import asyncio
import json
from typing import Any, Dict, List
from app.services.connector_service import ConnectorService
from langchain_core.messages import HumanMessage, SystemMessage
from langchain_core.runnables import RunnableConfig
from sqlalchemy.ext.asyncio import AsyncSession
from .configuration import Configuration, SearchMode
from .prompts import get_answer_outline_system_prompt, get_further_questions_system_prompt
from .state import State
from .sub_section_writer.graph import graph as sub_section_writer_graph
from .sub_section_writer.configuration import SubSectionType
from .qna_agent.graph import graph as qna_agent_graph
from .utils import AnswerOutline, get_connector_emoji, get_connector_friendly_name
from app.services.query_service import QueryService
from langgraph.types import StreamWriter
# Additional imports for document fetching
from sqlalchemy.future import select
from app.db import Document, SearchSpace
async def fetch_documents_by_ids(
document_ids: List[int],
user_id: str,
db_session: AsyncSession
) -> tuple[List[Dict[str, Any]], List[Dict[str, Any]]]:
"""
Fetch documents by their IDs with ownership check using DOCUMENTS mode approach.
This function ensures that only documents belonging to the user are fetched,
providing security by checking ownership through SearchSpace association.
Similar to SearchMode.DOCUMENTS, it fetches full documents and concatenates their chunks.
Also creates source objects for UI display, grouped by document type.
Args:
document_ids: List of document IDs to fetch
user_id: The user ID to check ownership
db_session: The database session
Returns:
Tuple of (source_objects, document_chunks) - similar to ConnectorService pattern
"""
if not document_ids:
return [], []
try:
# Query documents with ownership check
result = await db_session.execute(
select(Document)
.join(SearchSpace)
.filter(
Document.id.in_(document_ids),
SearchSpace.user_id == user_id
)
)
documents = result.scalars().all()
# Group documents by type for source object creation
documents_by_type = {}
formatted_documents = []
for doc in documents:
# Fetch associated chunks for this document (similar to DocumentHybridSearchRetriever)
from app.db import Chunk
chunks_query = select(Chunk).where(Chunk.document_id == doc.id).order_by(Chunk.id)
chunks_result = await db_session.execute(chunks_query)
chunks = chunks_result.scalars().all()
# Concatenate chunks content (similar to SearchMode.DOCUMENTS approach)
concatenated_chunks_content = " ".join([chunk.content for chunk in chunks]) if chunks else doc.content
# Format to match connector service return format
formatted_doc = {
"chunk_id": f"user_doc_{doc.id}",
"content": concatenated_chunks_content, # Use concatenated content like DOCUMENTS mode
"score": 0.5, # High score since user explicitly selected these
"document": {
"id": doc.id,
"title": doc.title,
"document_type": doc.document_type.value if doc.document_type else "UNKNOWN",
"metadata": doc.document_metadata or {},
},
"source": doc.document_type.value if doc.document_type else "UNKNOWN"
}
formatted_documents.append(formatted_doc)
# Group by document type for source objects
doc_type = doc.document_type.value if doc.document_type else "UNKNOWN"
if doc_type not in documents_by_type:
documents_by_type[doc_type] = []
documents_by_type[doc_type].append(doc)
# Create source objects for each document type (similar to ConnectorService)
source_objects = []
connector_id_counter = 100 # Start from 100 to avoid conflicts with regular connectors
for doc_type, docs in documents_by_type.items():
sources_list = []
for doc in docs:
metadata = doc.document_metadata or {}
# Create type-specific source formatting (similar to ConnectorService)
if doc_type == "LINEAR_CONNECTOR":
# Extract Linear-specific metadata
issue_identifier = metadata.get('issue_identifier', '')
issue_title = metadata.get('issue_title', doc.title)
issue_state = metadata.get('state', '')
comment_count = metadata.get('comment_count', 0)
# Create a more descriptive title for Linear issues
title = f"Linear: {issue_identifier} - {issue_title}" if issue_identifier else f"Linear: {issue_title}"
if issue_state:
title += f" ({issue_state})"
# Create description
description = doc.content[:100] + "..." if len(doc.content) > 100 else doc.content
if comment_count:
description += f" | Comments: {comment_count}"
# Create URL
url = f"https://linear.app/issue/{issue_identifier}" if issue_identifier else ""
elif doc_type == "SLACK_CONNECTOR":
# Extract Slack-specific metadata
channel_name = metadata.get('channel_name', 'Unknown Channel')
channel_id = metadata.get('channel_id', '')
message_date = metadata.get('start_date', '')
title = f"Slack: {channel_name}"
if message_date:
title += f" ({message_date})"
description = doc.content[:100] + "..." if len(doc.content) > 100 else doc.content
url = f"https://slack.com/app_redirect?channel={channel_id}" if channel_id else ""
elif doc_type == "NOTION_CONNECTOR":
# Extract Notion-specific metadata
page_title = metadata.get('page_title', doc.title)
page_id = metadata.get('page_id', '')
title = f"Notion: {page_title}"
description = doc.content[:100] + "..." if len(doc.content) > 100 else doc.content
