mirror of
https://github.com/trustgraph-ai/trustgraph.git
synced 2026-07-11 22:32:12 +02:00
feat: add MiniMax as LLM text completion provider
- Add MiniMax chat model provider using OpenAI-compatible API - Support MiniMax-M2.7 and MiniMax-M2.7-highspeed models - Temperature clamping to MiniMax valid range (0.0, 1.0] - Streaming support with token usage tracking - MINIMAX_API_KEY environment variable support - Add text-completion-minimax entry point - Add 15 unit tests and 3 integration tests - Update README with MiniMax in LLM APIs list
This commit is contained in:
PR Bot
parent
d30857b5c3
commit
50f24b8e2a
8 changed files with 960 additions and 4 deletions
|
|
@ -38,7 +38,7 @@ The platform:
|
|||
- [x] Deploy locally with Docker
|
||||
- [x] Deploy in cloud with Kubernetes
|
||||
- [x] Support for all major LLMs
|
||||
- [x] API support for Anthropic, Cohere, Gemini, Mistral, OpenAI, and others
|
||||
- [x] API support for Anthropic, Cohere, Gemini, MiniMax, Mistral, OpenAI, and others
|
||||
- [x] Model inferencing with vLLM, Ollama, TGI, LM Studio, and Llamafiles
|
||||
- [x] Developer friendly
|
||||
- [x] REST API [Docs](https://docs.trustgraph.ai/reference/apis/rest.html)
|
||||
|
|
@ -50,7 +50,7 @@ The platform:
|
|||
|
||||
How many times have you cloned a repo and opened the `.env.example` to see the dozens of API keys for 3rd party dependencies needed to make the services work? There are only 3 things in TrustGraph that might need an API key:
|
||||
|
||||
- 3rd party LLM services like Anthropic, Cohere, Gemini, Mistral, OpenAI, etc.
|
||||
- 3rd party LLM services like Anthropic, Cohere, Gemini, MiniMax, Mistral, OpenAI, etc.
|
||||
- 3rd party OCR like Mistral OCR
|
||||
- The API key *you set* for the TrustGraph API gateway
|
||||
|
||||
|
|
@ -161,6 +161,7 @@ TrustGraph provides component flexibility to optimize agent workflows.
|
|||
- Cohere<br>
|
||||
- Google AI Studio<br>
|
||||
- Google VertexAI<br>
|
||||
- MiniMax<br>
|
||||
- Mistral<br>
|
||||
- OpenAI<br>
|
||||
|
||||
|
|
|
|||
86
tests/integration/test_minimax_integration.py
Normal file
86
tests/integration/test_minimax_integration.py
Normal file
|
|
@ -0,0 +1,86 @@
|
|||
"""
|
||||
Integration tests for MiniMax text completion provider.
|
||||
Tests actual API calls to MiniMax when MINIMAX_API_KEY is available.
|
||||
"""
|
||||
|
||||
import os
|
||||
import pytest
|
||||
from openai import OpenAI
|
||||
|
||||
MINIMAX_API_KEY = os.getenv("MINIMAX_API_KEY")
|
||||
MINIMAX_BASE_URL = "https://api.minimax.io/v1"
|
||||
|
||||
|
||||
@pytest.mark.skipif(
|
||||
not MINIMAX_API_KEY,
|
||||
reason="MINIMAX_API_KEY not set"
|
||||
)
|
||||
class TestMiniMaxIntegration:
|
||||
"""Integration tests that call the real MiniMax API"""
|
||||
|
||||
def get_client(self):
|
||||
return OpenAI(api_key=MINIMAX_API_KEY, base_url=MINIMAX_BASE_URL)
|
||||
|
||||
def test_basic_chat_completion(self):
|
||||
"""Test basic chat completion with MiniMax-M2.7"""
|
||||
client = self.get_client()
|
||||
|
||||
resp = client.chat.completions.create(
|
||||
model="MiniMax-M2.7",
|
||||
messages=[{
|
||||
"role": "user",
|
||||
"content": 'Say "test passed" and nothing else.'
