Spaces:
Sleeping
Sleeping
File size: 10,719 Bytes
3194955 6795b19 3194955 1651eb7 3194955 1651eb7 3194955 4db8949 3194955 1651eb7 3194955 1651eb7 3194955 6eb6385 3194955 1651eb7 3194955 1651eb7 3194955 1651eb7 3194955 6eb6385 3194955 |
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 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 |
"""
ChatUI Adapters for LangGraph Workflow Streaming
"""
import logging
import asyncio
import json
from typing import AsyncGenerator, Dict, Any
from components.utils import build_conversation_context
logger = logging.getLogger(__name__)
async def process_query_streaming(
compiled_graph,
query: str,
file_upload=None,
reports_filter: str = "",
sources_filter: str = "",
subtype_filter: str = "",
year_filter: str = "",
conversation_context: str = None,
file_content: bytes = None,
filename: str = None
):
"""
Process a query through the LangGraph workflow with streaming.
COPIED FROM ORIGINAL ORCHESTRATOR. TO BE REPLACED WITH AGENTIC WORFLOW
"""
# Build metadata filters from filter parameters
metadata_filters = None
if any([reports_filter, sources_filter, subtype_filter, year_filter]):
metadata_filters = {}
if reports_filter:
metadata_filters["reports"] = reports_filter
if sources_filter:
metadata_filters["sources"] = sources_filter
if subtype_filter:
metadata_filters["subtype"] = subtype_filter
if year_filter:
metadata_filters["year"] = year_filter
initial_state = {
"query": query,
"metadata": {"session_type": "chatui"},
"raw_documents": [],
"conversation_context": conversation_context,
"metadata_filters": metadata_filters
}
# Add file content if present
if file_content and filename:
initial_state["file_content"] = file_content
initial_state["filename"] = filename
try:
async for output in compiled_graph.astream(initial_state, stream_mode="custom"):
if output.get("event") == "data":
yield {"type": "data", "content": output["data"]}
elif output.get("event") == "final_answer":
# Handle final_answer event with webSources
sources = output["data"].get("webSources", [])
if sources:
yield {"type": "sources", "content": sources}
elif output.get("event") == "error":
yield {"type": "error", "content": output["data"].get("error", "Unknown error")}
yield {"type": "end", "content": ""}
except Exception as e:
logger.error(f"Pipeline error: {e}", exc_info=True)
yield {"type": "error", "content": str(e)}
async def chatui_adapter(data, compiled_graph, max_turns: int = 3, max_chars: int = 8000):
"""Text-only adapter for ChatUI with structured message support"""
logger.debug(f"ChatUI adapter called with data type: {type(data)}")
try:
# Handle both dict and object access patterns
if isinstance(data, dict):
text_value = data.get('text', '')
messages_value = data.get('messages', None)
preprompt_value = data.get('preprompt', None)
else:
text_value = getattr(data, 'text', '')
messages_value = getattr(data, 'messages', None)
preprompt_value = getattr(data, 'preprompt', None)
# Convert dict messages to objects if needed
messages = []
if messages_value:
for msg in messages_value:
if isinstance(msg, dict):
messages.append(type('Message', (), {
'role': msg.get('role', 'unknown'),
'content': msg.get('content', '')
})())
else:
messages.append(msg)
# Extract latest user query
user_messages = [msg for msg in messages if msg.role == 'user']
query = user_messages[-1].content if user_messages else text_value
# Conversation metadata (troubleshooting purposes)
msg_metadata = {
'total': len(messages),
'user': len(user_messages),
'assistant': len([m for m in messages if m.role == 'assistant']),
'msg_lengths': [len(m.content) for m in messages]
}
logger.info(f"Processing query: {query[:20]}... | Conversation: {msg_metadata}")
# Build conversation context for generation (last N turns)
conversation_context = build_conversation_context(messages, max_turns=max_turns, max_chars=max_chars)
full_response = ""
sources_collected = None
async for result in process_query_streaming(
compiled_graph=compiled_graph,
query=query,
file_upload=None,
reports_filter="",
sources_filter="",
subtype_filter="",
year_filter="",