url = f"https://notion.so/{page_id.replace('-', '')}" if page_id else ""
elif doc_type == "GITHUB_CONNECTOR":
title = f"GitHub: {doc.title}"
description = metadata.get('description', doc.content[:100] + "..." if len(doc.content) > 100 else doc.content)
url = metadata.get('url', '')
elif doc_type == "YOUTUBE_VIDEO":
# Extract YouTube-specific metadata
video_title = metadata.get('video_title', doc.title)
video_id = metadata.get('video_id', '')
channel_name = metadata.get('channel_name', '')
title = video_title
if channel_name:
title += f" - {channel_name}"
description = metadata.get('description', doc.content[:100] + "..." if len(doc.content) > 100 else doc.content)
url = f"https://www.youtube.com/watch?v={video_id}" if video_id else ""
elif doc_type == "DISCORD_CONNECTOR":
# Extract Discord-specific metadata
channel_name = metadata.get('channel_name', 'Unknown Channel')
channel_id = metadata.get('channel_id', '')
guild_id = metadata.get('guild_id', '')
message_date = metadata.get('start_date', '')
title = f"Discord: {channel_name}"
if message_date:
title += f" ({message_date})"
description = doc.content[:100] + "..." if len(doc.content) > 100 else doc.content
if guild_id and channel_id:
url = f"https://discord.com/channels/{guild_id}/{channel_id}"
elif channel_id:
url = f"https://discord.com/channels/@me/{channel_id}"
else:
url = ""
elif doc_type == "JIRA_CONNECTOR":
# Extract Jira-specific metadata
issue_key = metadata.get('issue_key', 'Unknown Issue')
issue_title = metadata.get('issue_title', 'Untitled Issue')
status = metadata.get('status', '')
priority = metadata.get('priority', '')
issue_type = metadata.get('issue_type', '')
title = f"Jira: {issue_key} - {issue_title}"
if status:
title += f" ({status})"
description = doc.content[:100] + "..." if len(doc.content) > 100 else doc.content
# Construct Jira URL if we have the base URL
base_url = metadata.get('base_url', '')
if base_url and issue_key:
url = f"{base_url}/browse/{issue_key}"
else:
url = ""
elif doc_type == "EXTENSION":
# Extract Extension-specific metadata
webpage_title = metadata.get('VisitedWebPageTitle', doc.title)
webpage_url = metadata.get('VisitedWebPageURL', '')
visit_date = metadata.get('VisitedWebPageDateWithTimeInISOString', '')
title = webpage_title
if visit_date:
formatted_date = visit_date.split('T')[0] if 'T' in visit_date else visit_date
title += f" (visited: {formatted_date})"
description = doc.content[:100] + "..." if len(doc.content) > 100 else doc.content
url = webpage_url
elif doc_type == "CRAWLED_URL":
title = doc.title
description = metadata.get('og:description', metadata.get('ogDescription', doc.content[:100] + "..." if len(doc.content) > 100 else doc.content))
url = metadata.get('url', '')
else: # FILE and other types
title = doc.title
description = doc.content[:100] + "..." if len(doc.content) > 100 else doc.content
url = metadata.get('url', '')
# Create source entry
source = {
"id": doc.id,
"title": title,
"description": description,
"url": url
}
sources_list.append(source)
# Create source object for this document type
friendly_type_names = {
"LINEAR_CONNECTOR": "Linear Issues (Selected)",
"SLACK_CONNECTOR": "Slack (Selected)",
"NOTION_CONNECTOR": "Notion (Selected)",
"GITHUB_CONNECTOR": "GitHub (Selected)",
"YOUTUBE_VIDEO": "YouTube Videos (Selected)",
"DISCORD_CONNECTOR": "Discord (Selected)",
"JIRA_CONNECTOR": "Jira Issues (Selected)",
"EXTENSION": "Browser Extension (Selected)",
"CRAWLED_URL": "Web Pages (Selected)",
"FILE": "Files (Selected)"
}
source_object = {
"id": connector_id_counter,
"name": friendly_type_names.get(doc_type, f"{doc_type} (Selected)"),
"type": f"USER_SELECTED_{doc_type}",
"sources": sources_list,
}
source_objects.append(source_object)
connector_id_counter += 1
print(f"Fetched {len(formatted_documents)} user-selected documents (with concatenated chunks) from {len(document_ids)} requested IDs")
print(f"Created {len(source_objects)} source objects for UI display")
return source_objects, formatted_documents
except Exception as e:
print(f"Error fetching documents by IDs: {str(e)}")
return [], []
async def write_answer_outline(state: State, config: RunnableConfig, writer: StreamWriter) -> Dict[str, Any]:
"""
Create a structured answer outline based on the user query.
This node takes the user query and number of sections from the configuration and uses
an LLM to generate a comprehensive outline with logical sections and research questions
for each section.
Returns:
Dict containing the answer outline in the "answer_outline" key for state update.
"""
from app.services.llm_service import get_user_strategic_llm
from app.db import get_async_session
streaming_service = state.streaming_service
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ Generating answer outline..."
)
}
)
# Get configuration from runnable config
configuration = Configuration.from_runnable_config(config)
reformulated_query = state.reformulated_query
user_query = configuration.user_query
num_sections = configuration.num_sections
user_id = configuration.user_id
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f'๐ค Planning research approach for: "{user_query[:100]}..."'
