Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -35,33 +35,38 @@ def call_llm(messages, model="deepseek/deepseek-chat-v3-0324:free", max_tokens=3
|
|
| 35 |
|
| 36 |
response = requests.post(url, headers=headers, json=data)
|
| 37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
if response.status_code != 200:
|
| 39 |
-
|
| 40 |
-
error_msg = response.json().get("error", {}).get("message", "Unknown error from LLM API.")
|
| 41 |
-
except Exception:
|
| 42 |
-
error_msg = response.text
|
| 43 |
raise RuntimeError(f"LLM API error: {error_msg}")
|
| 44 |
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
raise RuntimeError("LLM API returned unexpected response format. Missing 'choices'.")
|
| 50 |
|
| 51 |
def get_sources(topic, domains=None):
|
| 52 |
query = topic
|
| 53 |
if domains:
|
| 54 |
domain_filters = [d.strip() for d in domains.split(",") if d.strip()]
|
| 55 |
query += " site:" + " OR site:".join(domain_filters)
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
|
| 66 |
def get_arxiv_papers(query):
|
| 67 |
from urllib.parse import quote_plus
|
|
@@ -130,8 +135,8 @@ def generate_latex(text):
|
|
| 130 |
def generate_download_button(file, label, mime_type):
|
| 131 |
b64 = base64.b64encode(file.read()).decode()
|
| 132 |
return f"""
|
| 133 |
-
<a href
|
| 134 |
-
|
| 135 |
</a>
|
| 136 |
"""
|
| 137 |
|
|
@@ -143,27 +148,27 @@ def fetch_related_image(topic):
|
|
| 143 |
st.set_page_config("Deep Research Bot", layout="wide")
|
| 144 |
|
| 145 |
with st.sidebar:
|
| 146 |
-
st.title("
|
| 147 |
-
topic = st.text_input("
|
| 148 |
-
report_type = st.selectbox("
|
| 149 |
"Summary - Short and fast (~2 min)",
|
| 150 |
"Detailed Report (~5 min)",
|
| 151 |
"Thorough Academic Research (~10 min)"
|
| 152 |
])
|
| 153 |
-
tone = st.selectbox("
|
| 154 |
"Objective - Impartial and unbiased presentation of facts and findings",
|
| 155 |
"Persuasive - Advocating a specific point of view",
|
| 156 |
"Narrative - Storytelling tone for layperson readers"
|
| 157 |
])
|
| 158 |
-
source_type = st.selectbox("
|
| 159 |
-
custom_domains = st.text_input("
|
| 160 |
-
research_button = st.button("
|
| 161 |
|
| 162 |
-
st.title("
|
| 163 |
|
| 164 |
if research_button and topic:
|
| 165 |
try:
|
| 166 |
-
with st.status("
|
| 167 |
st.info("Fetching from sources...")
