Normal file
View File

@@ -0,0 +1,78 @@
from unittest.mock import patch
from langchain.chat_models import AzureChatOpenAI as AzureChatOpenAILC
from langchain.schema.messages import (
SystemMessage,
HumanMessage,
AIMessage,
)
from kotaemon.llms.chats.openai import AzureChatOpenAI
from kotaemon.llms.base import LLMInterface
_openai_chat_completion_response = {
"id": "chatcmpl-7qyuw6Q1CFCpcKsMdFkmUPUa7JP2x",
"object": "chat.completion",
"created": 1692338378,
"model": "gpt-35-turbo",
"choices": [
{
"index": 0,
"finish_reason": "stop",
"message": {
"role": "assistant",
"content": "Hello! How can I assist you today?",
},
}
],
"usage": {"completion_tokens": 9, "prompt_tokens": 10, "total_tokens": 19},
}
@patch(
"openai.api_resources.chat_completion.ChatCompletion.create",
side_effect=lambda *args, **kwargs: _openai_chat_completion_response,
)
def test_azureopenai_model(openai_completion):
model = AzureChatOpenAI(
openai_api_base="https://test.openai.azure.com/",
openai_api_key="some-key",
openai_api_version="2023-03-15-preview",
deployment_name="gpt35turbo",
temperature=0,
request_timeout=60,
)
assert isinstance(
model.agent, AzureChatOpenAILC
), "Agent not wrapped in Langchain's AzureChatOpenAI"
# test for str input - stream mode
output = model("hello world")
assert isinstance(output, LLMInterface), "Output for single text is not LLMInterface"
openai_completion.assert_called()
# test for list[str] input - batch mode
output = model(["hello world"])
assert isinstance(output, list), "Output for batch string is not a list"
assert isinstance(output[0], LLMInterface), "Output for text is not LLMInterface"
openai_completion.assert_called()
# test for list[message] input - stream mode
messages = [
SystemMessage(content="You are a philosohper"),
HumanMessage(content="What is the meaning of life"),
AIMessage(content="42"),
HumanMessage(content="What is the meaning of 42"),
]
output = model(messages)
assert isinstance(output, LLMInterface), "Output for single text is not LLMInterface"
openai_completion.assert_called()
# test for list[list[message]] input - batch mode
output = model([messages])
assert isinstance(output, list), "Output for batch string is not a list"
assert isinstance(output[0], LLMInterface), "Output for text is not LLMInterface"
openai_completion.assert_called()

70
tests/test_llms_completion_models.py Normal file
View File

@@ -0,0 +1,70 @@
from unittest.mock import patch
from langchain.llms import AzureOpenAI as AzureOpenAILC, OpenAI as OpenAILC
from kotaemon.llms.completions.openai import AzureOpenAI, OpenAI
from kotaemon.llms.base import LLMInterface
_openai_completion_response = {
"id": "cmpl-7qyNoIo6gRSCJR0hi8o3ZKBH4RkJ0",
"object": "sample text_completion",
"created": 1392751226,
"model": "gpt-35-turbo",
"choices": [
{"text": "completion", "index": 0, "finish_reason": "length", "logprobs": None}
],
"usage": {"completion_tokens": 20, "prompt_tokens": 2, "total_tokens": 22},
}
@patch(
"openai.api_resources.completion.Completion.create",
side_effect=lambda *args, **kwargs: _openai_completion_response,
)
def test_azureopenai_model(openai_completion):
model = AzureOpenAI(
openai_api_base="https://test.openai.azure.com/",
openai_api_key="some-key",
openai_api_version="2023-03-15-preview",
deployment_name="gpt35turbo",
temperature=0,
request_timeout=60,
)
assert isinstance(
model.agent, AzureOpenAILC
), "Agent not wrapped in Langchain's AzureOpenAI"
output = model(["hello world"])
assert isinstance(output, list), "Output for batch is not a list"
assert isinstance(output[0], LLMInterface), "Output for text is not LLMInterface"
openai_completion.assert_called()
output = model("hello world")
assert isinstance(output, LLMInterface), "Output for single text is not LLMInterface"
@patch(
"openai.api_resources.completion.Completion.create",
side_effect=lambda *args, **kwargs: _openai_completion_response,
)
def test_openai_model(openai_completion):
model = OpenAI(
openai_api_base="https://test.openai.azure.com/",
openai_api_key="some-key",
openai_api_version="2023-03-15-preview",
deployment_name="gpt35turbo",
temperature=0,
request_timeout=60,
)
assert isinstance(
model.agent, OpenAILC
), "Agent is not wrapped in Langchain's OpenAI"
output = model(["hello world"])
assert isinstance(output, list), "Output for batch is not a list"
assert isinstance(output[0], LLMInterface), "Output for text is not LLMInterface"
openai_completion.assert_called()
output = model("hello world")
assert isinstance(output, LLMInterface), "Output for single text is not LLMInterface"

29
tests/test_telemetry.py
View File

@@ -1,3 +1,31 @@
import os
import sys
import pytest
@pytest.fixture
def clean_artifacts_for_telemetry():
try:
del sys.modules["kotaemon"]
except KeyError:
pass
try:
del sys.modules["haystack"]
except KeyError:
pass
try:
del sys.modules["haystack.telemetry"]
except KeyError:
pass
if "HAYSTACK_TELEMETRY_ENABLED" in os.environ:
del os.environ["HAYSTACK_TELEMETRY_ENABLED"]
@pytest.mark.usefixtures("clean_artifacts_for_telemetry")
def test_disable_telemetry_import_haystack_first():
"""Test that telemetry is disabled when kotaemon lib is initiated after"""
import os
@@ -10,6 +38,7 @@ def test_disable_telemetry_import_haystack_first():
assert os.environ.get("HAYSTACK_TELEMETRY_ENABLED", "True") == "False"
@pytest.mark.usefixtures("clean_artifacts_for_telemetry")
def test_disable_telemetry_import_haystack_after_kotaemon():
"""Test that telemetry is disabled when kotaemon lib is initiated before"""
import os