)
}
)
# Get user's strategic LLM
llm = await get_user_strategic_llm(state.db_session, user_id)
if not llm:
error_message = f"No strategic LLM configured for user {user_id}"
writer({"yield_value": streaming_service.format_error(error_message)})
raise RuntimeError(error_message)
# Create the human message content
human_message_content = f"""
Now Please create an answer outline for the following query:
User Query: {reformulated_query}
Number of Sections: {num_sections}
Remember to format your response as valid JSON exactly matching this structure:
{{
"answer_outline": [
{{
"section_id": 0,
"section_title": "Section Title",
"questions": [
"Question 1 to research for this section",
"Question 2 to research for this section"
]
}}
]
}}
Your output MUST be valid JSON in exactly this format. Do not include any other text or explanation.
"""
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ Designing structured outline with AI..."
)
}
)
# Create messages for the LLM
messages = [
SystemMessage(content=get_answer_outline_system_prompt()),
HumanMessage(content=human_message_content)
]
# Call the LLM directly without using structured output
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"โ๏ธ Processing answer structure..."
)
}
)
response = await llm.ainvoke(messages)
# Parse the JSON response manually
try:
# Extract JSON content from the response
content = response.content
# Find the JSON in the content (handle case where LLM might add additional text)
json_start = content.find('{')
json_end = content.rfind('}') + 1
if json_start >= 0 and json_end > json_start:
json_str = content[json_start:json_end]
# Parse the JSON string
parsed_data = json.loads(json_str)
# Convert to Pydantic model
answer_outline = AnswerOutline(**parsed_data)
total_questions = sum(len(section.questions) for section in answer_outline.answer_outline)
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"โ
Successfully generated outline with {len(answer_outline.answer_outline)} sections and {total_questions} research questions!"
)
}
)
print(
f"Successfully generated answer outline with {len(answer_outline.answer_outline)} sections"
)
# Return state update
return {"answer_outline": answer_outline}
else:
# If JSON structure not found, raise a clear error
error_message = (
f"Could not find valid JSON in LLM response. Raw response: {content}"
)
writer({"yield_value": streaming_service.format_error(error_message)})
raise ValueError(error_message)
except (json.JSONDecodeError, ValueError) as e:
# Log the error and re-raise it
error_message = f"Error parsing LLM response: {str(e)}"
writer({"yield_value": streaming_service.format_error(error_message)})
print(f"Error parsing LLM response: {str(e)}")
print(f"Raw response: {response.content}")
raise
async def fetch_relevant_documents(
research_questions: List[str],
user_id: str,
search_space_id: int,
db_session: AsyncSession,
connectors_to_search: List[str],
writer: StreamWriter = None,
state: State = None,
top_k: int = 10,
connector_service: ConnectorService = None,
search_mode: SearchMode = SearchMode.CHUNKS,
user_selected_sources: List[Dict[str, Any]] = None
) -> List[Dict[str, Any]]:
"""
Fetch relevant documents for research questions using the provided connectors.
This function searches across multiple data sources for information related to the
research questions. It provides user-friendly feedback during the search process by
displaying connector names (like "Web Search" instead of "TAVILY_API") and adding
relevant emojis to indicate the type of source being searched.
Args:
research_questions: List of research questions to find documents for
user_id: The user ID
search_space_id: The search space ID
db_session: The database session
connectors_to_search: List of connectors to search
writer: StreamWriter for sending progress updates
state: The current state containing the streaming service
top_k: Number of top results to retrieve per connector per question
connector_service: An initialized connector service to use for searching
Returns:
List of relevant documents
"""
# Initialize services
# connector_service = ConnectorService(db_session)
# Only use streaming if both writer and state are provided
streaming_service = state.streaming_service if state is not None else None
# Stream initial status update
if streaming_service and writer:
connector_names = [get_connector_friendly_name(connector) for connector in connectors_to_search]
connector_names_str = ", ".join(connector_names)
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Starting research on {len(research_questions)} questions using {connector_names_str} data sources"
)
}
)
all_raw_documents = [] # Store all raw documents
all_sources = [] # Store all sources
for i, user_query in enumerate(research_questions):
# Stream question being researched
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f'๐ง Researching question {i + 1}/{len(research_questions)}: "{user_query[:100]}..."'
)
}
)
# Use original research question as the query
reformulated_query = user_query
# Process each selected connector
for connector in connectors_to_search:
# Stream connector being searched
if streaming_service and writer:
connector_emoji = get_connector_emoji(connector)
friendly_name = get_connector_friendly_name(connector)
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"{connector_emoji} Searching {friendly_name} for relevant information..."