|
| 168 |
|
| 169 |
all_sources = []
|
|
@@ -188,7 +193,7 @@ if research_button and topic:
|
|
| 188 |
combined_text += f"- [{m['title']}]({m['url']})\n> {m.get('snippet', m.get('summary', ''))[:300]}...\n\n"
|
| 189 |
citations.append(generate_apa_citation(m['title'], m['url'], m['source']))
|
| 190 |
|
| 191 |
-
with st.spinner("
|
| 192 |
prompt = f"""
|
| 193 |
# Research Topic: {topic}
|
| 194 |
Tone: {tone}
|
|
@@ -204,19 +209,19 @@ Write the report in academic markdown with paragraphs (use bullet points only wh
|
|
| 204 |
"""
|
| 205 |
final_output = call_llm([{"role": "user", "content": prompt}])
|
| 206 |
|
| 207 |
-
st.markdown(f"###
|
| 208 |
st.markdown(final_output, unsafe_allow_html=True)
|
| 209 |
|
| 210 |
if report_type == "Thorough Academic Research (~10 min)":
|
| 211 |
image_url = fetch_related_image(topic)
|
| 212 |
st.image(image_url, caption="Related Image", use_column_width=True)
|
| 213 |
|
| 214 |
-
st.markdown("###
|
| 215 |
for cite in citations:
|
| 216 |
st.markdown(f"- {cite}")
|
| 217 |
|
| 218 |
if report_type == "Thorough Academic Research (~10 min)":
|
| 219 |
-
with st.spinner("
|
| 220 |
pdf_file = generate_pdf(final_output)
|
| 221 |
latex_file = generate_latex(final_output)
|
| 222 |
st.markdown(generate_download_button(pdf_file, "Research_Report.pdf", "application/pdf"), unsafe_allow_html=True)
|
|
@@ -224,11 +229,11 @@ Write the report in academic markdown with paragraphs (use bullet points only wh
|
|
| 224 |
|
| 225 |
overlaps = check_plagiarism(final_output, topic)
|
| 226 |
if overlaps:
|
| 227 |
-
st.warning("
|
| 228 |
for hit in overlaps:
|
| 229 |
st.markdown(f"- [{hit['title']}]({hit['url']})")
|
| 230 |
else:
|
| 231 |
-
st.success("
|
| 232 |
|
| 233 |
except Exception as e:
|
| 234 |
-
st.error(f"Error: {e}")
|
|
|
|
| 35 |
|
| 36 |
response = requests.post(url, headers=headers, json=data)
|
| 37 |
|
| 38 |
+
try:
|
| 39 |
+
result = response.json()
|
| 40 |
+
except Exception:
|
| 41 |
+
raise RuntimeError(f"LLM API returned invalid JSON: {response.text}")
|
| 42 |
+
|
| 43 |
if response.status_code != 200:
|
| 44 |
+
error_msg = result.get("error", {}).get("message", "Unknown error from LLM API.")
|
|
|
|
|
|
|
|
|
|
| 45 |
raise RuntimeError(f"LLM API error: {error_msg}")
|
| 46 |
|
| 47 |
+
if "choices" not in result:
|
| 48 |
+
raise RuntimeError(f"LLM API returned unexpected response format. Full response:\n{result}")
|
| 49 |
+
|
| 50 |
+
return result["choices"][0]["message"]["content"]
|
|
|
|
| 51 |
|
| 52 |
def get_sources(topic, domains=None):
|
| 53 |
query = topic
|
| 54 |
if domains:
|
| 55 |
domain_filters = [d.strip() for d in domains.split(",") if d.strip()]
|
| 56 |
query += " site:" + " OR site:".join(domain_filters)
|
| 57 |
+
try:
|
| 58 |
+
response = tavily.search(query=query, search_depth="advanced", max_results=10)
|
| 59 |
+
sources = []
|
| 60 |
+
for item in response.get("results", []):
|
| 61 |
+
sources.append({
|
| 62 |
+
"title": item.get("title"),
|
| 63 |
+
"url": item.get("url"),
|
| 64 |
+
"snippet": item.get("content", "")
|
| 65 |
+
})
|
| 66 |
+
return sources
|
| 67 |
+
except Exception as e:
|
| 68 |
+
st.error(f"Error retrieving sources: {e}")
|
| 69 |
+
return []
|
| 70 |
|
| 71 |
def get_arxiv_papers(query):
|
| 72 |
from urllib.parse import quote_plus
|
|
|
|
| 135 |
