Configuration Required
Missing API keys: {', '.join(api_errors)}
Please configure your .env file with the required API keys to continue.
""", unsafe_allow_html=True)
st.stop()
deepseek_client = OpenAI(
base_url="https://openrouter.ai/api/v1",
api_key=deepseek_api_key
)
gpt_client = OpenAI(
base_url="https://openrouter.ai/api/v1",
api_key=gpt_api_key
)
qwen_client = OpenAI(
base_url="https://openrouter.ai/api/v1",
api_key=qwen_api_key
)
def clean_code(text: str) -> str:
"""Remove markdown fences or extra text from code output."""
return text.strip().replace("```python", "").replace("```", "")
def generate_chart_code(columns: str, query: str) -> str:
messages = [
{"role": "system", "content": "You are an intelligent data visualization assistant."},
{"role": "user", "content": f"The user uploaded a dataset with the following columns:\n{columns}\n\nThe user asked: '{query}'\n\nWrite valid Python code that:\n1. Uses Plotly (px or go) to create the most appropriate chart.\n2. The dataframe is available as `df`.\n3. Save the chart as a variable named `fig`.\n\nOnly return the code. Do not include explanations."}
]
try:
response = deepseek_client.chat.completions.create(
model="deepseek/deepseek-chat-v3.1",
messages=messages,
temperature=0,
)
return clean_code(response.choices[0].message.content)
except Exception as e:
st.error(f" Error generating chart code: {e}")
return ""
def fix_chart_code(code: str, error: str) -> str:
messages = [
{"role": "system", "content": (
"You are a data visualization assistant. "
"IMPORTANT RULES:\n"
"1. Never create or inject fake/sample dataframes (like df = pd.DataFrame(...)).\n"
"2. Always assume the user already uploaded a dataframe named `df`.\n"
"3. If certain required columns are missing, insert a validation check that raises a ValueError with a clear message.\n"
"4. Always return corrected Python code that produces a Plotly chart named `fig`.\n"
"5. Only return the corrected code, no explanations."
)},
{"role": "user", "content": f"The following code failed:\n\n{code}\n\nError:\n{error}\n\nFix the code according to the rules."}
]
try:
response = deepseek_client.chat.completions.create(
model="deepseek/deepseek-chat-v3.1",
messages=messages,
temperature=0,
)
return clean_code(response.choices[0].message.content)
except Exception as e:
st.error(f"Error fixing chart code: {e}")
return code
def generate_explanation(query: str, columns: str, stats_summary: str) -> str:
code = generate_chart_code(columns, query)
messages = [
{"role": "system", "content": "You are a data visualization expert. Provide clear, concise explanations of generated chart.Don't show the <｜begin▁of▁sentence｜>"},
{"role": "user", "content": f"User query: '{query}'\nGenerated Graph code: '{code}'\nDataset columns: {columns}\nKey statistics from the chart data: {stats_summary}\n\nExplain the chart in natural language: Describe what it shows, key trends, insights, and implications. Keep it to 3-5 sentences. Be engaging and insightful."}
]
try:
response = deepseek_client.chat.completions.create(
model="deepseek/deepseek-chat-v3.1",
messages=messages,
temperature=0.3,
)
return response.choices[0].message.content.strip()
except Exception as e:
st.error(f"Error generating explanation: {e}")
return "Could not generate explanation due to API error."
def analyze_explanation_for_query(explanation: str, original_query: str, columns: str) -> str:
messages = [
{
"role": "system",
"content": (
"You are an academic and literature research assistant."
),
},
{
"role": "user",
"content": (
f"Dataset columns: {columns}\n"
f"Chart explanation: {explanation}\n"
f"Original user query: {original_query}\n\n"
"Craft precise, a domain-specific query (5–12 words) for Semantic Scholar "
"based on the dataset columns, chart explanation, and user query. "
"Generate a concise academic search query specific to this dataset."
"If dataset columns contain abbreviations or symbols (e.g., APPL), expand them into their real-world entity names (e.g., Apple). "
"Queries must incorporate relevant dataset entities (company names, regions, populations, etc.) "
"Focus on methods, variables, and domain terminology of dataset. "
"Target literature that can support or challenge the observed trend. "
"Do not include chart names or explanations. "
"your response should be a concise academic search query only, no reasoning and explanation! ."
