add huoshan_ark

config/examples/huoshan_ark.yaml

@@ -0,0 +1,8 @@
+# Full Example: https://github.com/geekan/MetaGPT/blob/main/config/config2.example.yaml
+# Reflected Code: https://github.com/geekan/MetaGPT/blob/main/metagpt/config2.py
+# Config Docs: https://docs.deepwisdom.ai/main/en/guide/get_started/configuration.html
+llm:
+  api_type: "ark"  # or azure / ollama / groq etc.
+  model: ""  # your model endpoint like ep-xxx
+  base_url: "https://ark.cn-beijing.volces.com/api/v3"  # or forward url / other llm url
+  api_key: "" # your api-key like ey……

metagpt/configs/llm_config.py

@@ -33,6 +33,7 @@ class LLMType(Enum):
     YI = "yi"  # lingyiwanwu
     OPENROUTER = "openrouter"
     BEDROCK = "bedrock"
+    ARK = "ark"
 
     def __missing__(self, key):
         return self.OPENAI

metagpt/provider/__init__.py

@@ -18,6 +18,7 @@ from metagpt.provider.qianfan_api import QianFanLLM
 from metagpt.provider.dashscope_api import DashScopeLLM
 from metagpt.provider.anthropic_api import AnthropicLLM
 from metagpt.provider.bedrock_api import BedrockLLM
+from metagpt.provider.ark_api import ArkLLM
 
 __all__ = [
     "GeminiLLM",
@@ -32,4 +33,5 @@ __all__ = [
     "DashScopeLLM",
     "AnthropicLLM",
     "BedrockLLM",
+    "ArkLLM"
 ]

metagpt/provider/ark_api.py

@@ -0,0 +1,43 @@
+from openai import AsyncStream
+from openai.types import CompletionUsage
+from openai.types.chat import ChatCompletion, ChatCompletionChunk
+
+from metagpt.provider.openai_api import OpenAILLM
+from metagpt.configs.llm_config import LLMType
+from metagpt.provider.llm_provider_registry import register_provider
+from metagpt.const import USE_CONFIG_TIMEOUT
+from metagpt.logs import log_llm_stream
+
+@register_provider(LLMType.ARK)
+class ArkLLM(OpenAILLM):
+    """
+    用于火山方舟的API
+    见：https://www.volcengine.com/docs/82379/1263482
+    """
+    
+    async def _achat_completion_stream(self, messages: list[dict], timeout=USE_CONFIG_TIMEOUT) -> str:
+        response: AsyncStream[ChatCompletionChunk] = await self.aclient.chat.completions.create(
+            **self._cons_kwargs(messages, timeout=self.get_timeout(timeout)), 
+            stream=True,
+            extra_body={"stream_options": {"include_usage": True}}
+        )
+        usage = None
+        collected_messages = []
+        async for chunk in response:
+            chunk_message = chunk.choices[0].delta.content or "" if chunk.choices else ""  # extract the message
+            log_llm_stream(chunk_message)
+            collected_messages.append(chunk_message)
+            if chunk.usage:
+                # the usage of ark when streaming is in the last chunk while others are null
+                usage=CompletionUsage(**chunk.usage)
+
+        log_llm_stream("\n")
+        full_reply_content = "".join(collected_messages)
+        self._update_costs(usage,chunk.model)
+        return full_reply_content
+    
+    async def _achat_completion(self, messages: list[dict], timeout=USE_CONFIG_TIMEOUT) -> ChatCompletion:
+        kwargs = self._cons_kwargs(messages, timeout=self.get_timeout(timeout))
+        rsp: ChatCompletion = await self.aclient.chat.completions.create(**kwargs)
+        self._update_costs(rsp.usage,rsp.model)
+        return rsp

metagpt/utils/token_counter.py

@@ -68,6 +68,13 @@ TOKEN_COSTS = {
     "openai/gpt-4-turbo-preview": {"prompt": 0.01, "completion": 0.03},
     "deepseek-chat": {"prompt": 0.00014, "completion": 0.00028},
     "deepseek-coder": {"prompt": 0.00014, "completion": 0.00028},
+    # For ark model
+    "doubao-lite-4k-240515": {"prompt": 0.000042, "completion": 0.000084},
+    "doubao-lite-32k-240515": {"prompt": 0.000042, "completion": 0.000084},
+    "doubao-lite-128k-240515": {"prompt": 0.00011, "completion": 0.00013},
+    "doubao-pro-4k-240515":{"prompt":0.00011,"completion":0.00028},
+    "doubao-pro-32k-240515":{"prompt":0.00011,"completion":0.00028},
+    "doubao-pro-128k-240515":{"prompt":0.0007,"completion":0.0012},
 }
 
 

volcengine

@@ -0,0 +1 @@
+Subproject commit e62feb00e583a64243a58bf77de49cfca2c96210