add support for Anthropic (#2148)

### What problem does this PR solve?

#1853  add support for Anthropic

### Type of change

- [x] New Feature (non-breaking change which adds functionality)

---------

Co-authored-by: Zhedong Cen <cenzhedong2@126.com>
Co-authored-by: Kevin Hu <kevinhu.sh@gmail.com>

conf/llm_factories.json

@@ -3240,6 +3240,56 @@
             "tags": "SPEECH2TEXT",
             "status": "1",
             "llm": []
+        },
+        {
+            "name": "Anthropic",
+            "logo": "",
+            "tags": "LLM",
+            "status": "1",
+            "llm": [
+                {
+                    "llm_name": "claude-3-5-sonnet-20240620",
+                    "tags": "LLM,CHAT,200k",
+                    "max_tokens": 204800,
+                    "model_type": "chat"
+                },
+                {
+                    "llm_name": "claude-3-opus-20240229",
+                    "tags": "LLM,CHAT,200k",
+                    "max_tokens": 204800,
+                    "model_type": "chat"
+                },
+                {
+                    "llm_name": "claude-3-sonnet-20240229",
+                    "tags": "LLM,CHAT,200k",
+                    "max_tokens": 204800,
+                    "model_type": "chat"
+                },
+                {
+                    "llm_name": "claude-3-haiku-20240307",
+                    "tags": "LLM,CHAT,200k",
+                    "max_tokens": 204800,
+                    "model_type": "chat"
+                },
+                {
+                    "llm_name": "claude-2.1",
+                    "tags": "LLM,CHAT,200k",
+                    "max_tokens": 204800,
+                    "model_type": "chat"
+                },
+                {
+                    "llm_name": "claude-2.0",
+                    "tags": "LLM,CHAT,100k",
+                    "max_tokens": 102400,
+                    "model_type": "chat"
+                },
+                {
+                    "llm_name": "claude-instant-1.2",
+                    "tags": "LLM,CHAT,100k",
+                    "max_tokens": 102400,
+                    "model_type": "chat"
+                }
+            ]
         }
     ]
 }

rag/llm/__init__.py

@@ -104,7 +104,8 @@ ChatModel = {
     "Replicate": ReplicateChat,
     "Tencent Hunyuan": HunyuanChat,
     "XunFei Spark": SparkChat,
-    "BaiduYiyan": BaiduYiyanChat
+    "BaiduYiyan": BaiduYiyanChat,
+    "Anthropic": AnthropicChat
 }
 
 

rag/llm/chat_model.py

@@ -1193,3 +1193,67 @@ class BaiduYiyanChat(Base):
             return ans + "\n**ERROR**: " + str(e), 0
 
         yield total_tokens
+
+
+class AnthropicChat(Base):
+    def __init__(self, key, model_name, base_url=None):
+        import anthropic
+
+        self.client = anthropic.Anthropic(api_key=key)
+        self.model_name = model_name
+        self.system = ""
+
+    def chat(self, system, history, gen_conf):
+        if system:
+            self.system = system
+        if "max_tokens" not in gen_conf:
+            gen_conf["max_tokens"] = 4096
+
+        try:
+            response = self.client.messages.create(
+                model=self.model_name,
+                messages=history,
+                system=self.system,
+                stream=False,
+                **gen_conf,
+            ).json()
+            ans = response["content"][0]["text"]
+            if response["stop_reason"] == "max_tokens":
+                ans += (
+                    "...\nFor the content length reason, it stopped, continue?"
+                    if is_english([ans])
+                    else "······\n由于长度的原因，回答被截断了，要继续吗？"
+                )
+            return (
+                ans,
+                response["usage"]["input_tokens"] + response["usage"]["output_tokens"],
+            )
+        except Exception as e:
+            return ans + "\n**ERROR**: " + str(e), 0
+
+    def chat_streamly(self, system, history, gen_conf):
+        if system:
+            self.system = system
+        if "max_tokens" not in gen_conf:
+            gen_conf["max_tokens"] = 4096
+
+        ans = ""
+        total_tokens = 0
+        try:
+            response = self.client.messages.create(
+                model=self.model_name,
+                messages=history,
+                system=self.system,
+                stream=True,
+                **gen_conf,
+            )
+            for res in response.iter_lines():
+                res = res.decode("utf-8")
+                if "content_block_delta" in res and "data" in res:
+                    text = json.loads(res[6:])["delta"]["text"]
+                    ans += text
+                    total_tokens += num_tokens_from_string(text)
+        except Exception as e:
+            yield ans + "\n**ERROR**: " + str(e)
+
+        yield total_tokens

requirements.txt

@@ -1,3 +1,4 @@
+anthropic===0.34.1   
 arxiv==2.1.3
 Aspose.Slides==24.2.0
 BCEmbedding==0.1.3

requirements_arm.txt

@@ -2,6 +2,7 @@ accelerate==0.27.2
 aiohttp==3.9.4
 aiosignal==1.3.1
 annotated-types==0.6.0
+anthropic===0.34.1   
 anyio==4.3.0
 argon2-cffi==23.1.0
 argon2-cffi-bindings==21.2.0

web/src/assets/svg/llm/anthropic.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1724743631411" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="4224" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M710 196.2H572.9l250 631.6H960L710 196.2z m-396 0L64 827.8h139.8l51.1-132.6h261.5l51.1 132.6h139.8l-250-631.6H314z m-13.9 381.7l85.5-222 85.5 222h-171z" p-id="4225"></path></svg>

web/src/pages/user-setting/setting-model/constant.ts

@@ -37,6 +37,7 @@ export const IconMap = {
   BaiduYiyan: 'yiyan',
   'Fish Audio': 'fish-audio',
   'Tencent Cloud': 'tencent-cloud',
+  Anthropic: 'anthropic',
 };
 
 export const BedrockRegionList = [