Merge pull request #48 from InternLM/dev

refactor the config.yaml to make the model setting looks more logical

README.md

@@ -12,7 +12,8 @@ English | [简体中文](docs/README_zh-CN.md)
 
 
 ## Latest Progress 🎉
-
+- \[January 2024\] refactor the config-template.yaml to control the backend and the frontend settings at the same time, [click](https://github.com/InternLM/OpenAOE/blob/main/docs/tech-report/config-template.md) to find more introduction about the `config-template.yaml`
+- \[January 2024\] Add internlm2-chat-7b model
 - \[January 2024\] Released version v0.0.1, officially open source！
 ______________________________________________________________________
 

docs/todo/TODO.md

@@ -1,6 +1,8 @@
 ### TODO
 
 - [x] set the workers to 3
-- [ ] add Gemini model
-- [x] refacotor the config.yaml to make the model setting looks more logical
+- [x] add internlm2-chat-7b model as default
+- [ ] add Gemini model as default
+- [x] refactor the config.yaml to make the model setting looks more logical
 - [ ] dynamic add new model by editing external python files and the config.yaml 
+- [ ] build frontend project when OpenAOE start up

openaoe/backend/config/biz_config.py

@@ -1,6 +1,7 @@
 import argparse
 import os
 import sys
+from copy import deepcopy
 
 import yaml
 
@@ -12,23 +13,53 @@
 
 class BizConfig:
     def __init__(self, **args):
+        # raw dict
         self.__dict__.update(args)
 
-    def get(self, field):
-        if field in self.__dict__:
-            return self.__dict__[field]
-        return None
-
-
-def init_config():
-    parser = argparse.ArgumentParser(description="Example app using a YAML config file.")
+    # provider -> model_name -> ModelConfig
+    @property
+    def models_map(self):
+        if "models" not in self.__dict__:
+            return {}
+
+        models_dict = self.__dict__["models"]
+        ret = {}
+        for model_name, model_config in models_dict.items():
+            if model_config["provider"] not in ret:
+                ret[model_config["provider"]] = {}
+            ret[model_config["provider"]][model_name] = ModelConfig(model_config["webui"], model_config["api"])
+        return ret
+
+    @models_map.getter
+    def __get_models_map(self):
+        return self.models_map
+
+    @property
+    def json(self):
+        ret = deepcopy(self.__dict__)
+        if "models" in ret:
+            models_config = ret["models"]
+        for model_name, config in models_config.items():
+            config.pop("api")
+            ret["models"][model_name] = config
+        return ret
+
+
+class ModelConfig:
+    def __init__(self, webui_config, api_config):
+        self.webui_config = webui_config
+        self.api_config = api_config
+
+
+def init_config() -> BizConfig:
+    parser = argparse.ArgumentParser(description="LLM group chat framework")
     parser.add_argument('-f', '--file', type=str, required=True, help='Path to the YAML config file.')
     config_path = parser.parse_args()
     logger.info(f"your config file is: {config_path.file}")
-    load_config(config_path.file)
+    return load_config(config_path.file)
 
 
-def load_config(config_path):
+def load_config(config_path) -> BizConfig:
     logger.info(f"start to init configuration from {config_path}.")
     if not os.path.isfile(config_path):
         logger.error(f"invalid path: {config_path}, not exist or not file")
@@ -43,32 +74,46 @@ def load_config(config_path):
         global biz_config
         biz_config = BizConfig(**m)
     logger.info("init configuration successfully.")
+    return biz_config
 
 
-def get_configuration(field):
-    return biz_config.get(field)
-
-
-def get_model_configuration(vendor: str, field):
-    models = get_configuration("models")
+def get_model_configuration(provider: str, field, model_name: str = None):
+    models = biz_config.models_map
     if not models:
         logger.error(f"invalid configuration file")
         sys.exit(-1)
 
