feat: add MiniMax as first-class LLM provider (M2.7/M2.5, 204K context)

.env.example

@@ -2,6 +2,10 @@
 MODEL_BASE_URL=
 MODEL_KEY=
 
+#MINIMAX (https://platform.minimax.io)
+#Models: MiniMax-M2.7, MiniMax-M2.7-highspeed, MiniMax-M2.5, MiniMax-M2.5-highspeed (204K context)
+MINIMAX_API_KEY=
+
 #CUSTOM MODEL
 QWEN_BASE_URL=
 QWEN_KEY=

README.md

@@ -34,6 +34,37 @@ MODEL_BASE_URL=<your model base url>
 MODEL_KEY=<your model key>
 ```
 
+## Supported LLM Providers
+
+Aser works with any OpenAI-compatible API. Use the built-in `providers` module for quick setup:
+
+| Provider | Models | Context |
+|----------|--------|---------|
+| OpenAI | gpt-4o, gpt-4.1-mini, … | up to 128K |
+| [MiniMax](https://platform.minimax.io) | MiniMax-M2.7, MiniMax-M2.7-highspeed, MiniMax-M2.5, MiniMax-M2.5-highspeed | **204K** |
+| Any OpenAI-compat API | set `MODEL_BASE_URL` + `MODEL_KEY` | depends on model |
+
+### Using MiniMax
+
+```bash
+export MINIMAX_API_KEY=your_api_key_here
+```
+
+```python
+from aser.agent import Agent
+from aser.providers import MINIMAX, MINIMAX_MODELS
+
+agent = Agent(
+    name="minimax_agent",
+    model=MINIMAX_MODELS["standard"],   # "MiniMax-M2.7"
+    model_config=MINIMAX,
+)
+response = agent.chat("What is MiniMax AI?")
+print(response)
+```
+
+See the full example at [examples/agent_minimax.py](./examples/agent_minimax.py).
+
 ## Usage
 
 ```python
@@ -66,6 +97,7 @@ If you clone the project source code, before running the examples, please run `p
 
 - [Aser Agent](./examples/agent.py): Your First AI Agent
 - [Model Config](./examples/agent_model.py): Customize the LLM configuration
+- [MiniMax](./examples/agent_minimax.py): Use MiniMax M2.7 / M2.5 models (204K context)
 - [Character](./examples/agent_character.py): Build an agent with character
 - [Memory](./examples/agent_memory.py): Build an agent with memory storage
 - [RAG](./examples/agent_knowledge.py): Build an agent with knowledge retrieval

README_CN.md

@@ -34,6 +34,37 @@ MODEL_BASE_URL=<您的模型基础URL>
 MODEL_KEY=<您的模型密钥>
 ```
 
+## 支持的 LLM 提供商
+
+Aser 兼容任何 OpenAI 兼容 API。通过内置 `providers` 模块可快速配置：
+
+| 提供商 | 模型 | 上下文长度 |
+|--------|------|-----------|
+| OpenAI | gpt-4o, gpt-4.1-mini, … | 最高 128K |
+| [MiniMax](https://platform.minimax.io) | MiniMax-M2.7, MiniMax-M2.7-highspeed, MiniMax-M2.5, MiniMax-M2.5-highspeed | **204K** |
+| 任意 OpenAI 兼容 API | 设置 `MODEL_BASE_URL` + `MODEL_KEY` | 取决于模型 |
+
+### 使用 MiniMax
+
+```bash
+export MINIMAX_API_KEY=your_api_key_here
+```
+
+```python
+from aser.agent import Agent
+from aser.providers import MINIMAX, MINIMAX_MODELS
+
+agent = Agent(
+    name="minimax_agent",
+    model=MINIMAX_MODELS["standard"],   # "MiniMax-M2.7"
+    model_config=MINIMAX,
+)
+response = agent.chat("你好，请介绍一下 MiniMax。")
+print(response)
+```
+
+完整示例参见 [examples/agent_minimax.py](./examples/agent_minimax.py)。
+
 ## 使用
 
 ```python
@@ -66,6 +97,7 @@ aser = Agent(
 
 - [Aser代理](./examples/agent.py): 您的第一个AI代理
 - [模型配置](./examples/agent_model.py): 自定义LLM配置
+- [MiniMax](./examples/agent_minimax.py): 使用 MiniMax M2.7/M2.5 模型（204K 上下文）
 - [角色](./examples/agent_character.py): 构建具有角色的代理
 - [记忆](./examples/agent_memory.py): 构建具有记忆存储的代理
 - [RAG](./examples/agent_knowledge.py): 构建具有知识检索的代理

aser/__init__.py

@@ -16,6 +16,7 @@
 from .text2sql import Text2SQL
 from .evolution import SelfCodingTool
 from . import social,storage,utils
+from . import providers
 