def generate_download_button(file, label, mime_type):
|
| 136 |
b64 = base64.b64encode(file.read()).decode()
|
| 137 |
return f"""
|
| 138 |
+
<a href=\"data:{mime_type};base64,{b64}\" download=\"{label}\">
|
| 139 |
+
\ud83d\udce5 Download {label}
|
| 140 |
</a>
|
| 141 |
"""
|
| 142 |
|
|
|
|
| 148 |
st.set_page_config("Deep Research Bot", layout="wide")
|
| 149 |
|
| 150 |
with st.sidebar:
|
| 151 |
+
st.title("\ud83e\udde0 Deep Research Assistant")
|
| 152 |
+
topic = st.text_input("\ud83d\udca1 Topic to research")
|
| 153 |
+
report_type = st.selectbox("\ud83d\udcc4 Type of report", [
|
| 154 |
"Summary - Short and fast (~2 min)",
|
| 155 |
"Detailed Report (~5 min)",
|
| 156 |
"Thorough Academic Research (~10 min)"
|
| 157 |
])
|
| 158 |
+
tone = st.selectbox("\ud83c\udfaf Tone of the report", [
|
| 159 |
"Objective - Impartial and unbiased presentation of facts and findings",
|
| 160 |
"Persuasive - Advocating a specific point of view",
|
| 161 |
"Narrative - Storytelling tone for layperson readers"
|
| 162 |
])
|
| 163 |
+
source_type = st.selectbox("\ud83c\udf10 Sources to include", ["Web Only", "Academic Only", "Hybrid"])
|
| 164 |
+
custom_domains = st.text_input("\ud83d\udd0d Query Domains (Optional)", placeholder="techcrunch.com, forbes.com")
|
| 165 |
+
research_button = st.button("\ud83d\ude80 Start Research")
|
| 166 |
|
| 167 |
+
st.title("\ud83d\udcc1 Research Output")
|
| 168 |
|
| 169 |
if research_button and topic:
|
| 170 |
try:
|
| 171 |
+
with st.status("\ud83d\udd0d Gathering data..."):
|
| 172 |
st.info("Fetching from sources...")
|
| 173 |
|
| 174 |
all_sources = []
|
|
|
|
| 193 |
combined_text += f"- [{m['title']}]({m['url']})\n> {m.get('snippet', m.get('summary', ''))[:300]}...\n\n"
|
| 194 |
citations.append(generate_apa_citation(m['title'], m['url'], m['source']))
|
| 195 |
|
| 196 |
+
with st.spinner("\u270d\ufe0f Synthesizing report..."):
|
| 197 |
prompt = f"""
|
| 198 |
# Research Topic: {topic}
|
| 199 |
Tone: {tone}
|
|
|
|
| 209 |
"""
|
| 210 |
final_output = call_llm([{"role": "user", "content": prompt}])
|
| 211 |
|
| 212 |
+
st.markdown(f"### \ud83d\udcc4 {report_type}")
|
| 213 |
st.markdown(final_output, unsafe_allow_html=True)
|
| 214 |
|
| 215 |
if report_type == "Thorough Academic Research (~10 min)":
|
| 216 |
image_url = fetch_related_image(topic)
|
| 217 |
st.image(image_url, caption="Related Image", use_column_width=True)
|
| 218 |
|
| 219 |
+
st.markdown("### \ud83d\udcda Citations (APA Format)")
|
| 220 |
for cite in citations:
|
| 221 |
st.markdown(f"- {cite}")
|
| 222 |
|
| 223 |
if report_type == "Thorough Academic Research (~10 min)":
|
| 224 |
+
with st.spinner("\ud83d\udce6 Preparing PDF and LaTeX..."):
|
| 225 |
pdf_file = generate_pdf(final_output)
|
| 226 |
latex_file = generate_latex(final_output)
|
| 227 |
st.markdown(generate_download_button(pdf_file, "Research_Report.pdf", "application/pdf"), unsafe_allow_html=True)
|
|
|
|
| 229 |
|
| 230 |
overlaps = check_plagiarism(final_output, topic)
|
| 231 |
if overlaps:
|
| 232 |
+
st.warning("\u26a0\ufe0f Potential overlaps detected:")
|
| 233 |
for hit in overlaps:
|
| 234 |
st.markdown(f"- [{hit['title']}]({hit['url']})")
|
| 235 |
else:
|
| 236 |
+
st.success("\u2705 No major overlaps found.")
|
| 237 |
|
| 238 |
except Exception as e:
|
| 239 |
+
st.error(f"Error: {e}")
|