Add MiniMax as a first-class LLM provider

README.md

@@ -36,7 +36,7 @@ Langfun is *simple and elegant*:
 
 *   An intuitive programming model, graspable in 5 minutes;
 *   Plug-and-play into any Python codebase, making an immediate difference;
-*   Comprehensive LLMs under a unified API: Gemini, GPT, Claude, Llama3, and more.
+*   Comprehensive LLMs under a unified API: Gemini, GPT, Claude, MiniMax, Llama3, and more.
 *   Designed for agile developement: offering intellisense, easy debugging, with minimal overhead;
 
 ## Hello, Langfun

langfun/core/llms/__init__.py

@@ -233,6 +233,11 @@
 from langfun.core.llms.deepseek import DeepSeekV3
 from langfun.core.llms.deepseek import DeepSeekR1
 
+# MiniMax models.
+from langfun.core.llms.minimax import MiniMax
+from langfun.core.llms.minimax import MiniMaxM27
+from langfun.core.llms.minimax import MiniMaxM27Highspeed
+
 # LLaMA C++ models.
 from langfun.core.llms.llama_cpp import LlamaCppRemote
 

langfun/core/llms/minimax.py

@@ -0,0 +1,179 @@
+# Copyright 2025 The Langfun Authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Language models from MiniMax."""
+
+import datetime
+import functools
+import os
+from typing import Annotated, Any, Final
+
+import langfun.core as lf
+from langfun.core.llms import openai_compatible
+import pyglove as pg
+
+
+class MiniMaxModelInfo(lf.ModelInfo):
+  """MiniMax model info."""
+
+  LINKS = dict(
+      models='https://platform.minimaxi.com/document/Models',
+      pricing='https://platform.minimaxi.com/document/Price',
+  )
+
+  provider: Final[str] = 'MiniMax'  # pylint: disable=invalid-name
+
+
+SUPPORTED_MODELS = [
+    MiniMaxModelInfo(
+        model_id='MiniMax-M2.7',
+        in_service=True,
+        model_type='instruction-tuned',
+        description='MiniMax M2.7 model with 204K context window.',
+        url='https://platform.minimaxi.com/document/Models',
+        release_date=datetime.datetime(2025, 3, 1),
+        input_modalities=lf.ModelInfo.TEXT_INPUT_ONLY,
+        context_length=lf.ModelInfo.ContextLength(
+            max_input_tokens=204_800,
+            max_output_tokens=65_536,
+        ),
+        pricing=lf.ModelInfo.Pricing(
+            cost_per_1m_input_tokens=1.1,
+            cost_per_1m_output_tokens=4.4,
+        ),
+        rate_limits=None,
+    ),
+    MiniMaxModelInfo(
+        model_id='MiniMax-M2.7-highspeed',
+        in_service=True,
+        model_type='instruction-tuned',
+        description='MiniMax M2.7 Highspeed model with 204K context window.',
+        url='https://platform.minimaxi.com/document/Models',
+        release_date=datetime.datetime(2025, 3, 1),
+        input_modalities=lf.ModelInfo.TEXT_INPUT_ONLY,
+        context_length=lf.ModelInfo.ContextLength(
+            max_input_tokens=204_800,
+            max_output_tokens=65_536,
+        ),
+        pricing=lf.ModelInfo.Pricing(
+            cost_per_1m_input_tokens=0.55,
+            cost_per_1m_output_tokens=2.2,
+        ),
+        rate_limits=None,
+    ),
+]
+
+_SUPPORTED_MODELS_BY_ID = {m.model_id: m for m in SUPPORTED_MODELS}
+
+
+# MiniMax API uses an API format compatible with OpenAI.
+# Reference: https://platform.minimaxi.com/document/ChatCompletion%20v2
+@lf.use_init_args(['model'])
+class MiniMax(openai_compatible.OpenAIChatCompletionAPI):
+  """MiniMax models.
+
+  **Quick Start:**
+
+  ```python
+  import langfun as lf
+
+  # Call MiniMax-M2.7 using API key from environment variable
+  # 'MINIMAX_API_KEY'.
+  lm = lf.llms.MiniMaxM27()
+  r = lm('Who are you?')
+  print(r)
+  ```
+
+  **Setting up API key:**
+
+  The MiniMax API key can be specified in following ways:
+
+  1. At model instantiation:
+
+     ```python
+     lm = lf.llms.MiniMaxM27(api_key='MY_API_KEY')
+     ```
+  2. via environment variable `MINIMAX_API_KEY`.
+
+  **References:**
+
+  *   https://platform.minimaxi.com/document/ChatCompletion%20v2
+  """
+
+  model: pg.typing.Annotated[
+      pg.typing.Enum(
+          pg.MISSING_VALUE, [m.model_id for m in SUPPORTED_MODELS]
+      ),
+      'The name of the model to use.',
+  ]
+
+  api_endpoint: str = 'https://api.minimax.io/v1/chat/completions'
+
+  api_key: Annotated[
+      str | None,
+      (
+          'API key. If None, the key will be read from environment variable '
+          "'MINIMAX_API_KEY'."
+      ),
+  ] = None
+
+  @property
+  def headers(self) -> dict[str, Any]:
+    api_key = self.api_key or os.environ.get('MINIMAX_API_KEY', None)
+    if not api_key:
+      raise ValueError(
+          'Please specify `api_key` during `__init__` or set environment '
+          'variable `MINIMAX_API_KEY` with your MiniMax API key.'
+      )
+    headers = super().headers
+    headers.update({
+        'Authorization': f'Bearer {api_key}',
+    })
+    return headers
+
+  @functools.cached_property
+  def model_info(self) -> MiniMaxModelInfo:
+    return _SUPPORTED_MODELS_BY_ID[self.model]
+
+  def _request_args(
+      self, options: lf.LMSamplingOptions) -> dict[str, Any]:
+    """Returns a dict as request arguments."""
+    args = super()._request_args(options)
+    # MiniMax requires temperature in (0.0, 1.0], clamp 0.0 to 0.01.
+    if 'temperature' in args and args['temperature'] == 0.0:
+      args['temperature'] = 0.01
+    return args
+
+  @classmethod
+  def dir(cls):
+    return [m.model_id for m in SUPPORTED_MODELS if m.in_service]
+
+
+class MiniMaxM27(MiniMax):  # pylint: disable=invalid-name
+  """MiniMax M2.7 model with 204K context window."""
+
+  model = 'MiniMax-M2.7'
+
+
+class MiniMaxM27Highspeed(MiniMax):  # pylint: disable=invalid-name
+  """MiniMax M2.7 Highspeed model with 204K context window."""
+
+  model = 'MiniMax-M2.7-highspeed'
+
+
+def _register_minimax_models():
+  """Registers MiniMax models."""
+  for m in SUPPORTED_MODELS:
+    lf.LanguageModel.register(m.model_id, MiniMax)
+
+_register_minimax_models()