-    if models.get(vendor) and models.get(vendor).get(field):
-        conf = models.get(vendor).get(field)
-        return conf
-
-    logger.error(f"vendor: {vendor} has no field: {field} configuration")
+    provider_config = models.get(provider)
+    if provider_config:
+        if model_name:
+            try:
+                return provider_config.get(model_name).api_config.get(field)
+            except:
+                for config_model_name, config in provider_config.items():
+                    if config_model_name.startswith(model_name):
+                        return config.api_config.get(field)
+                logger.error(f"{provider} get field: {field} for model: {model_name} failed")
+                return ""
+        elif not model_name:
+            # default the first provider ModelConfig
+            provider_models_config_list = list(provider_config.values())
+            try:
+                logger.info(f"{provider} get field: {field} for anonymous model, use the first one as default.")
+                return provider_models_config_list[0].api_config.get(field)
+            except:
+                logger.error(f"{provider} get field: {field} for anonymous model failed")
+                return ""
+
+    logger.error(f"provider: {provider} has no field: {field} configuration for model: {model_name}")
     return ""
 
 
-def get_base_url(vendor: str) -> str:
-    return get_model_configuration(vendor, "api_base")
+def get_base_url(provider: str, model_name: str = None) -> str:
+    return get_model_configuration(provider, "api_base", model_name)
 
 
-def get_api_key(vendor: str) -> str:
-    return get_model_configuration(vendor, "api_key")
+def get_api_key(provider: str, model_name: str = None) -> str:
+    return get_model_configuration(provider, "api_key", model_name)
 
 
 def app_abs_path():

openaoe/backend/config/config-template.yaml

@@ -1,29 +1,71 @@
 ---
 models:
-    gpt:
-        api_base: https://api.openai.com/v1
-        api_key:
-
-    claude:
-        api_key:
-
-    bard:
-        api_base: https://bard.google.com
-        api_key:
-
-    minimax:
-        api_base: https://api.minimax.chat
-        group_id:
-        jwt:
-
+    internlm-chat-7b:
+        provider: internlm
+        webui:
+            avatar: 'https://oss.openmmlab.com/frontend/OpenAOE/internlm.svg'
+            background: 'linear-gradient(rgb(3 26 108 / 85%) 0%, rgb(29 60 161 / 85%) 100%)'
+        api:
+            api_base: http://localhost:23333
+    gpt-3.5-turbo:
+        provider: openai
+        webui:
+            avatar: 'https://oss.openmmlab.com/frontend/OpenAOE/openai.svg'
+            background: 'linear-gradient(180deg, rgba(156, 206, 116, 0.15) 0%, #1a8d15 100%)'
+        api:
+            api_base: https://api.openai.com/v1
+            api_key:
+    gpt-4:
+        provider: openai
+        webui:
+            avatar: 'https://oss.openmmlab.com/frontend/OpenAOE/openai.svg'
+            background: 'linear-gradient(180deg, rgba(156, 206, 116, 0.15) 0%, #08be00 100%)'
+        api:
+            api_base: https://api.openai.com/v1
+            api_key:
+    claude-1:
+        provider: claude
+        webui:
+            avatar: 'https://oss.openmmlab.com/frontend/OpenAOE/claude.svg'
+            background: 'linear-gradient(180deg, rgba(141, 90, 181, 0.15) 0%, rgba(106, 39, 123, 0.7) 53.12%, #663E9A 100%)'
+        api:
+            api_base: https://api.anthropic.com
+            api_key:
+    claude-1-100k:
+        provider: claude
+        webui:
+            avatar: 'https://oss.openmmlab.com/frontend/OpenAOE/claude.svg'
+            background: 'linear-gradient(180deg, rgba(141, 90, 181, 0.15) 0%, rgba(106, 39, 123, 0.7) 53.12%, #663E9A 100%)'
+        api:
+            api_base: https://api.anthropic.com
+            api_key:
+    chat-bison-001:
+        provider: google
+        webui:
+            avatar: 'https://oss.openmmlab.com/frontend/OpenAOE/google-palm.webp'
+            isStream: false
+            background: 'linear-gradient(180deg, rgba(181, 90, 90, 0.15) 0%, #fa5ab1 100%)'
+        api:
+            api_base: https://generativelanguage.googleapis.com
+            api_key:
+    abab5-chat:
+        provider: minimax
+        webui:
+            avatar: 'https://oss.openmmlab.com/frontend/OpenAOE/minimax.png'
+            background: 'linear-gradient(180deg, rgba(207, 72, 72, 0.15) 0%, rgba(151, 43, 43, 0.7) 53.12%, #742828 100%)'
+        api:
+            api_base: https://api.minimax.chat
+            group_id:
+            jwt:
     spark:
-        api_base: wss://spark-api.xf-yun.com
-        app_id:
-        ak:
-        sk:
-
-    internlm:
-        api_base:
+        provider: spark
+        webui:
+            avatar: 'https://oss.openmmlab.com/frontend/OpenAOE/spark.svg'
+            isStream: false
+            background: 'linear-gradient(180deg, rgba(72, 72, 207, 0.15) 0%, #7498be 100%)'
+        api:
+            api_base: wss://spark-api.xf-yun.com/v2.1/chat
+            app_id:
+            ak:
+            sk:
 ...
-
-