|
||||
}],
|
||||
temperature=1.0,
|
||||
max_tokens=64,
|
||||
)
|
||||
|
||||
assert resp.choices[0].message.content is not None
|
||||
assert len(resp.choices[0].message.content) > 0
|
||||
assert resp.usage.prompt_tokens > 0
|
||||
assert resp.usage.completion_tokens > 0
|
||||
|
||||
def test_highspeed_model(self):
|
||||
"""Test chat completion with MiniMax-M2.7-highspeed"""
|
||||
client = self.get_client()
|
||||
|
||||
resp = client.chat.completions.create(
|
||||
model="MiniMax-M2.7-highspeed",
|
||||
messages=[{
|
||||
"role": "user",
|
||||
"content": 'Say "highspeed test passed" and nothing else.'
|
||||
}],
|
||||
temperature=1.0,
|
||||
max_tokens=64,
|
||||
)
|
||||
|
||||
assert resp.choices[0].message.content is not None
|
||||
assert len(resp.choices[0].message.content) > 0
|
||||
|
||||
def test_streaming_completion(self):
|
||||
"""Test streaming chat completion"""
|
||||
client = self.get_client()
|
||||
|
||||
response = client.chat.completions.create(
|
||||
model="MiniMax-M2.7",
|
||||
messages=[{
|
||||
"role": "user",
|
||||
"content": 'Say "stream test passed" and nothing else.'
|
||||
}],
|
||||
temperature=1.0,
|
||||
max_tokens=64,
|
||||
stream=True,
|
||||
stream_options={"include_usage": True}
|
||||
)
|
||||
|
||||
chunks = list(response)
|
||||
assert len(chunks) > 0
|
||||
|
||||
# Collect text from chunks
|
||||
texts = []
|
||||
for chunk in chunks:
|
||||
if chunk.choices and chunk.choices[0].delta.content:
|
||||
texts.append(chunk.choices[0].delta.content)
|
||||
|
||||
full_text = "".join(texts)
|
||||
assert len(full_text) > 0
|
||||
|
|
@ -494,6 +494,45 @@ def mock_llamafile_client():
|
|||
mock_response.choices[0].message.content = "Test response from LlamaFile"
|
||||
mock_response.usage.prompt_tokens = 14
|
||||
mock_response.usage.completion_tokens = 8
|
||||
|
||||
|
||||
mock_client.chat.completions.create.return_value = mock_response
|
||||
return mock_client
|
||||
return mock_client
|
||||
|
||||
|
||||
# === MiniMax Specific Fixtures ===
|
||||
|
||||
@pytest.fixture
|
||||
def minimax_processor_config(base_processor_config):
|
||||
"""Default configuration for MiniMax processor"""
|
||||
config = base_processor_config.copy()
|
||||
config.update({
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.0,
|
||||
'max_output': 4096
|
||||
})
|
||||
return config
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
def mock_minimax_client():
|
||||
"""Mock OpenAI client for MiniMax"""
|
||||
mock_client = MagicMock()
|
||||
|
||||
# Mock the response structure
|
||||
mock_response = MagicMock()
|
||||
mock_response.choices = [MagicMock()]
|
||||
mock_response.choices[0].message.content = "Test response from MiniMax"
|
||||
mock_response.usage.prompt_tokens = 18
|
||||
mock_response.usage.completion_tokens = 10
|
||||
|
||||
mock_client.chat.completions.create.return_value = mock_response
|
||||
return mock_client
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
def mock_minimax_rate_limit_error():
|
||||
"""Mock MiniMax (OpenAI) rate limit error"""
|
||||
from openai import RateLimitError
|
||||
return RateLimitError("Rate limit exceeded", response=MagicMock(), body=None)
|
||||
570
tests/unit/test_text_completion/test_minimax_processor.py
Normal file
570
tests/unit/test_text_completion/test_minimax_processor.py
Normal file
|
|
@ -0,0 +1,570 @@
|
|||
"""
|
||||
Unit tests for trustgraph.model.text_completion.minimax
|
||||
Following the same successful pattern as OpenAI and other provider tests
|
||||
"""
|
||||
|
||||
import pytest
|
||||
from unittest.mock import AsyncMock, MagicMock, patch
|
||||
from unittest import IsolatedAsyncioTestCase
|
||||
|
||||
# Import the service under test
|
||||
from trustgraph.model.text_completion.minimax.llm import Processor
|
||||
from trustgraph.base import LlmResult
|
||||
from trustgraph.exceptions import TooManyRequests
|
||||
|
||||
|
||||
class TestMiniMaxProcessorSimple(IsolatedAsyncioTestCase):
|
||||
"""Test MiniMax processor functionality"""
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_processor_initialization_basic(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test basic processor initialization"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.0,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
# Act
|
||||
processor = Processor(**config)
|
||||
|
||||
# Assert
|
||||
assert processor.default_model == 'MiniMax-M2.7'
|
||||
assert processor.temperature == 1.0
|
||||
assert processor.max_output == 4096
|
||||
assert hasattr(processor, 'openai')
|
||||
mock_openai_class.assert_called_once_with(base_url='https://api.minimax.io/v1', api_key='test-api-key')
