Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,6 +1,10 @@
|
|
1 |
# ------------------------------
|
2 |
-
# Imports &
|
3 |
# ------------------------------
|
|
|
|
|
|
|
|
|
4 |
from langchain_openai import OpenAIEmbeddings
|
5 |
from langchain_community.vectorstores import Chroma
|
6 |
from langchain_core.messages import HumanMessage, AIMessage, ToolMessage
|
@@ -13,7 +17,6 @@ from typing import Sequence, Dict, List, Optional, Any
|
|
13 |
import chromadb
|
14 |
import re
|
15 |
import os
|
16 |
-
import streamlit as st
|
17 |
import requests
|
18 |
import hashlib
|
19 |
import json
|
@@ -64,9 +67,9 @@ Format: Markdown with LaTeX mathematical notation where applicable
|
|
64 |
# Validate API key configuration
|
65 |
if not ResearchConfig.DEEPSEEK_API_KEY:
|
66 |
st.error("""**Research Portal Configuration Required**
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
st.stop()
|
71 |
|
72 |
# ------------------------------
|
@@ -87,7 +90,7 @@ class QuantumDocumentManager:
|
|
87 |
separators=["\n\n", "\n", "|||"]
|
88 |
)
|
89 |
docs = splitter.create_documents(documents)
|
90 |
-
#
|
91 |
st.write(f"Created {len(docs)} chunks for collection '{collection_name}'")
|
92 |
return Chroma.from_documents(
|
93 |
documents=docs,
|
@@ -202,7 +205,7 @@ class CognitiveProcessor:
|
|
202 |
valid = [r for r in results if "error" not in r]
|
203 |
if not valid:
|
204 |
return {"error": "All API requests failed"}
|
205 |
-
# Choose the result with the longest content
|
206 |
return max(valid, key=lambda x: len(x.get('choices', [{}])[0].get('message', {}).get('content', '')))
|
207 |
|
208 |
# ------------------------------
|
@@ -251,7 +254,6 @@ class ResearchWorkflow:
|
|
251 |
try:
|
252 |
query = state["context"]["raw_query"]
|
253 |
docs = retriever.retrieve(query, "research")
|
254 |
-
# Log the retrieval result for debugging
|
255 |
st.write(f"[DEBUG] Retrieved {len(docs)} documents from retrieval node.")
|
256 |
return {
|
257 |
"messages": [AIMessage(content=f"Retrieved {len(docs)} documents")],
|
@@ -265,11 +267,11 @@ class ResearchWorkflow:
|
|
265 |
|
266 |
def analyze_content(self, state: AgentState) -> Dict:
|
267 |
try:
|
268 |
-
# Ensure documents are present before proceeding
|
269 |
if "documents" not in state["context"] or not state["context"]["documents"]:
|
270 |
return self._error_state("No documents retrieved; please check your query or retrieval process.")
|
271 |
|
272 |
-
# Concatenate all document content for analysis
|
273 |
docs = "\n\n".join([d.page_content for d in state["context"]["documents"] if hasattr(d, "page_content")])
|
274 |
st.write(f"[DEBUG] Analyzing content from {len(state['context']['documents'])} documents.")