)
}
)
try:
if connector == "YOUTUBE_VIDEO":
source_object, youtube_chunks = await connector_service.search_youtube(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(youtube_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐น Found {len(youtube_chunks)} YouTube chunks related to your query"
)
}
)
elif connector == "EXTENSION":
source_object, extension_chunks = await connector_service.search_extension(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(extension_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐งฉ Found {len(extension_chunks)} Browser Extension chunks related to your query"
)
}
)
elif connector == "CRAWLED_URL":
source_object, crawled_urls_chunks = await connector_service.search_crawled_urls(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(crawled_urls_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Found {len(crawled_urls_chunks)} Web Pages chunks related to your query"
)
}
)
elif connector == "FILE":
source_object, files_chunks = await connector_service.search_files(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(files_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Found {len(files_chunks)} Files chunks related to your query"
)
}
)
elif connector == "SLACK_CONNECTOR":
source_object, slack_chunks = await connector_service.search_slack(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(slack_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ฌ Found {len(slack_chunks)} Slack messages related to your query"
)
}
)
elif connector == "NOTION_CONNECTOR":
source_object, notion_chunks = await connector_service.search_notion(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(notion_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Found {len(notion_chunks)} Notion pages/blocks related to your query"
)
}
)
elif connector == "GITHUB_CONNECTOR":
source_object, github_chunks = await connector_service.search_github(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(github_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Found {len(github_chunks)} GitHub files/issues related to your query"
)
}
)
elif connector == "LINEAR_CONNECTOR":
source_object, linear_chunks = await connector_service.search_linear(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(linear_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Found {len(linear_chunks)} Linear issues related to your query"
)
}
)
elif connector == "TAVILY_API":
source_object, tavily_chunks = await connector_service.search_tavily(
user_query=reformulated_query,
user_id=user_id,
top_k=top_k
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(tavily_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Found {len(tavily_chunks)} Web Search results related to your query"
)
}
)
elif connector == "LINKUP_API":
linkup_mode = "standard"
source_object, linkup_chunks = await connector_service.search_linkup(
user_query=reformulated_query,
user_id=user_id,
mode=linkup_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(linkup_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Found {len(linkup_chunks)} Linkup results related to your query"
)
}
)
elif connector == "DISCORD_CONNECTOR":
source_object, discord_chunks = await connector_service.search_discord(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(discord_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐จ๏ธ Found {len(discord_chunks)} Discord messages related to your query"
)
}
)
elif connector == "JIRA_CONNECTOR":
source_object, jira_chunks = await connector_service.search_jira(
user_query=reformulated_query,
user_id=user_id,
search_space_id=search_space_id,
top_k=top_k,
search_mode=search_mode
)
# Add to sources and raw documents
if source_object:
all_sources.append(source_object)
all_raw_documents.extend(jira_chunks)
# Stream found document count
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ซ Found {len(jira_chunks)} Jira issues related to your query"
)
}
)
except Exception as e:
error_message = f"Error searching connector {connector}: {str(e)}"
print(error_message)
# Stream error message
if streaming_service and writer:
friendly_name = get_connector_friendly_name(connector)
writer(
{
"yield_value": streaming_service.format_error(
f"Error searching {friendly_name}: {str(e)}"
)
}
)
# Continue with other connectors on error
continue
# Deduplicate source objects by ID before streaming
deduplicated_sources = []
seen_source_keys = set()
# First add user-selected sources (if any)
if user_selected_sources:
for source_obj in user_selected_sources:
source_id = source_obj.get('id')
source_type = source_obj.get('type')
if source_id and source_type:
source_key = f"{source_type}_{source_id}"
if source_key not in seen_source_keys:
seen_source_keys.add(source_key)
deduplicated_sources.append(source_obj)
else:
deduplicated_sources.append(source_obj)
# Then add connector sources
for source_obj in all_sources:
# Use combination of source ID and type as a unique identifier
# This ensures we don't accidentally deduplicate sources from different connectors
source_id = source_obj.get('id')
source_type = source_obj.get('type')
if source_id and source_type:
source_key = f"{source_type}_{source_id}"
current_sources_count = len(source_obj.get('sources', []))
if source_key not in seen_source_keys:
seen_source_keys.add(source_key)
deduplicated_sources.append(source_obj)
print(f"Debug: Added source - ID: {source_id}, Type: {source_type}, Key: {source_key}, Sources count: {current_sources_count}")
else:
# Check if this source object has more sources than the existing one
existing_index = None
for i, existing_source in enumerate(deduplicated_sources):
existing_id = existing_source.get('id')
existing_type = existing_source.get('type')
if existing_id == source_id and existing_type == source_type:
existing_index = i
break
if existing_index is not None:
existing_sources_count = len(deduplicated_sources[existing_index].get('sources', []))
if current_sources_count > existing_sources_count:
# Replace the existing source object with the new one that has more sources
deduplicated_sources[existing_index] = source_obj
print(f"Debug: Replaced source - ID: {source_id}, Type: {source_type}, Key: {source_key}, Sources count: {existing_sources_count} -> {current_sources_count}")
else:
print(f"Debug: Skipped duplicate source - ID: {source_id}, Type: {source_type}, Key: {source_key}, Sources count: {current_sources_count} <= {existing_sources_count}")
else:
print(f"Debug: Skipped duplicate source - ID: {source_id}, Type: {source_type}, Key: {source_key} (couldn't find existing)")
else:
# If there's no ID or type, just add it to be safe
deduplicated_sources.append(source_obj)
print(f"Debug: Added source without ID/type - {source_obj.get('name', 'UNKNOWN')}")
# Stream info about deduplicated sources
if streaming_service and writer:
user_source_count = len(user_selected_sources) if user_selected_sources else 0
connector_source_count = len(deduplicated_sources) - user_source_count
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Collected {len(deduplicated_sources)} total sources ({user_source_count} user-selected + {connector_source_count} from connectors)"
)
}
)
# After all sources are collected and deduplicated, stream them
if streaming_service and writer:
writer({"yield_value": streaming_service.format_sources_delta(deduplicated_sources)})
# Deduplicate raw documents based on chunk_id or content
seen_chunk_ids = set()
seen_content_hashes = set()
deduplicated_docs = []
for doc in all_raw_documents:
chunk_id = doc.get("chunk_id")
content = doc.get("content", "")
content_hash = hash(content)
# Skip if we've seen this chunk_id or content before
if (chunk_id and chunk_id in seen_chunk_ids) or content_hash in seen_content_hashes:
continue
# Add to our tracking sets and keep this document
if chunk_id:
seen_chunk_ids.add(chunk_id)
seen_content_hashes.add(content_hash)
deduplicated_docs.append(doc)
# Stream info about deduplicated documents
if streaming_service and writer:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐งน Found {len(deduplicated_docs)} unique document chunks after removing duplicates"
)
}
)
# Return deduplicated documents
return deduplicated_docs
async def process_sections(state: State, config: RunnableConfig, writer: StreamWriter) -> Dict[str, Any]:
"""
Process all sections in parallel and combine the results.