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_generate_content_success(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test successful content generation"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_response = MagicMock()
|
||||
mock_response.choices = [MagicMock()]
|
||||
mock_response.choices[0].message.content = "Generated response from MiniMax"
|
||||
mock_response.usage.prompt_tokens = 20
|
||||
mock_response.usage.completion_tokens = 12
|
||||
|
||||
mock_openai_client.chat.completions.create.return_value = mock_response
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.0,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
processor = Processor(**config)
|
||||
|
||||
# Act
|
||||
result = await processor.generate_content("System prompt", "User prompt")
|
||||
|
||||
# Assert
|
||||
assert isinstance(result, LlmResult)
|
||||
assert result.text == "Generated response from MiniMax"
|
||||
assert result.in_token == 20
|
||||
assert result.out_token == 12
|
||||
assert result.model == 'MiniMax-M2.7'
|
||||
|
||||
# Verify the API call
|
||||
mock_openai_client.chat.completions.create.assert_called_once_with(
|
||||
model='MiniMax-M2.7',
|
||||
messages=[{
|
||||
"role": "user",
|
||||
"content": [{
|
||||
"type": "text",
|
||||
"text": "System prompt\n\nUser prompt"
|
||||
}]
|
||||
}],
|
||||
temperature=1.0,
|
||||
max_tokens=4096
|
||||
)
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_generate_content_rate_limit_error(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test rate limit error handling"""
|
||||
# Arrange
|
||||
from openai import RateLimitError
|
||||
|
||||
mock_openai_client = MagicMock()
|
||||
mock_openai_client.chat.completions.create.side_effect = RateLimitError("Rate limit exceeded", response=MagicMock(), body=None)
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.0,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
processor = Processor(**config)
|
||||
|
||||
# Act & Assert
|
||||
with pytest.raises(TooManyRequests):
|
||||
await processor.generate_content("System prompt", "User prompt")
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_generate_content_generic_exception(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test handling of generic exceptions"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_openai_client.chat.completions.create.side_effect = Exception("API connection error")
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.0,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
processor = Processor(**config)
|
||||
|
||||
# Act & Assert
|
||||
with pytest.raises(Exception, match="API connection error"):
|
||||
await processor.generate_content("System prompt", "User prompt")
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_processor_initialization_without_api_key(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test processor initialization without API key (should fail)"""
|
||||
# Arrange
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': None, # No API key provided
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.0,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
# Act & Assert
|
||||
with pytest.raises(RuntimeError, match="MiniMax API key not specified"):
|
||||
processor = Processor(**config)
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_processor_initialization_with_custom_parameters(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test processor initialization with custom parameters"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7-highspeed',
|
||||
'api_key': 'custom-api-key',
|
||||
'url': 'https://custom-minimax-url.com/v1',
|
||||
'temperature': 0.7,
|
||||
'max_output': 2048,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
# Act
|
||||
processor = Processor(**config)
|
||||
|
||||
# Assert
|
||||
assert processor.default_model == 'MiniMax-M2.7-highspeed'
|
||||
assert processor.temperature == 0.7
|
||||
assert processor.max_output == 2048
|
||||
mock_openai_class.assert_called_once_with(base_url='https://custom-minimax-url.com/v1', api_key='custom-api-key')
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_processor_initialization_with_defaults(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test processor initialization with default values"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
# Only provide required fields, should use defaults
|
||||
config = {
|
||||
'api_key': 'test-api-key',
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
# Act
|
||||
processor = Processor(**config)
|
||||
|
||||
# Assert
|
||||
assert processor.default_model == 'MiniMax-M2.7' # default_model
|
||||
assert processor.temperature == 1.0 # default_temperature
|
||||
