feat: bump version to 0.3.31-fix2 (#1592)

api/core/model_providers/providers/anthropic_provider.py

@@ -32,9 +32,12 @@ class AnthropicProvider(BaseModelProvider):
         if model_type == ModelType.TEXT_GENERATION:
             return [
                 {
-                    'id': 'claude-instant-1',
-                    'name': 'claude-instant-1',
+                    'id': 'claude-2.1',
+                    'name': 'claude-2.1',
                     'mode': ModelMode.CHAT.value,
+                    'features': [
+                        ModelFeature.AGENT_THOUGHT.value
+                    ]
                 },
                 {
                     'id': 'claude-2',
@@ -44,6 +47,11 @@ class AnthropicProvider(BaseModelProvider):
                         ModelFeature.AGENT_THOUGHT.value
                     ]
                 },
+                {
+                    'id': 'claude-instant-1',
+                    'name': 'claude-instant-1',
+                    'mode': ModelMode.CHAT.value,
+                },
             ]
         else:
             return []
@@ -73,12 +81,18 @@ class AnthropicProvider(BaseModelProvider):
         :param model_type:
         :return:
         """
+        model_max_tokens = {
+            'claude-instant-1': 100000,
+            'claude-2': 100000,
+            'claude-2.1': 200000,
+        }
+
         return ModelKwargsRules(
             temperature=KwargRule[float](min=0, max=1, default=1, precision=2),
             top_p=KwargRule[float](min=0, max=1, default=0.7, precision=2),
             presence_penalty=KwargRule[float](enabled=False),
             frequency_penalty=KwargRule[float](enabled=False),
-            max_tokens=KwargRule[int](alias="max_tokens_to_sample", min=10, max=100000, default=256, precision=0),
+            max_tokens=KwargRule[int](alias="max_tokens_to_sample", min=10, max=model_max_tokens.get(model_name, 100000), default=256, precision=0),
         )
 
     @classmethod

api/core/model_providers/rules/anthropic.json

@@ -23,8 +23,14 @@
             "currency": "USD"
         },
         "claude-2": {
-            "prompt": "11.02",
-            "completion": "32.68",
+            "prompt": "8.00",
+            "completion": "24.00",
+            "unit": "0.000001",
+            "currency": "USD"
+        },
+        "claude-2.1": {
+            "prompt": "8.00",
+            "completion": "24.00",
             "unit": "0.000001",
             "currency": "USD"
         }

api/core/third_party/langchain/llms/anthropic_llm.py

@@ -1,7 +1,7 @@
 from typing import Dict
 
-from httpx import Limits
 from langchain.chat_models import ChatAnthropic
+from langchain.schema import ChatMessage, BaseMessage, HumanMessage, AIMessage, SystemMessage
 from langchain.utils import get_from_dict_or_env, check_package_version
 from pydantic import root_validator
 
@@ -29,8 +29,7 @@ class AnthropicLLM(ChatAnthropic):
                 base_url=values["anthropic_api_url"],
                 api_key=values["anthropic_api_key"],
                 timeout=values["default_request_timeout"],
-                max_retries=0,
-                connection_pool_limits=Limits(max_connections=200, max_keepalive_connections=100),
+                max_retries=0
             )
             values["async_client"] = anthropic.AsyncAnthropic(
                 base_url=values["anthropic_api_url"],
@@ -46,3 +45,16 @@ class AnthropicLLM(ChatAnthropic):
                 "Please it install it with `pip install anthropic`."
             )
         return values
+
+    def _convert_one_message_to_text(self, message: BaseMessage) -> str:
+        if isinstance(message, ChatMessage):
+            message_text = f"\n\n{message.role.capitalize()}: {message.content}"
+        elif isinstance(message, HumanMessage):
+            message_text = f"{self.HUMAN_PROMPT} {message.content}"
+        elif isinstance(message, AIMessage):
+            message_text = f"{self.AI_PROMPT} {message.content}"
+        elif isinstance(message, SystemMessage):
+            message_text = f"{message.content}"
+        else:
+            raise ValueError(f"Got unknown type {message}")
+        return message_text

api/requirements.txt

@@ -35,7 +35,7 @@ docx2txt==0.8
 pypdfium2==4.16.0
 resend~=0.5.1
 pyjwt~=2.6.0
-anthropic~=0.3.4
+anthropic~=0.7.2
 newspaper3k==0.2.8
 google-api-python-client==2.90.0
 wikipedia==1.4.0

