Switch LLM provider to z.ai GLM

README.md

@@ -1,6 +1,6 @@
 # LLM Trade Bot
 
-Автономный DeepSeek-агент для выделенного **Bybit Demo Trading** либо **Binance Testnet** аккаунта. Один процесс работает только с одной биржей, выбранной в `config.toml`; одновременная торговля на обеих не поддерживается. Бот анализирует Spot и `linear` USDT-рынки два раза в день, умеет искать рыночный контекст в интернете и управляется через Telegram.
+Автономный GLM-агент через **z.ai** для выделенного **Bybit Demo Trading** либо **Binance Testnet** аккаунта. Один процесс работает только с одной биржей, выбранной в `config.toml`; одновременная торговля на обеих не поддерживается. Бот анализирует Spot и `linear` USDT-рынки два раза в день, умеет искать рыночный контекст в интернете и управляется через Telegram.
 
 `linear` — публичное имя USDⓈ-M Futures: оно намеренно сохранено в инструментах агента и промптах. Production endpoints не поддерживаются. Для Binance адреса Spot Testnet и USDⓈ-M Futures Demo Testnet зафиксированы в коде и не настраиваются через TOML.
 
@@ -24,10 +24,12 @@
 2. Для Bybit создайте ключ в основном аккаунте после переключения в **Demo Trading**, а не на testnet. Настройте Unified account в `ISOLATED_MARGIN` и one-way position mode.
 3. Для Binance создайте два отдельных набора ключей: Spot Testnet на [testnet.binance.vision](https://testnet.binance.vision/) и USDⓈ-M Futures Testnet/Demo. У ключей должны быть права чтения и торговли. Futures-аккаунт должен использовать one-way mode; бот проверит это при запуске и выставит isolated margin/плечо перед входом.
 4. Создайте Telegram-бота и получите `api_id`/`api_hash` приложения Telegram.
-5. Получите DeepSeek API key и Brave Search API key.
+5. Получите z.ai API key и Brave Search API key.
 6. Скопируйте `config.example.toml` в `config.toml`, а `prompts.example.md` в `prompts.md`.
 7. Замените секреты только выбранного провайдера. Локальные конфиги исключены из Git.
 
+Для GLM используйте раздел `[zai]`: базовый URL `https://api.z.ai/api/paas/v4` и модель `glm-4.6` уже заданы в примере. Ключ создаётся в z.ai; раздел `[deepseek]` больше не поддерживается.
+
 ```powershell
 Copy-Item config.example.toml config.toml
 Copy-Item prompts.example.md prompts.md

config.example.toml

@@ -1,7 +1,7 @@
-[deepseek]
+[zai]
 api_key = "replace-me"
-base_url = "https://api.deepseek.com"
+base_url = "https://api.z.ai/api/paas/v4"
-model = "deepseek-flash"
+model = "glm-4.6"
 thinking = true
 max_tool_rounds = 256
 

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "llm-trade-bot"
 version = "0.1.0"
-description = "Autonomous DeepSeek trading agent for Bybit Demo"
+description = "Autonomous GLM trading agent for Bybit Demo"
 readme = "README.md"
 authors = [
     { name = "Azimkin", email = "112006710+Azimkin@users.noreply.github.com" }

src/llm_trade_bot/agent.py

@@ -15,7 +15,6 @@ from micro_agents import (
     Injected,
     LlmMessage,
     MessageAdded,
-    OpenAIBasedLLM,
     SessionContext,
     Tool,
     ToolCallFailed,
@@ -31,6 +30,7 @@ from llm_trade_bot.models import Category, OpenTradeRequest, OrderType, Side
 from llm_trade_bot.prompts import PromptStore
 from llm_trade_bot.risk import RiskEngine
 from llm_trade_bot.web_tools import SafeWebTools
+from llm_trade_bot.zai_llm import ZaiLLM
 
 
 class AgentTools:
@@ -353,18 +353,18 @@ class AgentRunner:
         self.exchange = exchange
         self.bybit = exchange
         self.prompts = PromptStore(config.prompts.file_path)
-        llm = OpenAIBasedLLM(
+        llm = ZaiLLM(
-            api_url=config.deepseek.base_url,
+            api_url=config.zai.base_url,
-            api_key=config.deepseek.api_key,
+            api_key=config.zai.api_key,
-            model=config.deepseek.model,
+            model=config.zai.model,
         )
         self.agent_tools = AgentTools(config, database, exchange)
         self.aggregator = Aggregator(
             llm,