assert processor.max_output == 4096 # default_max_output
|
||||
mock_openai_class.assert_called_once_with(base_url='https://api.minimax.io/v1', api_key='test-api-key')
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_temperature_clamping_zero(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test that temperature 0.0 is clamped to 0.01"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 0.0, # Should be clamped to 0.01
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
# Act
|
||||
processor = Processor(**config)
|
||||
|
||||
# Assert
|
||||
assert processor.temperature == 0.01 # Clamped from 0.0
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_temperature_clamping_above_one(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test that temperature above 1.0 is clamped to 1.0"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.5, # Should be clamped to 1.0
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
# Act
|
||||
processor = Processor(**config)
|
||||
|
||||
# Assert
|
||||
assert processor.temperature == 1.0 # Clamped from 1.5
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_temperature_clamping_at_runtime(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test that runtime temperature override is also clamped"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_response = MagicMock()
|
||||
mock_response.choices = [MagicMock()]
|
||||
mock_response.choices[0].message.content = "Response"
|
||||
mock_response.usage.prompt_tokens = 10
|
||||
mock_response.usage.completion_tokens = 5
|
||||
|
||||
mock_openai_client.chat.completions.create.return_value = mock_response
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 0.5,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
processor = Processor(**config)
|
||||
|
||||
# Act - Override temperature with 0.0 at runtime
|
||||
await processor.generate_content("System", "User", temperature=0.0)
|
||||
|
||||
# Assert - temperature should be clamped to 0.01
|
||||
call_kwargs = mock_openai_client.chat.completions.create.call_args.kwargs
|
||||
assert call_kwargs['temperature'] == 0.01
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_generate_content_temperature_override(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test temperature parameter override functionality"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_response = MagicMock()
|
||||
mock_response.choices = [MagicMock()]
|
||||
mock_response.choices[0].message.content = "Response with custom temperature"
|
||||
mock_response.usage.prompt_tokens = 15
|
||||
mock_response.usage.completion_tokens = 10
|
||||
|
||||
mock_openai_client.chat.completions.create.return_value = mock_response
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.0,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
processor = Processor(**config)
|
||||
|
||||
# Act - Override temperature at runtime
|
||||
result = await processor.generate_content(
|
||||
"System prompt",
|
||||
"User prompt",
|
||||
model=None,
|
||||
temperature=0.9
|
||||
)
|
||||
|
||||
# Assert
|
||||
assert isinstance(result, LlmResult)
|
||||
assert result.text == "Response with custom temperature"
|
||||
call_kwargs = mock_openai_client.chat.completions.create.call_args.kwargs
|
||||
assert call_kwargs['temperature'] == 0.9
|
||||
assert call_kwargs['model'] == 'MiniMax-M2.7'
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_generate_content_model_override(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test model parameter override functionality"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_response = MagicMock()
|
||||
mock_response.choices = [MagicMock()]
|
||||
mock_response.choices[0].message.content = "Response with custom model"
|
||||
mock_response.usage.prompt_tokens = 15
|
||||
mock_response.usage.completion_tokens = 10
|
||||
|
||||
mock_openai_client.chat.completions.create.return_value = mock_response
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 0.5,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
processor = Processor(**config)
|
||||
|
||||
# Act - Override model at runtime
|
||||
result = await processor.generate_content(
|
||||
"System prompt",
|
||||
"User prompt",
|
||||
model="MiniMax-M2.7-highspeed",
|
||||
temperature=None
|
||||
)
|
||||
|
||||
# Assert
|
||||
assert isinstance(result, LlmResult)
|
||||
assert result.text == "Response with custom model"
|
||||
call_kwargs = mock_openai_client.chat.completions.create.call_args.kwargs
|
||||
assert call_kwargs['model'] == 'MiniMax-M2.7-highspeed'
|
||||
assert call_kwargs['temperature'] == 0.5