openaoe/backend/config/constant.py

@@ -5,12 +5,12 @@
 DATE_PATTERN = "%Y%m%d"
 TIMEOUT_SECONDS = 30
 
-VENDOR_OPENAI = "gpt"
-VENDOR_MINIMAX = "minimax"
-VENDOR_GOOGLE = "bard"
-VENDOR_XUNFEI = "spark"
-VENDOR_CLAUDE = "claude"
-VENDOR_INTERNLM = "internlm"
+PROVIDER_OPENAI = "openai"
+PROVIDER_MINIMAX = "minimax"
+PROVIDER_GOOGLE = "google"
+PROVIDER_XUNFEI = "spark"
+PROVIDER_CLAUDE = "claude"
+PROVIDER_INTERNLM = "internlm"
 
 DEFAULT_TIMEOUT_SECONDS = 600
 

openaoe/backend/service/service_claude.py

@@ -4,9 +4,9 @@
 from anthropic import Anthropic, HUMAN_PROMPT, AI_PROMPT
 from sse_starlette.sse import EventSourceResponse
 
-from openaoe.backend.config.biz_config import get_api_key
+from openaoe.backend.config.biz_config import get_api_key, get_base_url
 from openaoe.backend.config.constant import TYPE_BOT, TYPE_USER, TYPE_SYSTEM
-from openaoe.backend.config.constant import VENDOR_CLAUDE
+from openaoe.backend.config.constant import PROVIDER_CLAUDE
 from openaoe.backend.model.aoe_response import AOEResponse
 from openaoe.backend.model.claude import ClaudeChatBody, ClaudeMessage
 
@@ -16,7 +16,8 @@ def claude_chat_stream_svc(request, body: ClaudeChatBody):
     stream api logic for Claude model
     use anthropic SDK: https://github.com/anthropics/anthropic-sdk-python
     """
-    api_key = get_api_key(VENDOR_CLAUDE)
+    api_key = get_api_key(PROVIDER_CLAUDE, body.model)
+    api_base = get_base_url(PROVIDER_CLAUDE, body.model)
     prompt = _gen_prompt(body.messages)
     if not prompt or len(prompt) == 0:
         return AOEResponse(
@@ -25,7 +26,7 @@ def claude_chat_stream_svc(request, body: ClaudeChatBody):
             data="prompt or messages must be set"
         )
 
-    anthropic = Anthropic(api_key=api_key)
+    anthropic = Anthropic(api_key=api_key, base_url=api_base)
 
     async def stream():
         try:

openaoe/backend/service/service_google.py

@@ -15,8 +15,8 @@ def palm_chat_svc(body: GooglePalmChatBody):
     """
     chat logic for google PaLM model
     """
-    api_key = get_api_key(VENDOR_GOOGLE)
-    url = get_base_url(VENDOR_GOOGLE)
+    api_key = get_api_key(PROVIDER_GOOGLE, body.model)
+    url = get_base_url(PROVIDER_GOOGLE, body.model)
     url = f"{url}/google/v1beta2/models/{body.model}:generateMessage?key={api_key}"
     messages = [
         {"content": msg.content, "author": msg.author}