 
 

aser/providers.py

@@ -0,0 +1,63 @@
+"""
+Provider presets for popular LLM APIs.
+
+Usage:
+    from aser.providers import MINIMAX, OPENAI
+    from aser.agent import Agent
+
+    agent = Agent(name="my_agent", model="MiniMax-M2.7", model_config=MINIMAX)
+"""
+
+import os
+
+
+def _make_config(base_url: str, api_key_env: str, api_key: str | None = None) -> dict:
+    return {
+        "base_url": base_url,
+        "api_key": api_key or os.getenv(api_key_env) or "MISSING_API_KEY",
+    }
+
+
+# ---------------------------------------------------------------------------
+# MiniMax — OpenAI-compatible endpoint
+# Models: MiniMax-M2.7, MiniMax-M2.7-highspeed, MiniMax-M2.5, MiniMax-M2.5-highspeed
+# Context: 204K tokens for all models
+# Docs: https://docs.minimax.io
+# Note: temperature must be in (0.0, 1.0]; do NOT use temperature=0
+# ---------------------------------------------------------------------------
+MINIMAX = _make_config(
+    base_url="https://api.minimax.io/v1",
+    api_key_env="MINIMAX_API_KEY",
+)
+
+# Recommended models
+MINIMAX_MODELS = {
+    "standard": "MiniMax-M2.7",
+    "fast": "MiniMax-M2.7-highspeed",
+    "standard_v25": "MiniMax-M2.5",
+    "fast_v25": "MiniMax-M2.5-highspeed",
+}
+
+# ---------------------------------------------------------------------------
+# OpenAI
+# ---------------------------------------------------------------------------
+OPENAI = _make_config(
+    base_url="https://api.openai.com/v1",
+    api_key_env="OPENAI_API_KEY",
+)
+
+# ---------------------------------------------------------------------------
+# Anthropic (OpenAI-compatible proxy via third-party gateways)
+# Use your gateway's base_url; Anthropic's native API is not OpenAI-compatible.
+# ---------------------------------------------------------------------------
+ANTHROPIC_COMPAT = _make_config(
+    base_url=os.getenv("ANTHROPIC_COMPAT_BASE_URL", "https://api.anthropic.com/v1"),
+    api_key_env="ANTHROPIC_API_KEY",
+)
+
+# ---------------------------------------------------------------------------
+# Generic helper: build a custom provider config on the fly
+# ---------------------------------------------------------------------------
+def custom_provider(base_url: str, api_key: str) -> dict:
+    """Return a model_config dict for any OpenAI-compatible provider."""
+    return {"base_url": base_url, "api_key": api_key}

examples/agent_minimax.py

@@ -0,0 +1,72 @@
+"""
+Example: Using Aser with MiniMax models.
+
+MiniMax offers OpenAI-compatible API access to its M2.7 and M2.5 model families,
+all featuring 204K-token context windows.
+
+Setup:
+  1. Get your API key from https://platform.minimax.io
+  2. Set the environment variable:
+       export MINIMAX_API_KEY=your_api_key_here
+  3. Run this example:
+       python examples/agent_minimax.py
+"""
+
+import os
+from dotenv import load_dotenv
+
+load_dotenv()
+
+from aser.agent import Agent
+from aser.providers import MINIMAX, MINIMAX_MODELS
+
+# ---------------------------------------------------------------------------
+# Basic usage: use the default MODEL_BASE_URL / MODEL_KEY env vars
+# ---------------------------------------------------------------------------
+# You can also point MODEL_BASE_URL=https://api.minimax.io/v1 and
+# MODEL_KEY=<your MINIMAX_API_KEY> in your .env file and use Agent without
+# explicitly passing model_config:
+#
+#   agent = Agent(name="aser agent", model="MiniMax-M2.7")
+
+# ---------------------------------------------------------------------------
+# Explicit usage: pass model_config=MINIMAX
+# ---------------------------------------------------------------------------
+print("=== MiniMax M2.7 (standard) ===")
+agent = Agent(
+    name="minimax_agent",
+    description="You are a helpful AI assistant powered by MiniMax.",
+    model=MINIMAX_MODELS["standard"],   # "MiniMax-M2.7"
+    model_config=MINIMAX,               # base_url + MINIMAX_API_KEY
+)
+response = agent.chat("What is MiniMax AI? Introduce yourself briefly.")
+print(response)
+
+# ---------------------------------------------------------------------------
+# High-speed variant — same capability, optimised for latency
+# ---------------------------------------------------------------------------
+print("\n=== MiniMax M2.7-highspeed (fast) ===")
+fast_agent = Agent(
+    name="minimax_fast_agent",
+    description="You are a concise assistant. Always answer in one sentence.",
+    model=MINIMAX_MODELS["fast"],       # "MiniMax-M2.7-highspeed"
+    model_config=MINIMAX,
+)
+response = fast_agent.chat("What is the capital of France?")
+print(response)
+
+# ---------------------------------------------------------------------------
+# Tip: you can also build the config inline without importing MINIMAX
+# ---------------------------------------------------------------------------
+print("\n=== Inline model_config ===")
+inline_agent = Agent(
+    name="inline_minimax_agent",
+    description="Helpful assistant",
+    model="MiniMax-M2.5",
+    model_config={
+        "base_url": "https://api.minimax.io/v1",
+        "api_key": os.getenv("MINIMAX_API_KEY"),
+    },
+)
+response = inline_agent.chat("Say hello in three different languages.")
+print(response)