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_supports_streaming(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test that MiniMax provider reports streaming support"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.0,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
processor = Processor(**config)
|
||||
|
||||
# Assert
|
||||
assert processor.supports_streaming() is True
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_generate_content_empty_prompts(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test content generation with empty prompts"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_response = MagicMock()
|
||||
mock_response.choices = [MagicMock()]
|
||||
mock_response.choices[0].message.content = "Default response"
|
||||
mock_response.usage.prompt_tokens = 2
|
||||
mock_response.usage.completion_tokens = 3
|
||||
|
||||
mock_openai_client.chat.completions.create.return_value = mock_response
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 1.0,
|
||||
'max_output': 4096,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
processor = Processor(**config)
|
||||
|
||||
# Act
|
||||
result = await processor.generate_content("", "")
|
||||
|
||||
# Assert
|
||||
assert isinstance(result, LlmResult)
|
||||
assert result.text == "Default response"
|
||||
call_args = mock_openai_client.chat.completions.create.call_args
|
||||
expected_prompt = "\n\n"
|
||||
assert call_args[1]['messages'][0]['content'][0]['text'] == expected_prompt
|
||||
|
||||
@patch('trustgraph.model.text_completion.minimax.llm.OpenAI')
|
||||
@patch('trustgraph.base.async_processor.AsyncProcessor.__init__')
|
||||
@patch('trustgraph.base.llm_service.LlmService.__init__')
|
||||
async def test_generate_content_message_structure(self, mock_llm_init, mock_async_init, mock_openai_class):
|
||||
"""Test that messages are structured correctly"""
|
||||
# Arrange
|
||||
mock_openai_client = MagicMock()
|
||||
mock_response = MagicMock()
|
||||
mock_response.choices = [MagicMock()]
|
||||
mock_response.choices[0].message.content = "Response with proper structure"
|
||||
mock_response.usage.prompt_tokens = 25
|
||||
mock_response.usage.completion_tokens = 15
|
||||
|
||||
mock_openai_client.chat.completions.create.return_value = mock_response
|
||||
mock_openai_class.return_value = mock_openai_client
|
||||
|
||||
mock_async_init.return_value = None
|
||||
mock_llm_init.return_value = None
|
||||
|
||||
config = {
|
||||
'model': 'MiniMax-M2.7',
|
||||
'api_key': 'test-api-key',
|
||||
'url': 'https://api.minimax.io/v1',
|
||||
'temperature': 0.5,
|
||||
'max_output': 1024,
|
||||
'concurrency': 1,
|
||||
'taskgroup': AsyncMock(),
|
||||
'id': 'test-processor'
|
||||
}
|
||||
|
||||
processor = Processor(**config)
|
||||
|
||||
# Act
|
||||
result = await processor.generate_content("You are a helpful assistant", "What is AI?")
|
||||
|
||||
# Assert
|
||||
assert result.text == "Response with proper structure"
|
||||
assert result.in_token == 25
|
||||
assert result.out_token == 15
|
||||
|
||||
call_args = mock_openai_client.chat.completions.create.call_args
|
||||
messages = call_args[1]['messages']
|
||||
|
||||
assert len(messages) == 1
|
||||
assert messages[0]['role'] == 'user'
|
||||
assert messages[0]['content'][0]['type'] == 'text'
|
||||
assert messages[0]['content'][0]['text'] == "You are a helpful assistant\n\nWhat is AI?"
|
||||
|
||||
assert call_args[1]['model'] == 'MiniMax-M2.7'
|
||||
assert call_args[1]['temperature'] == 0.5
|
||||
assert call_args[1]['max_tokens'] == 1024
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
pytest.main([__file__])
|
||||
|
|
@ -108,6 +108,7 @@ text-completion-claude = "trustgraph.model.text_completion.claude:run"
|
|||
text-completion-cohere = "trustgraph.model.text_completion.cohere:run"
|
||||
text-completion-llamafile = "trustgraph.model.text_completion.llamafile:run"
|
||||
text-completion-lmstudio = "trustgraph.model.text_completion.lmstudio:run"
|
||||
text-completion-minimax = "trustgraph.model.text_completion.minimax:run"
|
||||
text-completion-mistral = "trustgraph.model.text_completion.mistral:run"
|
||||
text-completion-ollama = "trustgraph.model.text_completion.ollama:run"
|
||||
text-completion-openai = "trustgraph.model.text_completion.openai:run"
|
||||
|
|
|
|||
|
|
@ -0,0 +1,2 @@
|
|||
|
||||
from . llm import *
|
||||
|
|
@ -0,0 +1,6 @@
|
|||
#!/usr/bin/env python3
|
||||
|
||||
from . llm import run
|
||||
|
||||
if __name__ == '__main__':
|
||||
run()
|
||||
251
trustgraph-flow/trustgraph/model/text_completion/minimax/llm.py
Normal file
251
trustgraph-flow/trustgraph/model/text_completion/minimax/llm.py
Normal file
|
|
@ -0,0 +1,251 @@
|
|||
|
||||
"""
|
||||
Simple LLM service, performs text prompt completion using MiniMax.
|
||||
Input is prompt, output is response.