             ToolManager(self.agent_tools.tools()),
             master_prompt=_master_prompt,
-            max_tool_rounds=config.deepseek.max_tool_rounds,
+            max_tool_rounds=config.zai.max_tool_rounds,
-            thinking=config.deepseek.thinking,
+            thinking=config.zai.thinking,
         )
 
     async def run(self, trigger: str) -> str:

src/llm_trade_bot/cli.py

@@ -11,7 +11,7 @@ from llm_trade_bot.service import TradeBotService
 
 
 def main() -> None:
-    parser = argparse.ArgumentParser(description="Autonomous DeepSeek trader for Bybit Demo or Binance Testnet")
+    parser = argparse.ArgumentParser(description="Autonomous GLM trader for Bybit Demo or Binance Testnet")
     parser.add_argument("--config", default="config.toml", help="Path to TOML configuration")
     parser.add_argument("--log-level", default="INFO", choices=("DEBUG", "INFO", "WARNING", "ERROR"))
     args = parser.parse_args()

src/llm_trade_bot/config.py

@@ -13,10 +13,10 @@ class ConfigError(ValueError):
 
 
 @dataclass(frozen=True, slots=True)
-class DeepSeekConfig:
+class ZaiConfig:
     api_key: str
-    base_url: str = "https://api.deepseek.com"
+    base_url: str = "https://api.z.ai/api/paas/v4"
-    model: str = "deepseek-flash"
+    model: str = "glm-4.6"
     thinking: bool = True
     max_tool_rounds: int = 16
 
@@ -99,7 +99,7 @@ class StorageConfig:
 
 @dataclass(frozen=True, slots=True)
 class AppConfig:
-    deepseek: DeepSeekConfig
+    zai: ZaiConfig
     bybit: BybitConfig | None
     telegram: TelegramConfig
     brave: BraveConfig
@@ -121,7 +121,7 @@ def load_config(path: str | Path) -> AppConfig:
     except tomllib.TOMLDecodeError as error:
         raise ConfigError(f"Invalid TOML in {config_path}: {error}") from error
 
-    deepseek_raw = _section(raw, "deepseek")
+    zai_raw = _section(raw, "zai")
     exchange_raw = raw.get("exchange")
     if exchange_raw is None:
         if not isinstance(raw.get("bybit"), dict):
@@ -142,15 +142,15 @@ def load_config(path: str | Path) -> AppConfig:
     risk_raw = _section(raw, "risk")
     storage_raw = _section(raw, "storage")
 
-    deepseek = DeepSeekConfig(
+    zai = ZaiConfig(
-        api_key=_secret(deepseek_raw, "api_key"),
+        api_key=_secret(zai_raw, "api_key"),
-        base_url=_string(deepseek_raw, "base_url", "https://api.deepseek.com").rstrip("/"),
+        base_url=_string(zai_raw, "base_url", "https://api.z.ai/api/paas/v4").rstrip("/"),
-        model=_string(deepseek_raw, "model", "deepseek-flash"),
+        model=_string(zai_raw, "model", "glm-4.6"),
-        thinking=_boolean(deepseek_raw, "thinking", True),
+        thinking=_boolean(zai_raw, "thinking", True),
-        max_tool_rounds=_integer(deepseek_raw, "max_tool_rounds", 16, minimum=1, maximum=256),
+        max_tool_rounds=_integer(zai_raw, "max_tool_rounds", 16, minimum=1, maximum=256),
     )
-    if deepseek.base_url != "https://api.deepseek.com":
+    if zai.base_url != "https://api.z.ai/api/paas/v4":
-        raise ConfigError("deepseek.base_url must be https://api.deepseek.com")
+        raise ConfigError("zai.base_url must be https://api.z.ai/api/paas/v4")
 
     bybit: BybitConfig | None = None
     binance: BinanceConfig | None = None
@@ -228,7 +228,7 @@ def load_config(path: str | Path) -> AppConfig:
         max_notes=_integer(storage_raw, "max_notes", 50, minimum=1, maximum=1000),
         max_notes_chars=_integer(storage_raw, "max_notes_chars", 20_000, minimum=100, maximum=200_000),
     )
-    return AppConfig(deepseek, bybit, telegram, brave, prompts, schedule, scanner, risk, storage, exchange, binance)
+    return AppConfig(zai, bybit, telegram, brave, prompts, schedule, scanner, risk, storage, exchange, binance)
 
 
 def _reject_binance_endpoint_overrides(section: dict[str, Any]) -> None:

src/llm_trade_bot/zai_llm.py

@@ -0,0 +1,148 @@
+"""z.ai adapter for the provider-neutral micro-agents interface."""