|
275 |
prompt = ResearchConfig.ANALYSIS_TEMPLATE.format(context=docs)
|
@@ -320,7 +322,6 @@ Improve:
|
|
320 |
|
321 |
def _quality_check(self, state: AgentState) -> str:
|
322 |
content = state["messages"][-1].content
|
323 |
-
# Check for the keyword "VALID" in the output; if missing, trigger refinement
|
324 |
return "valid" if "VALID" in content else "invalid"
|
325 |
|
326 |
def _error_state(self, message: str) -> Dict:
|
@@ -337,14 +338,7 @@ Improve:
|
|
337 |
class ResearchInterface:
|
338 |
def __init__(self):
|
339 |
self.workflow = ResearchWorkflow()
|
340 |
-
|
341 |
-
|
342 |
-
def _initialize_interface(self):
|
343 |
-
st.set_page_config(
|
344 |
-
page_title="NeuroResearch AI",
|
345 |
-
layout="wide",
|
346 |
-
initial_sidebar_state="expanded"
|
347 |
-
)
|
348 |
self._inject_styles()
|
349 |
self._build_sidebar()
|
350 |
self._build_main_interface()
|
@@ -410,7 +404,7 @@ class ResearchInterface:
|
|
410 |
def _build_main_interface(self):
|
411 |
st.title("🧠 NeuroResearch AI")
|
412 |
query = st.text_area("Research Query:", height=200,
|
413 |
-
|
414 |
|
415 |
if st.button("Execute Analysis", type="primary"):
|
416 |
self._execute_analysis(query)
|
@@ -421,10 +415,8 @@ class ResearchInterface:
|
|
421 |
results = self.workflow.app.stream(
|
422 |
{"messages": [HumanMessage(content=query)], "context": {}, "metadata": {}}
|
423 |
)
|
424 |
-
|
425 |
for event in results:
|
426 |
self._render_event(event)
|
427 |
-
|
428 |
st.success("✅ Analysis Completed Successfully")
|
429 |
except Exception as e:
|
430 |
st.error(f"""**Analysis Failed**
|
@@ -438,7 +430,6 @@ Potential issues:
|
|
438 |
if 'ingest' in event:
|
439 |
with st.container():
|
440 |
st.success("✅ Query Ingested")
|
441 |
-
|
442 |
elif 'retrieve' in event:
|
443 |
with st.container():
|
444 |
docs = event['retrieve']['context']['documents']
|
@@ -447,13 +438,11 @@ Potential issues:
|
|
447 |
for i, doc in enumerate(docs, 1):
|
448 |
st.markdown(f"**Document {i}**")
|
449 |
st.code(doc.page_content, language='text')
|
450 |
-
|
451 |
elif 'analyze' in event:
|
452 |
with st.container():
|
453 |
content = event['analyze']['messages'][0].content
|
454 |
with st.expander("Technical Analysis Report", expanded=True):
|
455 |
st.markdown(content)
|
456 |
-
|
457 |
elif 'validate' in event:
|
458 |
with st.container():
|
459 |
content = event['validate']['messages'][0].content
|
|
|
1 |
# ------------------------------
|
2 |
+
# Imports & Initial Configuration
|
3 |
# ------------------------------
|
4 |
+
import streamlit as st
|
5 |
+
# Set the page configuration immediately—this must be the first Streamlit command.
|
6 |
+
st.set_page_config(page_title="NeuroResearch AI", layout="wide", initial_sidebar_state="expanded")
|
7 |
+
|
8 |
from langchain_openai import OpenAIEmbeddings
|
9 |
from langchain_community.vectorstores import Chroma
|
10 |
from langchain_core.messages import HumanMessage, AIMessage, ToolMessage
|
|
|
17 |
import chromadb
|
18 |
import re
|
19 |
import os
|
|
|
20 |
import requests
|
21 |
import hashlib
|
22 |
import json
|
|
|
67 |
# Validate API key configuration
|
68 |
if not ResearchConfig.DEEPSEEK_API_KEY:
|
69 |
st.error("""**Research Portal Configuration Required**
|
70 |
+
1. Obtain DeepSeek API key: [platform.deepseek.com](https://platform.deepseek.com/)
|
71 |
+
2. Configure secret: `DEEPSEEK_API_KEY` in Space settings
|
72 |
+
3. Rebuild deployment""")
|
73 |
st.stop()
|
74 |
|
75 |
# ------------------------------
|
|
|
90 |
separators=["\n\n", "\n", "|||"]
|
91 |
)
|
92 |
docs = splitter.create_documents(documents)