),
},
]
try:
response = deepseek_client.chat.completions.create(
model="deepseek/deepseek-chat-v3.1",
messages=messages,
temperature=0.1,
)
return response.choices[0].message.content.strip()
except Exception as e:
st.error(f"Error generating academic query: {e}")
return original_query
def search_semantic_scholar(search_query: str, retries: int = 3, backoff: int = 5):
"""Search Semantic Scholar API for the most relevant papers with retries & exponential backoff."""
url = "https://api.semanticscholar.org/graph/v1/paper/search"
params = {
"query": search_query,
"limit": 5,
"sort": "relevance",
"fields": "title,authors,year,abstract,citationCount,url"
}
headers = {"x-api-key": SEMANTIC_SCHOLAR_API_KEY}
for attempt in range(retries):
try:
response = requests.get(url, params=params, headers=headers, timeout=10)
if response.status_code == 429:
wait_time = backoff * (2 ** attempt)
st.warning(f"Semantic Scholar rate limit hit (429). Retrying in {wait_time} seconds...")
time.sleep(wait_time)
continue
response.raise_for_status()
data = response.json()
if "data" in data and data["data"]:
papers = []
for paper in data["data"]:
papers.append({
"Title": paper.get("title", "N/A"),
"Authors": ", ".join([a.get("name", "N/A") for a in paper.get("authors", [])]),
"Year": paper.get("year", "N/A"),
"Citations": paper.get("citationCount", 0),
"Abstract Snippet": paper.get("abstract", "N/A")[:200] + "..." if paper.get("abstract") else "N/A",
"URL": paper.get("url", "N/A")
})
return pd.DataFrame(papers)
else:
return pd.DataFrame({"Message": ["No relevant papers found."]})
except Exception as e:
st.error(f"Semantic Scholar API error: {e}")
if attempt == retries - 1:
return pd.DataFrame({"Message": ["Search failed."]})
time.sleep(backoff)
return pd.DataFrame({"Message": ["Failed after retries due to rate limits."]})
def interpret_graph_with_qwen(image_base64: str, question: str) -> str:
"""Interpret a graph image"""
try:
messages = [
{
"role": "user",
"content": [
{
"type": "text",
"text": f"Please analyze this graph/chart image and answer the following question: {question}\n\nProvide a concise and precise interpretation including trends, patterns, key insights, and any notable observations. If you are not clear about the graph/chart, say so and ask the user to provide more information.Answers should be no more than 6 to 7 lines."
},
{
"type": "image_url",
"image_url": {
"url": f"data:image/jpeg;base64,{image_base64}"
}
}
]
}
]
response = qwen_client.chat.completions.create(
model="qwen/qwen3-vl-32b-instruct",
messages=messages,
max_tokens=1000,
temperature=0.3
)
return response.choices[0].message.content.strip()
except Exception as e:
return f"Error interpreting graph with qwen: {str(e)}"
def create_sample_data_from_drive(sample_choice: str) -> pd.DataFrame:
"""Load dataset from Google Drive for given sample_choice."""
try:
if sample_choice == "Population Data":
# Use the share link you provided
share_url = "https://drive.google.com/file/d/1NttDmesL-sInl27PvYlFZRCUZNgpJQL8/view?usp=sharing"
file_id = share_url.split("/")[-2]
download_url = f"https://drive.google.com/uc?export=download&id={file_id}"
df = pd.read_csv(download_url)
return df
if sample_choice == "Immigration Data":
# Use the share link you provided
share_url = "https://drive.google.com/file/d/1DB9107DwAphw9cfTnkO6LxqbHH6SgGzy/view?usp=drive_link"
file_id = share_url.split("/")[-2]
download_url = f"https://drive.google.com/uc?export=download&id={file_id}"
df = pd.read_csv(download_url)
return df
except Exception as e:
st.error(f"⚠️ Error creating sample data: {e}")
return None
if 'df' not in st.session_state:
st.session_state.df = None
if 'fig' not in st.session_state:
st.session_state.fig = None
# Show intro only if no dataset is loaded
if st.session_state.df is None:
st.markdown("""
What is BioVix?
BioVix is an innovative web-based tool that integrates data visualization, AI-powered interpretation, and relevant literature search—all from a single query. It allows users to visualize their uploaded data and discover pertinent research articles seamlessly. Additionally, BioVix supports both conversational and uploaded graph interpretation to enhance users’ understanding of their data trends and patterns.
""", unsafe_allow_html=True)
st.header("Overview")
with st.container():
# Centering using columns
col1, col2, col3 = st.columns([1, 2, 1])
with col2:
st.image("overview.jpg", width=900)
with st.sidebar:
st.markdown("""
File Processing Error
Unable to read the uploaded file: {str(e)}
Please ensure your file is in the correct format (CSV, TSV, or Excel).
""", unsafe_allow_html=True)
st.session_state.df = None
elif sample_choice != "None" and uploaded_file is None:
current_df = create_sample_data_from_drive(sample_choice)
if current_df is not None:
st.session_state.df = current_df
st.sidebar.success(f"Loaded {sample_choice} dataset.")
else:
st.sidebar.error("Failed to create sample dataset.")
if st.session_state.df is not None:
df = st.session_state.df
rows, columns = df.shape
col1, col2, col3, col4 = st.columns(4)
with col1:
st.markdown(f"""