|
||||
"""
|
||||
|
||||
from openai import OpenAI, RateLimitError
|
||||
import os
|
||||
import logging
|
||||
|
||||
from .... exceptions import TooManyRequests
|
||||
from .... base import LlmService, LlmResult, LlmChunk
|
||||
|
||||
# Module logger
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
default_ident = "text-completion"
|
||||
|
||||
default_model = 'MiniMax-M2.7'
|
||||
default_temperature = 1.0
|
||||
default_max_output = 4096
|
||||
default_api_key = os.getenv("MINIMAX_API_KEY")
|
||||
default_base_url = os.getenv("MINIMAX_BASE_URL")
|
||||
|
||||
if default_base_url is None or default_base_url == "":
|
||||
default_base_url = "https://api.minimax.io/v1"
|
||||
|
||||
class Processor(LlmService):
|
||||
|
||||
def __init__(self, **params):
|
||||
|
||||
model = params.get("model", default_model)
|
||||
api_key = params.get("api_key", default_api_key)
|
||||
base_url = params.get("url", default_base_url)
|
||||
temperature = params.get("temperature", default_temperature)
|
||||
max_output = params.get("max_output", default_max_output)
|
||||
|
||||
if api_key is None:
|
||||
raise RuntimeError("MiniMax API key not specified")
|
||||
|
||||
# Clamp temperature to MiniMax's valid range (0.0, 1.0]
|
||||
if temperature is not None:
|
||||
if temperature <= 0.0:
|
||||
temperature = 0.01
|
||||
elif temperature > 1.0:
|
||||
temperature = 1.0
|
||||
|
||||
super(Processor, self).__init__(
|
||||
**params | {
|
||||
"model": model,
|
||||
"temperature": temperature,
|
||||
"max_output": max_output,
|
||||
"base_url": base_url,
|
||||
}
|
||||
)
|
||||
|
||||
self.default_model = model
|
||||
self.temperature = temperature
|
||||
self.max_output = max_output
|
||||
|
||||
self.openai = OpenAI(base_url=base_url, api_key=api_key)
|
||||
|
||||
logger.info("MiniMax LLM service initialized")
|
||||
|
||||
def _clamp_temperature(self, temperature):
|
||||
"""Clamp temperature to MiniMax's valid range (0.0, 1.0]"""
|
||||
if temperature is None:
|
||||
return self.temperature
|
||||
if temperature <= 0.0:
|
||||
return 0.01
|
||||
if temperature > 1.0:
|
||||
return 1.0
|
||||
return temperature
|
||||
|
||||
async def generate_content(self, system, prompt, model=None, temperature=None):
|
||||
|
||||
# Use provided model or fall back to default
|
||||
model_name = model or self.default_model
|
||||
# Use provided temperature or fall back to default
|
||||
effective_temperature = self._clamp_temperature(temperature)
|
||||
|
||||
logger.debug(f"Using model: {model_name}")
|
||||
logger.debug(f"Using temperature: {effective_temperature}")
|
||||
|
||||
prompt = system + "\n\n" + prompt
|
||||
|
||||
try:
|
||||
|
||||
resp = self.openai.chat.completions.create(
|
||||
model=model_name,
|
||||
messages=[
|
||||
{
|
||||
"role": "user",
|
||||
"content": [
|
||||
{
|
||||
"type": "text",
|
||||
"text": prompt
|
||||
}
|
||||
]
|
||||
}
|
||||
],
|
||||
temperature=effective_temperature,
|
||||
max_tokens=self.max_output,
|
||||
)
|
||||
|
||||
inputtokens = resp.usage.prompt_tokens
|
||||
outputtokens = resp.usage.completion_tokens
|
||||
logger.debug(f"LLM response: {resp.choices[0].message.content}")
|
||||
logger.info(f"Input Tokens: {inputtokens}")
|
||||
logger.info(f"Output Tokens: {outputtokens}")
|
||||
|
||||
resp = LlmResult(
|
||||
text = resp.choices[0].message.content,
|
||||
in_token = inputtokens,
|
||||
out_token = outputtokens,
|
||||
model = model_name
|
||||
)
|
||||
|
||||
return resp
|
||||
|
||||
# FIXME: Wrong exception, don't know what this LLM throws
|
||||
# for a rate limit
|
||||
except RateLimitError:
|
||||
|
||||
# Leave rate limit retries to the base handler
|
||||
raise TooManyRequests()
|
||||
|
||||
except Exception as e:
|
||||
|
||||
# Apart from rate limits, treat all exceptions as unrecoverable
|
||||
|
||||
logger.error(f"MiniMax LLM exception ({type(e).__name__}): {e}", exc_info=True)
|
||||
raise e
|
||||
|
||||
def supports_streaming(self):
|
||||
"""MiniMax supports streaming"""
|
||||
return True
|
||||
|
||||
async def generate_content_stream(self, system, prompt, model=None, temperature=None):
|
||||
"""
|
||||
Stream content generation from MiniMax.