|
93 |
+
# Debug: log the number of chunks created for the collection.
|
94 |
st.write(f"Created {len(docs)} chunks for collection '{collection_name}'")
|
95 |
return Chroma.from_documents(
|
96 |
documents=docs,
|
|
|
205 |
valid = [r for r in results if "error" not in r]
|
206 |
if not valid:
|
207 |
return {"error": "All API requests failed"}
|
208 |
+
# Choose the result with the longest content for robustness.
|
209 |
return max(valid, key=lambda x: len(x.get('choices', [{}])[0].get('message', {}).get('content', '')))
|
210 |
|
211 |
# ------------------------------
|
|
|
254 |
try:
|
255 |
query = state["context"]["raw_query"]
|
256 |
docs = retriever.retrieve(query, "research")
|
|
|
257 |
st.write(f"[DEBUG] Retrieved {len(docs)} documents from retrieval node.")
|
258 |
return {
|
259 |
"messages": [AIMessage(content=f"Retrieved {len(docs)} documents")],
|
|
|
267 |
|
268 |
def analyze_content(self, state: AgentState) -> Dict:
|
269 |
try:
|
270 |
+
# Ensure documents are present before proceeding.
|
271 |
if "documents" not in state["context"] or not state["context"]["documents"]:
|
272 |
return self._error_state("No documents retrieved; please check your query or retrieval process.")
|
273 |
|
274 |
+
# Concatenate all document content for analysis.
|
275 |
docs = "\n\n".join([d.page_content for d in state["context"]["documents"] if hasattr(d, "page_content")])
|
276 |
st.write(f"[DEBUG] Analyzing content from {len(state['context']['documents'])} documents.")
|
277 |
prompt = ResearchConfig.ANALYSIS_TEMPLATE.format(context=docs)
|
|
|
322 |
|
323 |
def _quality_check(self, state: AgentState) -> str:
|
324 |
content = state["messages"][-1].content
|
|
|
325 |
return "valid" if "VALID" in content else "invalid"
|
326 |
|
327 |
def _error_state(self, message: str) -> Dict:
|
|
|
338 |
class ResearchInterface:
|
339 |
def __init__(self):
|
340 |
self.workflow = ResearchWorkflow()
|
341 |
+
# Do not call st.set_page_config here because it has already been called at the top.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
342 |
self._inject_styles()
|
343 |
self._build_sidebar()
|
344 |
self._build_main_interface()
|
|
|
404 |
def _build_main_interface(self):
|
405 |
st.title("🧠 NeuroResearch AI")
|
406 |
query = st.text_area("Research Query:", height=200,
|
407 |
+
placeholder="Enter technical research question...")
|
408 |
|
409 |
if st.button("Execute Analysis", type="primary"):
|
410 |
self._execute_analysis(query)
|
|
|
415 |
results = self.workflow.app.stream(
|
416 |
{"messages": [HumanMessage(content=query)], "context": {}, "metadata": {}}
|
417 |
)
|
|
|
418 |
for event in results:
|
419 |
self._render_event(event)
|
|
|
420 |
st.success("✅ Analysis Completed Successfully")
|
421 |
except Exception as e:
|
422 |
st.error(f"""**Analysis Failed**
|
|
|
430 |
if 'ingest' in event:
|
431 |
with st.container():
|
432 |
st.success("✅ Query Ingested")
|
|
|
433 |
elif 'retrieve' in event:
|
434 |
with st.container():
|
435 |
docs = event['retrieve']['context']['documents']
|
|
|
438 |
for i, doc in enumerate(docs, 1):
|
439 |
st.markdown(f"**Document {i}**")
|
440 |
st.code(doc.page_content, language='text')
|
|
|
441 |
elif 'analyze' in event:
|
442 |
with st.container():
|
443 |
content = event['analyze']['messages'][0].content
|
444 |
with st.expander("Technical Analysis Report", expanded=True):
|
445 |
st.markdown(content)
|
|
|
446 |
elif 'validate' in event:
|
447 |
with st.container():
|
448 |
content = event['validate']['messages'][0].content
|