This node takes the answer outline from the previous step, fetches relevant documents
for all questions across all sections once, and then processes each section in parallel
using the sub_section_writer graph with the shared document pool.
Returns:
Dict containing the final written report in the "final_written_report" key.
"""
# Get configuration and answer outline from state
configuration = Configuration.from_runnable_config(config)
answer_outline = state.answer_outline
streaming_service = state.streaming_service
# Initialize a dictionary to track content for all sections
# This is used to maintain section content while streaming multiple sections
section_contents = {}
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ Starting to process research sections..."
)
}
)
print(f"Processing sections from outline: {answer_outline is not None}")
if not answer_outline:
error_message = "No answer outline was provided. Cannot generate report."
writer({"yield_value": streaming_service.format_error(error_message)})
return {
"final_written_report": "No answer outline was provided. Cannot generate final report."
}
# Collect all questions from all sections
all_questions = []
for section in answer_outline.answer_outline:
all_questions.extend(section.questions)
print(f"Collected {len(all_questions)} questions from all sections")
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐งฉ Found {len(all_questions)} research questions across {len(answer_outline.answer_outline)} sections"
)
}
)
# Fetch relevant documents once for all questions
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ Searching for relevant information across all connectors..."
)
}
)
if configuration.num_sections == 1:
TOP_K = 10
elif configuration.num_sections == 3:
TOP_K = 20
elif configuration.num_sections == 6:
TOP_K = 30
else:
TOP_K = 10
relevant_documents = []
user_selected_documents = []
user_selected_sources = []
try:
# First, fetch user-selected documents if any
if configuration.document_ids_to_add_in_context:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Including {len(configuration.document_ids_to_add_in_context)} user-selected documents..."
)
}
)
user_selected_sources, user_selected_documents = await fetch_documents_by_ids(
document_ids=configuration.document_ids_to_add_in_context,
user_id=configuration.user_id,
db_session=state.db_session
)
if user_selected_documents:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"โ
Successfully added {len(user_selected_documents)} user-selected documents to context"
)
}
)
# Create connector service using state db_session
connector_service = ConnectorService(state.db_session, user_id=configuration.user_id)
await connector_service.initialize_counter()
relevant_documents = await fetch_relevant_documents(
research_questions=all_questions,
user_id=configuration.user_id,
search_space_id=configuration.search_space_id,
db_session=state.db_session,
connectors_to_search=configuration.connectors_to_search,
writer=writer,
state=state,
top_k=TOP_K,
connector_service=connector_service,
search_mode=configuration.search_mode,
user_selected_sources=user_selected_sources
)
except Exception as e:
error_message = f"Error fetching relevant documents: {str(e)}"
print(error_message)
writer({"yield_value": streaming_service.format_error(error_message)})
# Log the error and continue with an empty list of documents
# This allows the process to continue, but the report might lack information
relevant_documents = []
# Combine user-selected documents with connector-fetched documents
all_documents = user_selected_documents + relevant_documents
print(f"Fetched {len(relevant_documents)} relevant documents for all sections")
print(f"Added {len(user_selected_documents)} user-selected documents for all sections")
print(f"Total documents for sections: {len(all_documents)}")
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"โจ Starting to draft {len(answer_outline.answer_outline)} sections using {len(all_documents)} total document chunks ({len(user_selected_documents)} user-selected + {len(relevant_documents)} connector-found)"
)
}
)
# Create tasks to process each section in parallel with the same document set
section_tasks = []
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"โ๏ธ Creating processing tasks for each section..."
)
}
)
for i, section in enumerate(answer_outline.answer_outline):
if i == 0:
sub_section_type = SubSectionType.START
elif i == len(answer_outline.answer_outline) - 1:
sub_section_type = SubSectionType.END
else:
sub_section_type = SubSectionType.MIDDLE
# Initialize the section_contents entry for this section
section_contents[i] = {
"title": section.section_title,
"content": "",
"index": i
}
section_tasks.append(
process_section_with_documents(
section_id=i,
section_title=section.section_title,
section_questions=section.questions,
user_query=configuration.user_query,
user_id=configuration.user_id,
search_space_id=configuration.search_space_id,
relevant_documents=all_documents, # Use combined documents
state=state,
writer=writer,
sub_section_type=sub_section_type,
section_contents=section_contents
)
)
# Run all section processing tasks in parallel
print(f"Running {len(section_tasks)} section processing tasks in parallel")
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"โณ Processing {len(section_tasks)} sections simultaneously..."
