feat: add MiniMax LLM provider support (M2.7 default)

README.md

@@ -45,6 +45,7 @@ Choose **one** of:
     - Add your OPENAI_API_KEY in .env
 - Custom model:
     - See [docs](https://gpt-engineer.readthedocs.io/en/latest/open_models.html), supports local model, azure, etc.
+- For MiniMax models: `export MINIMAX_API_KEY=[your api key]`
 
 Check the [Windows README](./WINDOWS_README.md) for Windows usage.
 
@@ -102,7 +103,26 @@ E.g. `gpte projects/example-vision gpt-4-vision-preview --prompt_file prompt/tex
 
 ### Open source, local and alternative models
 
-By default, gpt-engineer supports OpenAI Models via the OpenAI API or Azure OpenAI API, as well as Anthropic models.
+By default, gpt-engineer supports OpenAI Models via the OpenAI API or Azure OpenAI API, as well as Anthropic models and [MiniMax](https://platform.minimax.io) models.
+
+#### MiniMax Models
+
+To use MiniMax models, set the `MINIMAX_API_KEY` environment variable and pass a MiniMax model name:
+
+```bash
+export MINIMAX_API_KEY=your-minimax-api-key
+gpte my-project -m MiniMax-M2.7
+```
+
+Supported MiniMax models:
+| Model | Description |
+|-------|-------------|
+| `MiniMax-M2.7` | Latest flagship model with enhanced reasoning and coding. |
+| `MiniMax-M2.7-highspeed` | High-speed version of M2.7 for low-latency scenarios. |
+| `MiniMax-M2.5` | Peak Performance. Ultimate Value. 204K context window. |
+| `MiniMax-M2.5-highspeed` | Same performance, faster and more agile. 204K context window. |
+
+You can also set a custom API base URL via `MINIMAX_API_BASE` (defaults to `https://api.minimax.io/v1`). For users in China, set it to `https://api.minimaxi.com/v1`.
 
 With a little extra setup, you can also run with open source models like WizardCoder. See the [documentation](https://gpt-engineer.readthedocs.io/en/latest/open_models.html) for example instructions.
 

gpt_engineer/applications/cli/main.py

@@ -89,6 +89,11 @@ def load_env_if_needed():
     if os.getenv("ANTHROPIC_API_KEY") is None:
         load_dotenv(dotenv_path=os.path.join(os.getcwd(), ".env"))
 
+    if os.getenv("MINIMAX_API_KEY") is None:
+        load_dotenv()
+    if os.getenv("MINIMAX_API_KEY") is None:
+        load_dotenv(dotenv_path=os.path.join(os.getcwd(), ".env"))
+
 
 def concatenate_paths(base_path, sub_path):
     # Compute the relative path from base_path to sub_path
@@ -551,6 +556,8 @@ def main(
 
     if ai.token_usage_log.is_openai_model():
         print("Total api cost: $ ", ai.token_usage_log.usage_cost())
+    elif ai.token_usage_log.is_minimax_model():
+        print("Total tokens used: ", ai.token_usage_log.total_tokens())
     elif os.getenv("LOCAL_MODEL"):
         print("Total api cost: $ 0.0 since we are using local LLM.")
     else:

gpt_engineer/core/ai.py

@@ -362,6 +362,26 @@ def _create_chat_model(self) -> BaseChatModel:
                 streaming=self.streaming,
                 max_tokens_to_sample=4096,
             )
+        elif "minimax" in self.model_name.lower():
+            # MiniMax requires temperature in (0.0, 1.0], zero is not accepted
+            minimax_temperature = max(self.temperature, 0.01)
+            minimax_api_key = os.getenv("MINIMAX_API_KEY")
+            if not minimax_api_key:
+                raise ValueError(
+                    "MINIMAX_API_KEY environment variable is required for MiniMax models. "
+                    "Get your API key at https://platform.minimax.io"
+                )
+            minimax_base_url = os.getenv(
+                "MINIMAX_API_BASE", "https://api.minimax.io/v1"
+            )
+            return ChatOpenAI(
+                model=self.model_name,
+                temperature=minimax_temperature,
+                streaming=self.streaming,
+                callbacks=[StreamingStdOutCallbackHandler()],
+                openai_api_key=minimax_api_key,
+                openai_api_base=minimax_base_url,
+            )
         elif self.vision:
             return ChatOpenAI(
                 model=self.model_name,

gpt_engineer/core/token_usage.py

@@ -259,6 +259,17 @@ def is_openai_model(self) -> bool:
         """
         return "gpt" in self.model_name.lower()
 
+    def is_minimax_model(self) -> bool:
+        """
+        Check if the model is a MiniMax model.
+
+        Returns
+        -------
+        bool
+            True if the model is a MiniMax model, False otherwise.
+        """
+        return "minimax" in self.model_name.lower()
+
     def total_tokens(self) -> int:
         """
         Return the total number of tokens used in the conversation.

tests/core/test_ai.py

@@ -1,3 +1,7 @@
+import os
+
+from unittest.mock import patch
+
 from langchain.chat_models.base import BaseChatModel
 from langchain_community.chat_models.fake import FakeListChatModel
 
@@ -51,3 +55,53 @@ def test_token_logging(monkeypatch):
     # assert
     assert usageCostAfterStart > 0
     assert usageCostAfterNext > usageCostAfterStart
+
+
+def test_minimax_model_creation(monkeypatch):
+    monkeypatch.setattr(AI, "_create_chat_model", mock_create_chat_model)
+
+    ai = AI("MiniMax-M2.7")
+
+    # act
+    response_messages = ai.start("system prompt", "user prompt", step_name="step name")
+
+    # assert
+    assert response_messages[-1].content == "response1"
+    assert ai.token_usage_log.is_minimax_model()
+    assert not ai.token_usage_log.is_openai_model()
+
+
+def test_minimax_temperature_constraint():
+    """Test that MiniMax temperature is clamped to avoid zero."""
+    ai_instance = AI.__new__(AI)
+    ai_instance.temperature = 0.0
+    ai_instance.model_name = "MiniMax-M2.7"
+    ai_instance.azure_endpoint = None
+    ai_instance.streaming = True
+    ai_instance.vision = False
+
+    with patch.dict(os.environ, {"MINIMAX_API_KEY": "test-key"}):
+        with patch("gpt_engineer.core.ai.ChatOpenAI") as mock_chat:
+            mock_chat.return_value = FakeListChatModel(responses=["response"])
+            ai_instance._create_chat_model()
+            # Verify temperature was clamped above 0
+            call_kwargs = mock_chat.call_args[1]
+            assert call_kwargs["temperature"] >= 0.01
+
+
+def test_minimax_missing_api_key():
+    """Test that MiniMax raises error when API key is missing."""
+    ai_instance = AI.__new__(AI)
+    ai_instance.temperature = 0.1
+    ai_instance.model_name = "MiniMax-M2.7"
+    ai_instance.azure_endpoint = None
+    ai_instance.streaming = True
+    ai_instance.vision = False
+
+    with patch.dict(os.environ, {}, clear=True):
+        os.environ.pop("MINIMAX_API_KEY", None)
+        try:
+            ai_instance._create_chat_model()
+            assert False, "Should have raised ValueError"
+        except ValueError as e:
+            assert "MINIMAX_API_KEY" in str(e)