api/tests/unit_tests/model_providers/test_anthropic_provider.py

@@ -31,12 +31,12 @@ def mock_chat_generate_invalid(messages: List[BaseMessage],
                                run_manager: Optional[CallbackManagerForLLMRun] = None,
                                **kwargs: Any):
     raise anthropic.APIStatusError('Invalid credentials',
-                                   request=httpx._models.Request(
-                                       method='POST',
-                                       url='https://api.anthropic.com/v1/completions',
-                                   ),
                                    response=httpx._models.Response(
                                        status_code=401,
+                                       request=httpx._models.Request(
+                                           method='POST',
+                                           url='https://api.anthropic.com/v1/completions',
+                                       )
                                    ),
                                    body=None
                                 )

docker/docker-compose.yaml

@@ -2,7 +2,7 @@ version: '3.1'
 services:
   # API service
   api:
-    image: langgenius/dify-api:0.3.31-fix1
+    image: langgenius/dify-api:0.3.31-fix2
     restart: always
     environment:
       # Startup mode, 'api' starts the API server.
@@ -128,7 +128,7 @@ services:
   # worker service
   # The Celery worker for processing the queue.
   worker:
-    image: langgenius/dify-api:0.3.31-fix1
+    image: langgenius/dify-api:0.3.31-fix2
     restart: always
     environment:
       # Startup mode, 'worker' starts the Celery worker for processing the queue.
@@ -196,7 +196,7 @@ services:
 
   # Frontend web application.
   web:
-    image: langgenius/dify-web:0.3.31-fix1
+    image: langgenius/dify-web:0.3.31-fix2
     restart: always
     environment:
       EDITION: SELF_HOSTED

web/global.d.ts

@@ -1,5 +1,3 @@
 declare module 'lamejs';
 declare module 'react-18-input-autosize';
-declare module 'fetch-readablestream' {
-  export default function fetchReadableStream(url: string, options?: RequestInit): Promise<Response>
-}
+

web/package.json

@@ -36,7 +36,6 @@
     "echarts": "^5.4.1",
     "echarts-for-react": "^3.0.2",
     "emoji-mart": "^5.5.2",
-    "fetch-readablestream": "^0.2.0",
     "i18next": "^22.4.13",
     "i18next-resources-to-backend": "^1.1.3",
     "immer": "^9.0.19",

web/service/base.ts

@@ -1,11 +1,8 @@
-import fetchStream from 'fetch-readablestream'
 import { API_PREFIX, IS_CE_EDITION, PUBLIC_API_PREFIX } from '@/config'
 import Toast from '@/app/components/base/toast'
 import type { MessageEnd, MessageReplace, ThoughtItem } from '@/app/components/app/chat/type'
-import { isSupportNativeFetchStream } from '@/utils/stream'
 
 const TIME_OUT = 100000
-const supportNativeFetchStream = isSupportNativeFetchStream()
 
 const ContentType = {
   json: 'application/json',
@@ -223,9 +220,6 @@ const baseFetch = <T>(
   if (body && bodyStringify)
     options.body = JSON.stringify(body)
 
-  // for those do not support native fetch stream, we use fetch-readablestream as polyfill
-  const doFetch = supportNativeFetchStream ? globalThis.fetch : fetchStream
-
   // Handle timeout
   return Promise.race([
     new Promise((resolve, reject) => {
@@ -234,7 +228,7 @@ const baseFetch = <T>(
       }, TIME_OUT)
     }),
     new Promise((resolve, reject) => {
-      doFetch(urlWithPrefix, options as RequestInit)
+      globalThis.fetch(urlWithPrefix, options as RequestInit)
         .then((res) => {
           const resClone = res.clone()
           // Error handler

web/utils/stream.ts

@@ -1,21 +0,0 @@
-// https://developer.chrome.com/articles/fetch-streaming-requests/#feature-detection
-export const isSupportNativeFetchStream = () => {
-  const supportsRequestStreams = (() => {
-    let duplexAccessed = false
-
-    const params = {
-      body: new ReadableStream(),
-      method: 'POST',
-      get duplex() {
-        duplexAccessed = true
-        return 'half'
-      },
-    }
-
-    const hasContentType = new Request('', params).headers.has('Content-Type')
-
-    return duplexAccessed && !hasContentType
-  })()
-
-  return supportsRequestStreams
-}