)
}
)
section_results = await asyncio.gather(*section_tasks, return_exceptions=True)
# Handle any exceptions in the results
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐งต Combining section results into final report..."
)
}
)
processed_results = []
for i, result in enumerate(section_results):
if isinstance(result, Exception):
section_title = answer_outline.answer_outline[i].section_title
error_message = f"Error processing section '{section_title}': {str(result)}"
print(error_message)
writer({"yield_value": streaming_service.format_error(error_message)})
processed_results.append(error_message)
else:
processed_results.append(result)
# Combine the results into a final report with section titles
final_report = []
for i, (section, content) in enumerate(zip(answer_outline.answer_outline, processed_results)):
# Skip adding the section header since the content already contains the title
final_report.append(content)
final_report.append("\n")
# Stream each section with its title
writer(
{
"yield_value": state.streaming_service.format_text_chunk(f"# {section.section_title}\n\n{content}")
}
)
# Join all sections with newlines
final_written_report = "\n".join(final_report)
print(f"Generated final report with {len(final_report)} parts")
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ Final research report generated successfully!"
)
}
)
# Use the shared documents for further question generation
# Since all sections used the same document pool, we can use it directly
return {
"final_written_report": final_written_report,
"reranked_documents": all_documents
}
async def process_section_with_documents(
section_id: int,
section_title: str,
section_questions: List[str],
user_id: str,
search_space_id: int,
relevant_documents: List[Dict[str, Any]],
user_query: str,
state: State = None,
writer: StreamWriter = None,
sub_section_type: SubSectionType = SubSectionType.MIDDLE,
section_contents: Dict[int, Dict[str, Any]] = None
) -> str:
"""
Process a single section using pre-fetched documents.
Args:
section_id: The ID of the section
section_title: The title of the section
section_questions: List of research questions for this section
user_id: The user ID
search_space_id: The search space ID
relevant_documents: Pre-fetched documents to use for this section
state: The current state
writer: StreamWriter for sending progress updates
sub_section_type: The type of section (start, middle, end)
section_contents: Dictionary to track content across multiple sections
Returns:
The written section content
"""
try:
# Use the provided documents
documents_to_use = relevant_documents
# Send status update via streaming if available
if state and state.streaming_service and writer:
writer(
{
"yield_value": state.streaming_service.format_terminal_info_delta(
f'๐ Writing section: "{section_title}" with {len(section_questions)} research questions'
)
}
)
# Fallback if no documents found
if not documents_to_use:
print(f"No relevant documents found for section: {section_title}")
if state and state.streaming_service and writer:
writer(
{
"yield_value": state.streaming_service.format_error(
f'Warning: No relevant documents found for section: "{section_title}"'
)
}
)
documents_to_use = [
{"content": f"No specific information was found for: {question}"}
for question in section_questions
]
# Call the sub_section_writer graph with the appropriate config
config = {
"configurable": {
"sub_section_title": section_title,
"sub_section_questions": section_questions,
"sub_section_type": sub_section_type,
"user_query": user_query,
"relevant_documents": documents_to_use,
"user_id": user_id,
"search_space_id": search_space_id,
}
}
# Create the initial state with db_session and chat_history
sub_state = {
"db_session": state.db_session,
"chat_history": state.chat_history
}
# Invoke the sub-section writer graph with streaming
print(f"Invoking sub_section_writer for: {section_title}")
if state and state.streaming_service and writer:
writer(
{
"yield_value": state.streaming_service.format_terminal_info_delta(
f'๐ง Analyzing information and drafting content for section: "{section_title}"'
)
}
)
# Variables to track streaming state
complete_content = "" # Tracks the complete content received so far
async for chunk_type, chunk in sub_section_writer_graph.astream(sub_state, config, stream_mode=["values"]):
if "final_answer" in chunk:
new_content = chunk["final_answer"]
if new_content and new_content != complete_content:
# Extract only the new content (delta)
delta = new_content[len(complete_content):]
# Update what we've processed so far
complete_content = new_content
# Only stream if there's actual new content
if delta and state and state.streaming_service and writer:
# Update terminal with real-time progress indicator
writer(
{
"yield_value": state.streaming_service.format_terminal_info_delta(
f"โ๏ธ Writing section {section_id + 1}... ({len(complete_content.split())} words)"
)
}
)
# Update section_contents with just the new delta
section_contents[section_id]["content"] += delta
# Build UI-friendly content for all sections
complete_answer = []
for i in range(len(section_contents)):
if i in section_contents and section_contents[i]["content"]:
# Add section header
complete_answer.append(f"# {section_contents[i]['title']}")
complete_answer.append("") # Empty line after title
# Add section content
content_lines = section_contents[i]["content"].split("\n")
complete_answer.extend(content_lines)
complete_answer.append("") # Empty line after content
# Set default if no content was received
if not complete_content:
complete_content = "No content was generated for this section."
section_contents[section_id]["content"] = complete_content
# Final terminal update
if state and state.streaming_service and writer:
writer(
{
"yield_value": state.streaming_service.format_terminal_info_delta(
f'โ
Completed section: "{section_title}"'
)
}
)
return complete_content
except Exception as e:
print(f"Error processing section '{section_title}': {str(e)}")
# Send error update via streaming if available
if state and state.streaming_service and writer:
writer(
{
"yield_value": state.streaming_service.format_error(
f'Error processing section "{section_title}": {str(e)}'
)
}
)
return f"Error processing section: {section_title}. Details: {str(e)}"
async def reformulate_user_query(state: State, config: RunnableConfig, writer: StreamWriter) -> Dict[str, Any]:
"""
Reforms the user query based on the chat history.