langfun/core/llms/minimax_test.py

@@ -0,0 +1,103 @@
+# Copyright 2025 The Langfun Authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import os
+import unittest
+import langfun.core as lf
+from langfun.core.llms import minimax
+
+
+class MiniMaxTest(unittest.TestCase):
+  """Tests for MiniMax language model."""
+
+  def test_dir(self):
+    self.assertIn('MiniMax-M2.7', minimax.MiniMax.dir())
+    self.assertIn('MiniMax-M2.7-highspeed', minimax.MiniMax.dir())
+
+  def test_key(self):
+    old_key = os.environ.pop('MINIMAX_API_KEY', None)
+    try:
+      with self.assertRaisesRegex(ValueError, 'Please specify `api_key`'):
+        _ = minimax.MiniMaxM27().headers
+      self.assertEqual(
+          minimax.MiniMaxM27(api_key='test_key').headers,
+          {
+              'Content-Type': 'application/json',
+              'Authorization': 'Bearer test_key',
+          }
+      )
+    finally:
+      if old_key is not None:
+        os.environ['MINIMAX_API_KEY'] = old_key
+
+  def test_key_from_env(self):
+    os.environ['MINIMAX_API_KEY'] = 'env_key'
+    try:
+      self.assertEqual(
+          minimax.MiniMaxM27().headers,
+          {
+              'Content-Type': 'application/json',
+              'Authorization': 'Bearer env_key',
+          }
+      )
+    finally:
+      del os.environ['MINIMAX_API_KEY']
+
+  def test_model_id(self):
+    self.assertEqual(
+        minimax.MiniMaxM27(api_key='test_key').model_id,
+        'MiniMax-M2.7',
+    )
+
+  def test_resource_id(self):
+    self.assertEqual(
+        minimax.MiniMaxM27(api_key='test_key').resource_id,
+        'minimax://MiniMax-M2.7',
+    )
+
+  def test_model_info(self):
+    lm = minimax.MiniMaxM27(api_key='test_key')
+    self.assertEqual(lm.model_info.provider, 'MiniMax')
+    self.assertEqual(lm.model_info.model_id, 'MiniMax-M2.7')
+    self.assertTrue(lm.model_info.in_service)
+
+  def test_request(self):
+    request = minimax.MiniMaxM27(api_key='test_key').request(
+        lf.UserMessage('hi'), lf.LMSamplingOptions(temperature=0.5),
+    )
+    self.assertEqual(request['model'], 'MiniMax-M2.7')
+    self.assertEqual(request['temperature'], 0.5)
+
+  def test_temperature_clamping(self):
+    request = minimax.MiniMaxM27(api_key='test_key').request(
+        lf.UserMessage('hi'), lf.LMSamplingOptions(temperature=0.0),
+    )
+    self.assertEqual(request['temperature'], 0.01)
+
+  def test_highspeed_model(self):
+    self.assertEqual(
+        minimax.MiniMaxM27Highspeed(api_key='test_key').model_id,
+        'MiniMax-M2.7-highspeed',
+    )
+
+  def test_lm_get(self):
+    self.assertIsInstance(
+        lf.LanguageModel.get('MiniMax-M2.7'), minimax.MiniMax
+    )
+    self.assertIsInstance(
+        lf.LanguageModel.get('MiniMax-M2.7-highspeed'), minimax.MiniMax
+    )
+
+
+if __name__ == '__main__':
+  unittest.main()