openaoe/backend/service/service_internlm.py

@@ -30,7 +30,7 @@ def chat_completion_v1(request, body: InternlmChatCompletionBody):
     }
     msgs.append(msg_item)
     # restful api
-    url = get_base_url(VENDOR_INTERNLM) + "/v1/chat/completions"
+    url = get_base_url(PROVIDER_INTERNLM, body.model) + "/v1/chat/completions"
     headers = {
         'accept': 'application/json',
         'Content-Type': 'application/json'
@@ -42,7 +42,7 @@ def chat_completion_v1(request, body: InternlmChatCompletionBody):
         "top_p": body.top_p,
         "n": body.n,
         "max_tokens": body.max_tokens,
-        "stop": False,
+        "stop": "false",
         "stream": body.stream,
         "presence_penalty": body.presence_penalty,
         "frequency_penalty": body.frequency_penalty,

openaoe/backend/service/service_minimax.py

@@ -12,9 +12,9 @@
 
 
 def _get_req_param(body):
-    group_id = get_model_configuration(VENDOR_MINIMAX, "group_id")
-    jwt = get_model_configuration(VENDOR_MINIMAX, "jwt")
-    api_base = get_base_url(VENDOR_MINIMAX)
+    group_id = get_model_configuration(PROVIDER_MINIMAX, "group_id", body.model)
+    jwt = get_model_configuration(PROVIDER_MINIMAX, "jwt", body.model)
+    api_base = get_base_url(PROVIDER_MINIMAX, body.model)
     headers = {
         "Authorization": jwt,
         "Content-Type": "application/json"

openaoe/backend/service/service_openai.py

@@ -43,9 +43,9 @@ def chat_completion_stream(request, body):
     async def event_generator():
         while True:
             client = OpenAI(
-                api_key=get_api_key(VENDOR_OPENAI),
+                api_key=get_api_key(PROVIDER_OPENAI, body.model),
                 timeout=body.timeout,
-                base_url=get_base_url(VENDOR_OPENAI)
+                base_url=get_base_url(PROVIDER_OPENAI, body.model)
             )
 
             stop_flag = False
@@ -80,9 +80,9 @@ async def event_generator():
     async def event_generator_json():
         while True:
             client = OpenAI(
-                api_key=get_api_key(VENDOR_OPENAI),
+                api_key=get_api_key(PROVIDER_OPENAI, body.model),
                 timeout=body.timeout,
-                base_url=get_base_url(VENDOR_OPENAI)
+                base_url=get_base_url(PROVIDER_OPENAI, body.model)
             )
             stop_flag = False
             response = ""

openaoe/backend/service/service_xunfei.py

@@ -8,7 +8,7 @@
 from websocket import create_connection
 
 from openaoe.backend.config.biz_config import get_model_configuration, get_base_url
-from openaoe.backend.config.constant import VENDOR_XUNFEI
+from openaoe.backend.config.constant import PROVIDER_XUNFEI
 from openaoe.backend.model.aoe_response import AOEResponse
 from openaoe.backend.model.xunfei import XunfeiSparkChatBody
 from openaoe.backend.util.log import log
@@ -59,10 +59,10 @@ def spark_chat_svc(body: XunfeiSparkChatBody):
     """
     chat logic for spark model
     """
-    api_base = get_base_url(VENDOR_XUNFEI)
-    app_id = get_model_configuration(VENDOR_XUNFEI, "app_id")
-    ak = get_model_configuration(VENDOR_XUNFEI, "ak")
-    sk = get_model_configuration(VENDOR_XUNFEI, "sk")
+    api_base = get_base_url(PROVIDER_XUNFEI)
+    app_id = get_model_configuration(PROVIDER_XUNFEI, "app_id")
+    ak = get_model_configuration(PROVIDER_XUNFEI, "ak")
+    sk = get_model_configuration(PROVIDER_XUNFEI, "sk")
 
     url_parse = urllib.parse.urlparse(api_base)
     host = url_parse.hostname