"""
configuration = Configuration.from_runnable_config(config)
user_query = configuration.user_query
chat_history_str = await QueryService.langchain_chat_history_to_str(state.chat_history)
if len(state.chat_history) == 0:
reformulated_query = user_query
else:
reformulated_query = await QueryService.reformulate_query_with_chat_history(user_query=user_query, session=state.db_session, user_id=configuration.user_id, chat_history_str=chat_history_str)
return {
"reformulated_query": reformulated_query
}
async def handle_qna_workflow(state: State, config: RunnableConfig, writer: StreamWriter) -> Dict[str, Any]:
"""
Handle the QNA research workflow.
This node fetches relevant documents for the user query and then uses the QNA agent
to generate a comprehensive answer with proper citations.
Returns:
Dict containing the final answer in the "final_written_report" key for consistency.
"""
streaming_service = state.streaming_service
configuration = Configuration.from_runnable_config(config)
reformulated_query = state.reformulated_query
user_query = configuration.user_query
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ค Starting Q&A research workflow..."
)
}
)
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f'๐ Researching: "{user_query[:100]}..."'
)
}
)
# Fetch relevant documents for the QNA query
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ Searching for relevant information across all connectors..."
)
}
)
# Use a reasonable top_k for QNA - not too many documents to avoid overwhelming the LLM
TOP_K = 15
relevant_documents = []
user_selected_documents = []
user_selected_sources = []
try:
# First, fetch user-selected documents if any
if configuration.document_ids_to_add_in_context:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ Including {len(configuration.document_ids_to_add_in_context)} user-selected documents..."
)
}
)
user_selected_sources, user_selected_documents = await fetch_documents_by_ids(
document_ids=configuration.document_ids_to_add_in_context,
user_id=configuration.user_id,
db_session=state.db_session
)
if user_selected_documents:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"โ
Successfully added {len(user_selected_documents)} user-selected documents to context"
)
}
)
# Create connector service using state db_session
connector_service = ConnectorService(state.db_session, user_id=configuration.user_id)
await connector_service.initialize_counter()
# Use the reformulated query as a single research question
research_questions = [reformulated_query, user_query]
relevant_documents = await fetch_relevant_documents(
research_questions=research_questions,
user_id=configuration.user_id,
search_space_id=configuration.search_space_id,
db_session=state.db_session,
connectors_to_search=configuration.connectors_to_search,
writer=writer,
state=state,
top_k=TOP_K,
connector_service=connector_service,
search_mode=configuration.search_mode,
user_selected_sources=user_selected_sources
)
except Exception as e:
error_message = f"Error fetching relevant documents for QNA: {str(e)}"
print(error_message)
writer({"yield_value": streaming_service.format_error(error_message)})
# Continue with empty documents - the QNA agent will handle this gracefully
relevant_documents = []
# Combine user-selected documents with connector-fetched documents
all_documents = user_selected_documents + relevant_documents
print(f"Fetched {len(relevant_documents)} relevant documents for QNA")
print(f"Added {len(user_selected_documents)} user-selected documents for QNA")
print(f"Total documents for QNA: {len(all_documents)}")
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"๐ง Generating comprehensive answer using {len(all_documents)} total sources ({len(user_selected_documents)} user-selected + {len(relevant_documents)} connector-found)..."
)
}
)
# Prepare configuration for the QNA agent
qna_config = {
"configurable": {
"user_query": user_query, # Use the reformulated query
"reformulated_query": reformulated_query,
"relevant_documents": all_documents, # Use combined documents
"user_id": configuration.user_id,
"search_space_id": configuration.search_space_id
}
}
# Create the state for the QNA agent (it has a different state structure)
qna_state = {
"db_session": state.db_session,
"chat_history": state.chat_history
}
try:
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"โ๏ธ Writing comprehensive answer with citations..."
)
}
)
# Track streaming content for real-time updates
complete_content = ""
captured_reranked_documents = []
# Call the QNA agent with streaming
async for _chunk_type, chunk in qna_agent_graph.astream(qna_state, qna_config, stream_mode=["values"]):
if "final_answer" in chunk:
new_content = chunk["final_answer"]
if new_content and new_content != complete_content:
# Extract only the new content (delta)
delta = new_content[len(complete_content):]
complete_content = new_content
# Stream the real-time answer if there's new content
if delta:
# Update terminal with progress
word_count = len(complete_content.split())
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"โ๏ธ Writing answer... ({word_count} words)"
)
}
)
writer(
{"yield_value": streaming_service.format_text_chunk(delta)}
)
# Capture reranked documents from QNA agent for further question generation
if "reranked_documents" in chunk:
captured_reranked_documents = chunk["reranked_documents"]
# Set default if no content was received
if not complete_content:
complete_content = "I couldn't find relevant information in your knowledge base to answer this question."
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ Q&A answer generated successfully!"