+
+from __future__ import annotations
+
+from typing import Any
+
+import orjson
+from micro_agents import OpenAIBasedLLM
+from micro_agents.conversation import (
+    Conversation,
+    LlmMessage,
+    Message,
+    ReasoningMessage,
+    ToolCallMessage,
+    ToolCallResponseMessage,
+    UserMessage,
+)
+import micro_agents.tools as tools_package
+
+
+class ZaiLLM(OpenAIBasedLLM):
+    """Use z.ai's OpenAI-compatible Chat Completions endpoint.
+
+    ``OpenAIBasedLLM`` in micro-agents targets OpenAI's Responses API, which
+    z.ai does not expose.  This adapter retains its tool-schema inference while
+    translating the agent conversation to the Chat Completions format.
+    """
+
+    async def ask(
+        self,
+        conversation: Conversation,
+        tools: list[tools_package.Tool],
+        *,
+        instructions: str | None = None,
+        thinking: bool | None = None,
+    ) -> list[Message]:
+        extra_body: dict[str, object] | None = None
+        if thinking is not None:
+            extra_body = {"thinking": {"type": "enabled" if thinking else "disabled"}}
+        response = await self._ai.chat.completions.create(
+            model=self._model,
+            messages=self._serialize_conversation(conversation, instructions),
+            tools=self._serialize_chat_tools(tools),
+            stream=False,
+            extra_body=extra_body,
+        )
+        if not response.choices:
+            raise ValueError("z.ai returned no choices")
+        message = response.choices[0].message
+        tool_calls = message.tool_calls or []
+        out: list[Message] = []
+
+        # GLM returns this field when thinking is enabled.  Keeping it in the
+        # conversation lets the next call include it with the tool invocation.
+        reasoning = getattr(message, "reasoning_content", None)
+        if isinstance(reasoning, str) and reasoning:
+            out.append(ReasoningMessage(reasoning))
+
+        # A Chat Completions response with tool calls is represented by one
+        # assistant message.  The content is normally empty in that case; do
+        # not create a second assistant message before the tool-call message.
+        if not tool_calls and isinstance(message.content, str) and message.content:
+            out.append(LlmMessage(message.content))
+        for call in tool_calls:
+            function = call.function
+            if not isinstance(call.id, str) or not isinstance(function.name, str):
+                raise ValueError("z.ai returned an invalid function call")
+            try:
+                arguments = orjson.loads(function.arguments)
+            except orjson.JSONDecodeError as error:
+                raise ValueError("z.ai returned invalid JSON function arguments") from error
+            if not isinstance(arguments, dict):
+                raise ValueError("z.ai function-call arguments must be an object")
+            out.append(ToolCallMessage(call.id, function.name, arguments))
+        return out
+
+    def _serialize_chat_tools(self, tools: list[tools_package.Tool]) -> list[dict[str, object]]:
+        result: list[dict[str, object]] = []
+        for tool in tools:
+            schema = self._serialize_tool(tool)
+            result.append(
+                {
+                    "type": "function",
+                    "function": {
+                        "name": schema["name"],
+                        "description": schema["description"],
+                        "parameters": schema["parameters"],
+                    },
+                }
+            )
+        return result
+
+    @staticmethod
+    def _serialize_conversation(conversation: Conversation, instructions: str | None) -> list[dict[str, Any]]:
+        messages: list[dict[str, Any]] = []
+        if instructions:
+            messages.append({"role": "system", "content": instructions})
+
+        items = conversation.messages
+        index = 0
+        pending_reasoning: str | None = None
+        while index < len(items):
+            item = items[index]
+            if isinstance(item, ReasoningMessage):
+                pending_reasoning = item.text()
+                index += 1
+                continue
+            if isinstance(item, UserMessage):
+                messages.append({"role": "user", "content": item.message()})
+            elif isinstance(item, LlmMessage):
+                assistant: dict[str, Any] = {"role": "assistant", "content": item.message()}
+                if pending_reasoning:
+                    assistant["reasoning_content"] = pending_reasoning
+                    pending_reasoning = None
+                messages.append(assistant)
