added SVG for Groq model model providers (#1470)

#1432  #1447 
This PR adds support for the GROQ LLM (Large Language Model).

Groq is an AI solutions company delivering ultra-low latency inference
with the first-ever LPU™ Inference Engine. The Groq API enables
developers to integrate state-of-the-art LLMs, such as Llama-2 and
llama3-70b-8192, into low latency applications with the request limits
specified below. Learn more at [groq.com](https://groq.com/).
Supported Models


| ID | Requests per Minute | Requests per Day | Tokens per Minute |

|----------------------|---------------------|------------------|-------------------|
| gemma-7b-it | 30 | 14,400 | 15,000 |
| gemma2-9b-it | 30 | 14,400 | 15,000 |
| llama3-70b-8192 | 30 | 14,400 | 6,000 |
| llama3-8b-8192 | 30 | 14,400 | 30,000 |
| mixtral-8x7b-32768 | 30 | 14,400 | 5,000 |

---------

Co-authored-by: paresh0628 <paresh.tuvoc@gmail.com>
Co-authored-by: Kevin Hu <kevinhu.sh@gmail.com>

rag/llm/__init__.py

@@ -32,7 +32,8 @@ EmbeddingModel = {
     "Jina": JinaEmbed,
     "BAAI": DefaultEmbedding,
     "Mistral": MistralEmbed,
-    "Bedrock": BedrockEmbed
+    "Bedrock": BedrockEmbed,
+    "Groq": GroqChat
 }
 
 

rag/llm/chat_model.py

@@ -23,6 +23,7 @@ from ollama import Client
 from volcengine.maas.v2 import MaasService
 from rag.nlp import is_english
 from rag.utils import num_tokens_from_string
+from groq import Groq
 
 
 class Base(ABC):
@@ -681,4 +682,63 @@ class GeminiChat(Base):
         except Exception as e:
             yield ans + "\n**ERROR**: " + str(e)
 
-        yield  response._chunks[-1].usage_metadata.total_token_count
+        yield  response._chunks[-1].usage_metadata.total_token_count
+
+
+
+class GroqChat:
+    def __init__(self, key, model_name,base_url=''):
+        self.client = Groq(api_key=key)
+        self.model_name = model_name
+
+    def chat(self, system, history, gen_conf):
+        if system:
+            history.insert(0, {"role": "system", "content": system})
+        for k in list(gen_conf.keys()):
+            if k not in ["temperature", "top_p", "max_tokens"]:
+                del gen_conf[k]
+
+        ans = ""
+        try:
+            response = self.client.chat.completions.create(
+                model=self.model_name,
+                messages=history,
+                **gen_conf
+            )
+            ans = response.choices[0].message.content
+            if response.choices[0].finish_reason == "length":
+                ans += "...\nFor the content length reason, it stopped, continue?" if self.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.total_tokens
+        except Exception as e:
+            return ans + "\n**ERROR**: " + str(e), 0
+
+    def chat_streamly(self, system, history, gen_conf):
+        if system:
+            history.insert(0, {"role": "system", "content": system})
+        for k in list(gen_conf.keys()):
+            if k not in ["temperature", "top_p", "max_tokens"]:
+                del gen_conf[k]
+        ans = ""
+        total_tokens = 0
+        try:
+            response = self.client.chat.completions.create(
+                model=self.model_name,
+                messages=history,
+                stream=True,
+                **gen_conf
+            )
+            for resp in response:
+                if not resp.choices or not resp.choices[0].delta.content:
+                    continue
+                ans += resp.choices[0].delta.content
+                total_tokens += 1
+                if resp.choices[0].finish_reason == "length":
+                    ans += "...\nFor the content length reason, it stopped, continue?" if self.is_english(
+                        [ans]) else "······\n由于长度的原因，回答被截断了，要继续吗？"
+                yield ans
+
+        except Exception as e:
+            yield ans + "\n**ERROR**: " + str(e)
+
+        yield total_tokens