)
}
)
# Return the final answer and captured reranked documents for further question generation
return {
"final_written_report": complete_content,
"reranked_documents": captured_reranked_documents
}
except Exception as e:
error_message = f"Error generating QNA answer: {str(e)}"
print(error_message)
writer({"yield_value": streaming_service.format_error(error_message)})
return {"final_written_report": f"Error generating answer: {str(e)}"}
async def generate_further_questions(state: State, config: RunnableConfig, writer: StreamWriter) -> Dict[str, Any]:
"""
Generate contextually relevant follow-up questions based on chat history and available documents.
This node takes the chat history and reranked documents from sub-agents (qna_agent or sub_section_writer)
and uses an LLM to generate follow-up questions that would naturally extend the conversation
and provide additional value to the user.
Returns:
Dict containing the further questions in the "further_questions" key for state update.
"""
from app.services.llm_service import get_user_fast_llm
# Get configuration and state data
configuration = Configuration.from_runnable_config(config)
chat_history = state.chat_history
user_id = configuration.user_id
streaming_service = state.streaming_service
# Get reranked documents from the state (will be populated by sub-agents)
reranked_documents = getattr(state, 'reranked_documents', None) or []
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ค Generating follow-up questions..."
)
}
)
# Get user's fast LLM
llm = await get_user_fast_llm(state.db_session, user_id)
if not llm:
error_message = f"No fast LLM configured for user {user_id}"
print(error_message)
writer({"yield_value": streaming_service.format_error(error_message)})
# Stream empty further questions to UI
writer({"yield_value": streaming_service.format_further_questions_delta([])})
return {"further_questions": []}
# Format chat history for the prompt
chat_history_xml = "\n"
for message in chat_history:
if hasattr(message, 'type'):
if message.type == "human":
chat_history_xml += f"{message.content}\n"
elif message.type == "ai":
chat_history_xml += f"{message.content}\n"
else:
# Handle other message types if needed
chat_history_xml += f"{str(message)}\n"
chat_history_xml += ""
# Format available documents for the prompt
documents_xml = "\n"
for i, doc in enumerate(reranked_documents):
document_info = doc.get("document", {})
source_id = document_info.get("id", f"doc_{i}")
source_type = document_info.get("document_type", "UNKNOWN")
content = doc.get("content", "")
documents_xml += f"\n"
documents_xml += f"\n"
documents_xml += f"{source_id}\n"
documents_xml += f"{source_type}\n"
documents_xml += f"\n"
documents_xml += f"\n{content}\n"
documents_xml += f"\n"
documents_xml += ""
# Create the human message content
human_message_content = f"""
{chat_history_xml}
{documents_xml}
Based on the chat history and available documents above, generate 3-5 contextually relevant follow-up questions that would naturally extend the conversation and provide additional value to the user. Make sure the questions can be reasonably answered using the available documents or knowledge base.
Your response MUST be valid JSON in exactly this format:
{{
"further_questions": [
{{
"id": 0,
"question": "further qn 1"
}},
{{
"id": 1,
"question": "further qn 2"
}}
]
}}
Do not include any other text or explanation. Only return the JSON.
"""
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
"๐ง Analyzing conversation context to suggest relevant questions..."
)
}
)
# Create messages for the LLM
messages = [
SystemMessage(content=get_further_questions_system_prompt()),
HumanMessage(content=human_message_content)
]
try:
# Call the LLM
response = await llm.ainvoke(messages)
# Parse the JSON response
content = response.content
# Find the JSON in the content
json_start = content.find('{')
json_end = content.rfind('}') + 1
if json_start >= 0 and json_end > json_start:
json_str = content[json_start:json_end]
# Parse the JSON string
parsed_data = json.loads(json_str)
# Extract the further_questions array
further_questions = parsed_data.get("further_questions", [])
writer(
{
"yield_value": streaming_service.format_terminal_info_delta(
f"โ
Generated {len(further_questions)} contextual follow-up questions!"
)
}
)
# Stream the further questions to the UI
writer(
{
"yield_value": streaming_service.format_further_questions_delta(
further_questions
)
}
)
print(f"Successfully generated {len(further_questions)} further questions")
return {"further_questions": further_questions}
else:
# If JSON structure not found, return empty list
error_message = (
"Could not find valid JSON in LLM response for further questions"
)
print(error_message)
writer(
{
"yield_value": streaming_service.format_error(
f"Warning: {error_message}"
)
}
)
# Stream empty further questions to UI
writer(
{"yield_value": streaming_service.format_further_questions_delta([])}
)
return {"further_questions": []}
except (json.JSONDecodeError, ValueError) as e:
# Log the error and return empty list
error_message = f"Error parsing further questions response: {str(e)}"
print(error_message)
writer(
{"yield_value": streaming_service.format_error(f"Warning: {error_message}")}
)
# Stream empty further questions to UI
writer({"yield_value": streaming_service.format_further_questions_delta([])})
return {"further_questions": []}
except Exception as e:
# Handle any other errors
error_message = f"Error generating further questions: {str(e)}"
print(error_message)
writer(
{"yield_value": streaming_service.format_error(f"Warning: {error_message}")}
)
# Stream empty further questions to UI
writer({"yield_value": streaming_service.format_further_questions_delta([])})
return {"further_questions": []}