File size: 5,176 Bytes
432dbc1 41ec323 432dbc1 41ec323 432dbc1 41ec323 432dbc1 41ec323 432dbc1 41ec323 79e40b4 41ec323 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 |
#############################################################################################################
# Title: Gradio Interface to LLM-chatbot (for recommending AI) with RAG-funcionality and ChromaDB on HF-Hub
# Author: Andreas Fischer
# Date: December 30th, 2023
# Last update: December 30th, 2023
##############################################################################################################
# Chroma-DB
#-----------
import os
import chromadb
dbPath="/home/af/Schreibtisch/gradio/Chroma/db"
if(os.path.exists(dbPath)==False):
dbPath="/home/user/app/db"
print(dbPath)
#client = chromadb.Client()
path=dbPath
client = chromadb.PersistentClient(path=path)
print(client.heartbeat())
print(client.get_version())
print(client.list_collections())
from chromadb.utils import embedding_functions
default_ef = embedding_functions.DefaultEmbeddingFunction()
sentence_transformer_ef = embedding_functions.SentenceTransformerEmbeddingFunction(model_name="T-Systems-onsite/cross-en-de-roberta-sentence-transformer")
#instructor_ef = embedding_functions.InstructorEmbeddingFunction(model_name="hkunlp/instructor-large", device="cuda")
print(str(client.list_collections()))
global collection
if("name=ChromaDB1" in str(client.list_collections())):
print("ChromaDB1 found!")
collection = client.get_collection(name="ChromaDB1", embedding_function=sentence_transformer_ef)
else:
print("ChromaDB1 created!")
collection = client.create_collection(
"ChromaDB1",
embedding_function=sentence_transformer_ef,
metadata={"hnsw:space": "cosine"})
collection.add(
documents=[
"Text generating AI model mistralai/Mixtral-8x7B-Instruct-v0.1: Perfect for text generation, e.g., social media content, marketing copy, blog posts, short stories, etc.",
"Image generating AI model stabilityai/sdxl-turbo: Perfect for image generation, e.g., illustrations, graphics, AI art, etc.",
"Audio transcribing AI model openai/whisper-large-v3: Perfect for audio-transcription in different languages",
"Speech synthesizing AI model coqui/XTTS-v2: Perfect for generating audio from text and for voice-cloning"
],
metadatas=[{"source": "AF"}, {"source": "AF"}, {"source": "AF"}, {"source": "AF"}],
ids=["ai1", "ai2", "ai3", "ai4"],
)
print("Database ready!")
print(collection.count())
# Model
#-------
from huggingface_hub import InferenceClient
import gradio as gr
client = InferenceClient(
"mistralai/Mixtral-8x7B-Instruct-v0.1"
#"mistralai/Mistral-7B-Instruct-v0.1"
)
# Gradio-GUI
#------------
import gradio as gr
import json
def format_prompt(message, history):
prompt = "<s>"
for user_prompt, bot_response in history:
prompt += f"[INST] {user_prompt} [/INST]"
prompt += f" {bot_response}</s> "
prompt += f"[INST] {message} [/INST]"
return prompt
def response(
prompt, history, temperature=0.9, max_new_tokens=500, top_p=0.95, repetition_penalty=1.0,
):
temperature = float(temperature)
if temperature < 1e-2: temperature = 1e-2
top_p = float(top_p)
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=max_new_tokens,
top_p=top_p,
repetition_penalty=repetition_penalty,
do_sample=True,
seed=42,
)
addon=""
results=collection.query(
query_texts=[prompt],
n_results=2,
#where={"source": "google-docs"}
#where_document={"$contains":"search_string"}
)
dists=["<small>(relevance: "+str(round((1-d)*100)/100)+";" for d in results['distances'][0]]
sources=["source: "+s["source"]+")</small>" for s in results['metadatas'][0]]
results=results['documents'][0]
combination = zip(results,dists,sources)
combination = [' '.join(triplets) for triplets in combination]
print(combination)
if(len(results)>1):
addon=" Bitte berücksichtige bei deiner Antwort ggf. folgende Auszüge aus unserer Datenbank, sofern sie für die Antwort erforderlich sind. Beantworte die Frage knapp und präzise. Ignoriere unpassende Datenbank-Auszüge OHNE sie zu kommentieren, zu erwähnen oder aufzulisten:\n"+"\n".join(results)
system="Du bist ein deutschsprachiges KI-basiertes Assistenzsystem, das zu jedem Anliegen möglichst geeignete KI-Tools empfiehlt."+addon+"\n\nUser-Anliegen:"
#body={"prompt":system+"### Instruktion:\n"+message+"\n\n### Antwort:","max_tokens":500, "echo":"False","stream":"True"} #e.g. SauerkrautLM
formatted_prompt = format_prompt(system+"\n"+prompt, history)
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
output += response.token.text
yield output
output=output+"\n\n<br><details open><summary><strong>Sources</strong></summary><br><ul>"+ "".join(["<li>" + s + "</li>" for s in combination])+"</ul></details>"
yield output
gr.ChatInterface(response, chatbot=gr.Chatbot(render_markdown=True),title="German AI-RAG-Interface to the Hugging Face Hub").queue().launch(share=True) #False, server_name="0.0.0.0", server_port=7864)
print("Interface up and running!") |