|
||||
Yields LlmChunk objects with is_final=True on the last chunk.
|
||||
"""
|
||||
# Use provided model or fall back to default
|
||||
model_name = model or self.default_model
|
||||
# Use provided temperature or fall back to default
|
||||
effective_temperature = self._clamp_temperature(temperature)
|
||||
|
||||
logger.debug(f"Using model (streaming): {model_name}")
|
||||
logger.debug(f"Using temperature: {effective_temperature}")
|
||||
|
||||
prompt = system + "\n\n" + prompt
|
||||
|
||||
try:
|
||||
response = self.openai.chat.completions.create(
|
||||
model=model_name,
|
||||
messages=[
|
||||
{
|
||||
"role": "user",
|
||||
"content": [
|
||||
{
|
||||
"type": "text",
|
||||
"text": prompt
|
||||
}
|
||||
]
|
||||
}
|
||||
],
|
||||
temperature=effective_temperature,
|
||||
max_tokens=self.max_output,
|
||||
stream=True,
|
||||
stream_options={"include_usage": True}
|
||||
)
|
||||
|
||||
total_input_tokens = 0
|
||||
total_output_tokens = 0
|
||||
|
||||
# Stream chunks
|
||||
for chunk in response:
|
||||
if chunk.choices and chunk.choices[0].delta.content:
|
||||
yield LlmChunk(
|
||||
text=chunk.choices[0].delta.content,
|
||||
in_token=None,
|
||||
out_token=None,
|
||||
model=model_name,
|
||||
is_final=False
|
||||
)
|
||||
|
||||
# Capture usage from final chunk
|
||||
if chunk.usage:
|
||||
total_input_tokens = chunk.usage.prompt_tokens
|
||||
total_output_tokens = chunk.usage.completion_tokens
|
||||
|
||||
# Send final chunk with token counts
|
||||
yield LlmChunk(
|
||||
text="",
|
||||
in_token=total_input_tokens,
|
||||
out_token=total_output_tokens,
|
||||
model=model_name,
|
||||
is_final=True
|
||||
)
|
||||
|
||||
logger.debug("Streaming complete")
|
||||
|
||||
except RateLimitError:
|
||||
logger.warning("Hit rate limit during streaming")
|
||||
raise TooManyRequests()
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"MiniMax streaming exception ({type(e).__name__}): {e}", exc_info=True)
|
||||
raise e
|
||||
|
||||
@staticmethod
|
||||
def add_args(parser):
|
||||
|
||||
LlmService.add_args(parser)
|
||||
|
||||
parser.add_argument(
|
||||
'-m', '--model',
|
||||
default="MiniMax-M2.7",
|
||||
help=f'LLM model (default: MiniMax-M2.7)'
|
||||
)
|
||||
|
||||
parser.add_argument(
|
||||
'-k', '--api-key',
|
||||
default=default_api_key,
|
||||
help=f'MiniMax API key'
|
||||
)
|
||||
|
||||
parser.add_argument(
|
||||
'-u', '--url',
|
||||
default=default_base_url,
|
||||
help=f'MiniMax service base URL'
|
||||
)
|
||||
|
||||
parser.add_argument(
|
||||
'-t', '--temperature',
|
||||
type=float,
|
||||
default=default_temperature,
|
||||
help=f'LLM temperature parameter (default: {default_temperature})'
|
||||
)
|
||||
|
||||
parser.add_argument(
|
||||
'-x', '--max-output',
|
||||
type=int,
|
||||
default=default_max_output,
|
||||
help=f'LLM max output tokens (default: {default_max_output})'
|
||||
)
|
||||
|
||||
def run():
|
||||
|
||||
Processor.launch(default_ident, __doc__)
|
||||
Loading…
Add table
Add a link
Reference in a new issue