File size: 4,184 Bytes
dd73cae
 
75ac77a
 
dd73cae
 
 
6e0c650
 
 
 
 
dd73cae
4a61c36
899f53c
 
 
 
14cfc79
75ac77a
dd73cae
4a61c36
75ac77a
 
0c1e99a
6154019
1963515
dd73cae
4a61c36
dd73cae
fd967d6
3debbd5
f6a820d
8ee8377
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3debbd5
f6a820d
3debbd5
dd73cae
7d10598
75ac77a
8ee8377
 
 
f6a820d
8ee8377
 
3b53eb2
8c3856c
75ac77a
dd73cae
8ee8377
97739c9
dd73cae
fd967d6
 
dd73cae
fd967d6
 
 
 
dd73cae
 
 
6b6e206
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
import gradio as gr
import os
from huggingface_hub import hf_hub_download
from llama_cpp import Llama

hf_token = os.environ.get("HF_TOKEN")

# repo_id = "JAZL/ChemGist-Phi3-Q4_K_M-GGUF" 
# filename = "chemgist-phi3-q4_k_m.gguf" 

repo_id = "JAZL/ChemGist-Phi3-Q8_0-GGUF" 
filename = "chemgist-phi3-q8_0.gguf" 

print(f"Downloading {filename} from {repo_id}...")
model_path = hf_hub_download(
    repo_id=repo_id, 
    filename=filename, 
    token=hf_token,
    force_download=False
)

print("Loading model into llama.cpp engine...")
llm = Llama(
    model_path=model_path,
    n_ctx=4096,        
    n_threads=2,      
    verbose=True     
)
print("Model loaded and ready!")

def generate_answer(message: str):
    system_prompt = (
        "You are ChemGist, a highly intelligent and conversational chemistry AI assistant. "
        "You are fine-tuned only for Organic Chemistry topics and chemical profile information found within your fine-tuning scope. "
        "Your purpose is to answer questions only about organic chemistry, organic compounds, functional groups, molecular structures, "
        "SMILES, InChIKeys, molecular formulas, physical properties, theoretical properties, uses, history, and organic chemical concepts. "

        "You must strictly follow these rules: "

        "1. Only answer questions that are clearly related to Organic Chemistry or the chemical information covered in your fine-tuned scope. "

        "2. If the user asks about topics outside Organic Chemistry, such as physics, biology, inorganic chemistry, analytical chemistry, politics, "
        "programming, mathematics, general trivia, personal advice, or unrelated subjects, politely refuse and say that ChemGist can only answer "
        "questions related to Organic Chemistry. "

        "3. If the user asks about a compound, structure, SMILES, InChIKey, molecular formula, property, use, or history that is not known or not confidently "
        "available within your fine-tuned scope, do not guess. Instead, say that the information is not available within ChemGist's fine-tuned organic chemistry scope. "

        "4. Do not invent chemical identifiers, SMILES strings, InChIKeys, formulas, structures, physical properties, theoretical properties, uses, or historical details. "

        "5. Answer only the specific question asked by the user. Do not include unrelated sections or unnecessary extra information. "

        "6. When asked about a chemical compound, respond naturally and conversationally. Use Markdown bullet points only when listing physical properties, "
        "theoretical properties, identifiers, or uses. "

        "7. Keep responses brief, direct, and clear unless the user asks for a detailed explanation. "

        "8. Never repeat words, stutter, loop, or generate irrelevant text. "

        "9. If the question is ambiguous, ask the user to clarify the organic chemistry compound, property, or concept they are asking about. "

        "10. Your default refusal response should be: "
        "'I’m ChemGist, and I can only answer questions related to Organic Chemistry within my fine-tuned scope. Please ask about an organic compound, "
        "functional group, molecular structure, reaction, SMILES, InChIKey, properties, uses, or related organic chemistry topic.'"
    )

    prompt = f"<s><|system|>\n{system_prompt}<|end|>\n<|user|>\n{message}<|end|>\n<|assistant|>\n"
    
    response = llm(
        prompt,
        max_tokens=512,
        temperature=0.03,
        top_p=0.9,
        min_p=0.05,
        repeat_penalty=1.12,
        stop=["<|end|>", "<|user|>", "<s>", "</s>"],
        echo=False,
        stream=False,
    )
    
    answer = response["choices"][0]["text"].strip()
    return answer

# Interface
demo = gr.Interface(
    fn=generate_answer,
    inputs=gr.Textbox(label="User Question", placeholder="Type your chemistry question here... (e.g., Tell me about Ammonia)"),
    outputs=gr.Textbox(label="ChemGist Answer"),
    title="ChemGist API Server",
    description="Ask me about chemical definitions, physical properties, or molecular structures."
)

if __name__ == "__main__":
    demo.launch(ssr_mode=False)