feat(llm): add MiniMax provider

.env.example

@@ -37,6 +37,16 @@ ATLAS_API_KEY=
 # ATLAS_BASE_URL=https://api.atlascloud.ai/v1
 # ATLAS_MODEL=anthropic/claude-sonnet-4.6
 
+# --- Optional: MiniMax (OpenAI- and Anthropic-compatible) -------------------
+# OpenAI SDK: global https://api.minimax.io/v1
+# OpenAI SDK: China  https://api.minimaxi.com/v1
+# Anthropic SDK: global https://api.minimax.io/anthropic
+# Anthropic SDK: China  https://api.minimaxi.com/anthropic
+# Select with LLM_PROVIDER=minimax or call get_provider("minimax").
+MINIMAX_API_KEY=
+# MINIMAX_MODEL=MiniMax-M3  # Also supports MiniMax-M2.7
+# MINIMAX_BASE_URL=https://api.minimax.io/v1
+
 # ─── ECC agent data (multi-harness isolation) ───────────────────────────────
 # Memory hooks (sessions, learned skills, aliases, metrics). Default: ~/.claude
 # Use a separate root when running ECC in Cursor alongside Claude Code:

src/llm/cli/selector.py

@@ -98,6 +98,7 @@ def interactive_select(
     if providers is None:
         providers = [
             ("claude", "Anthropic Claude ( Sonnet, Opus, Haiku)"),
+            ("minimax", "MiniMax (M3, M2.7)"),
             ("openai", "OpenAI GPT (4o, 4o-mini, 3.5-turbo)"),
             ("ollama", "Local Ollama models"),
         ]
@@ -109,6 +110,10 @@ def interactive_select(
                 ("claude-sonnet-4-6", "Claude Sonnet 4.6 - Balanced"),
                 ("claude-haiku-4-5", "Claude Haiku 4.5 - Fast"),
             ],
+            "minimax": [
+                ("MiniMax-M3", "MiniMax M3 - Long-context multimodal model"),
+                ("MiniMax-M2.7", "MiniMax M2.7 - Long-context text model"),
+            ],
             "openai": [
                 ("gpt-4o", "GPT-4o - Most capable"),
                 ("gpt-4o-mini", "GPT-4o-mini - Fast & affordable"),

src/llm/core/types.py

@@ -21,15 +21,17 @@ class ProviderType(str, Enum):
     ASTRAFLOW = "astraflow"
     ASTRAFLOW_CN = "astraflow_cn"
     ATLAS = "atlas"
+    MINIMAX = "minimax"
 
 
 @dataclass(frozen=True)
 class Message:
     role: Role
-    content: str
+    content: str | list[dict[str, Any]]
     name: str | None = None
     tool_call_id: str | None = None
     tool_calls: list[ToolCall] | None = None
+    metadata: dict[str, Any] = field(default_factory=dict)
 
     def to_dict(self) -> dict[str, Any]:
         result: dict[str, Any] = {"role": self.role.value, "content": self.content}

src/llm/prompt/builder.py

@@ -5,10 +5,7 @@
 from dataclasses import dataclass
 from typing import Any
 
-from llm.core.types import LLMInput, Message, Role, ToolDefinition
-from llm.providers.claude import ClaudeProvider
-from llm.providers.openai import OpenAIProvider
-from llm.providers.ollama import OllamaProvider
+from llm.core.types import Message, Role, ToolDefinition
 
 
 @dataclass
@@ -36,13 +33,16 @@ def __init__(
             raise ValueError("Pass either config or PromptBuilder keyword options, not both")
 
         if config is None:
-            overrides = {
-                "system_template": system_template,
-                "user_template": user_template,
-                "include_tools_in_system": include_tools_in_system,
-                "tool_format": tool_format,
-            }
-            config = PromptConfig(**{key: value for key, value in overrides.items() if value is not None})
+            config = PromptConfig(
+                system_template=system_template,
+                user_template=user_template,
+                include_tools_in_system=(
+                    include_tools_in_system
+                    if include_tools_in_system is not None
+                    else True
+                ),
+                tool_format=tool_format if tool_format is not None else "native",
+            )
 
         self.config = config
 
@@ -61,7 +61,10 @@ def build(self, messages: list[Message], tools: list[ToolDefinition] | None = No
             system_parts.append(f"\n\n## Available Tools\n{tools_desc}")
 
         if messages[0].role == Role.SYSTEM:
-            system_parts.insert(0, messages[0].content)
+            system_content = messages[0].content
+            if not isinstance(system_content, str):
+                raise TypeError("PromptBuilder system messages must contain text")
+            system_parts.insert(0, system_content)
             result.insert(0, Message(role=Role.SYSTEM, content="\n\n".join(system_parts)))
             result.extend(messages[1:])
         else:

src/llm/providers/__init__.py

@@ -3,6 +3,7 @@
 from llm.providers.astraflow import AstraflowCNProvider, AstraflowProvider
 from llm.providers.atlas import AtlasProvider
 from llm.providers.claude import ClaudeProvider
+from llm.providers.minimax import MiniMaxProvider
 from llm.providers.openai import OpenAIProvider
 from llm.providers.ollama import OllamaProvider
 from llm.providers.resolver import get_provider, register_provider
@@ -12,6 +13,7 @@
     "AstraflowProvider",
     "AtlasProvider",
     "ClaudeProvider",
+    "MiniMaxProvider",
     "OpenAIProvider",
     "OllamaProvider",
     "get_provider",

src/llm/providers/claude.py

@@ -58,7 +58,13 @@ def __init__(self, api_key: str | None = None, base_url: str | None = None) -> N
     def generate(self, input: LLMInput) -> LLMOutput:
         try:
             model = input.model or _DEFAULT_MODEL
-            system_parts = [msg.content for msg in input.messages if msg.role == Role.SYSTEM]
+            system_parts: list[str] = []
+            for message in input.messages:
+                if message.role != Role.SYSTEM:
+                    continue
+                if not isinstance(message.content, str):
+                    raise TypeError("System messages must contain text")
+                system_parts.append(message.content)
             api_messages = [
                 msg.to_dict() for msg in input.messages if msg.role not in (Role.SYSTEM,)
             ]