+            elif isinstance(item, ToolCallMessage):
+                calls: list[dict[str, object]] = []
+                while index < len(items) and isinstance(items[index], ToolCallMessage):
+                    call = items[index]
+                    calls.append(
+                        {
+                            "id": call.unique_id(),
+                            "type": "function",
+                            "function": {
+                                "name": call.tool_name(),
+                                "arguments": orjson.dumps(call.arguments()).decode("utf-8"),
+                            },
+                        }
+                    )
+                    index += 1
+                assistant = {"role": "assistant", "content": None, "tool_calls": calls}
+                if pending_reasoning:
+                    assistant["reasoning_content"] = pending_reasoning
+                    pending_reasoning = None
+                messages.append(assistant)
+                continue
+            elif isinstance(item, ToolCallResponseMessage):
+                messages.append(
+                    {
+                        "role": "tool",
+                        "tool_call_id": item.unique_id(),
+                        "content": item.text(),
+                    }
+                )
+            else:
+                raise ValueError(f"Unknown type of message {type(item).__name__}")
+            index += 1
+        return messages

tests/test_agent_tools.py

@@ -10,7 +10,7 @@ from llm_trade_bot.config import (
     AppConfig,
     BraveConfig,
     BybitConfig,
-    DeepSeekConfig,
+    ZaiConfig,
     PromptsConfig,
     RiskConfig,
     ScannerConfig,
@@ -48,7 +48,7 @@ class FakeBybit:
 
 def make_config(database_path: Path) -> AppConfig:
     return AppConfig(
-        DeepSeekConfig("key"),
+        ZaiConfig("key"),
         BybitConfig("key", "secret"),
         TelegramConfig(1, "hash", "token", frozenset({1})),
         BraveConfig("key"),

tests/test_config.py

@@ -5,8 +5,8 @@ import pytest
 from llm_trade_bot.config import ConfigError, load_config
 
 CONFIG = """
-[deepseek]
+[zai]
-api_key = "deepseek-key"
+api_key = "zai-key"
 [bybit]
 api_key = "bybit-key"
 api_secret = "bybit-secret"
@@ -51,7 +51,7 @@ def write_config(tmp_path: Path, content: str = CONFIG) -> Path:
 def test_loads_and_resolves_database_path(tmp_path: Path):
     path = write_config(tmp_path)
     config = load_config(path)
-    assert config.deepseek.model == "deepseek-flash"
+    assert config.zai.model == "glm-4.6"
     assert config.telegram.admin_user_ids == frozenset({42})
     assert config.storage.database_path == tmp_path / "state.db"
     assert config.prompts.file_path == tmp_path / "prompts.md"
@@ -63,8 +63,8 @@ def test_rejects_risk_above_hard_limit(tmp_path: Path):
         load_config(path)
 
 
-def test_rejects_non_official_deepseek_endpoint(tmp_path: Path):
+def test_rejects_non_official_zai_endpoint(tmp_path: Path):
-    changed = CONFIG.replace("api_key = \"deepseek-key\"", 'api_key = "deepseek-key"\nbase_url = "https://example.com"', 1)
+    changed = CONFIG.replace("api_key = \"zai-key\"", 'api_key = "zai-key"\nbase_url = "https://example.com"', 1)
     path = write_config(tmp_path, changed)
     with pytest.raises(ConfigError, match="base_url"):
         load_config(path)

tests/test_zai_llm.py

@@ -0,0 +1,34 @@
+from micro_agents.conversation import Conversation, ReasoningMessage, ToolCallMessage, ToolCallResponseMessage, UserMessage
+
+from llm_trade_bot.zai_llm import ZaiLLM
+
+
+def test_serializes_glm_tool_round_in_chat_completions_format():
+    conversation = Conversation(
+        [
+            UserMessage("analyse BTC"),
+            ReasoningMessage("I need market data."),
+            ToolCallMessage("call-1", "get_market_data", {"symbol": "BTCUSDT"}),
+            ToolCallResponseMessage("call-1", '{"last_price":"100000"}'),
+        ]
+    )
+
+    messages = ZaiLLM._serialize_conversation(conversation, "Follow the risk limits.")
+
+    assert messages == [
+        {"role": "system", "content": "Follow the risk limits."},
+        {"role": "user", "content": "analyse BTC"},
+        {
+            "role": "assistant",
+            "content": None,
+            "reasoning_content": "I need market data.",
+            "tool_calls": [
+                {
+                    "id": "call-1",
+                    "type": "function",
+                    "function": {"name": "get_market_data", "arguments": '{"symbol":"BTCUSDT"}'},
+                }
+            ],
+        },
+        {"role": "tool", "tool_call_id": "call-1", "content": '{"last_price":"100000"}'},
+    ]