conversation_context=conversation_context
):
if isinstance(result, dict):
result_type = result.get("type", "data")
content = result.get("content", "")
if result_type == "data":
full_response += content
yield content
elif result_type == "sources":
sources_collected = content
elif result_type == "end":
if sources_collected:
# Send sources as markdown with doc:// URLs for ChatUI to parse
sources_text = "\n\n**Sources:**\n"
for i, source in enumerate(sources_collected, 1):
title = source.get('title', 'Unknown')
uri = source.get('uri') or 'doc://#'
sources_text += f"{i}. [{title}]({uri})\n"
logger.info(f"Sending markdown sources with doc:// scheme")
yield sources_text
elif result_type == "error":
yield f"Error: {content}"
else:
yield str(result)
await asyncio.sleep(0)
except Exception as e:
logger.error(f"ChatUI error: {str(e)}")
logger.error(f"Full traceback:", exc_info=True)
yield f"Error: {str(e)}"
async def chatui_file_adapter(data, compiled_graph, max_turns: int = 3, max_chars: int = 8000):
"""File upload adapter for ChatUI with structured message support"""
try:
# Handle both dict and object access patterns
if isinstance(data, dict):
text_value = data.get('text', '')
messages_value = data.get('messages', None)
files_value = data.get('files', None)
preprompt_value = data.get('preprompt', None)
else:
text_value = getattr(data, 'text', '')
messages_value = getattr(data, 'messages', None)
files_value = getattr(data, 'files', None)
preprompt_value = getattr(data, 'preprompt', None)
# Extract query - prefer structured messages
conversation_context = None
if messages_value and len(messages_value) > 0:
# Convert dict messages to objects
messages = []
for msg in messages_value:
if isinstance(msg, dict):
messages.append(type('Message', (), {
'role': msg.get('role', 'unknown'),
'content': msg.get('content', '')
})())
else:
messages.append(msg)
user_messages = [msg for msg in messages if msg.role == 'user']
query = user_messages[-1].content if user_messages else text_value
# Conversation metadata (troubleshooting purposes)
msg_metadata = {
'total': len(messages),
'user': len(user_messages),
'assistant': len([m for m in messages if m.role == 'assistant']),
'msg_lengths': [len(m.content) for m in messages]
}
logger.info(f"Processing query with file: {query[:20]}... | Conversation: {msg_metadata}")
conversation_context = build_conversation_context(messages, max_turns=max_turns, max_chars=max_chars)
else:
query = text_value
file_content = None
filename = None
if files_value and len(files_value) > 0:
file_info = files_value[0]
logger.info(f"Processing file: {file_info.get('name', 'unknown')}")
if file_info.get('type') == 'base64' and file_info.get('content'):
try:
import base64
file_content = base64.b64decode(file_info['content'])
filename = file_info.get('name', 'uploaded_file')
except Exception as e:
logger.error(f"Error decoding base64 file: {str(e)}")
yield f"Error: Failed to decode uploaded file - {str(e)}"
return
sources_collected = None
async for result in process_query_streaming(
compiled_graph=compiled_graph,
query=query,
file_upload=None,
reports_filter="",
sources_filter="",
subtype_filter="",
year_filter="",
conversation_context=conversation_context,
file_content=file_content,
filename=filename
):
if isinstance(result, dict):
result_type = result.get("type", "data")
content = result.get("content", "")
if result_type == "data":
yield content
elif result_type == "sources":
sources_collected = content
elif result_type == "end":
if sources_collected:
# Send sources as markdown with doc:// URLs for ChatUI to parse
sources_text = "\n\n**Sources:**\n"
for i, source in enumerate(sources_collected, 1):
if isinstance(source, dict):
title = source.get('title', 'Unknown')
uri = source.get('uri') or 'doc://#'
sources_text += f"{i}. [{title}]({uri})\n"
else:
sources_text += f"{i}. {str(source)}\n"
logger.info(f"Sending markdown sources with doc:// scheme (file)")
yield sources_text
elif result_type == "error":
yield f"Error: {content}"
else:
yield str(result)
await asyncio.sleep(0)
except Exception as e:
logger.error(f"ChatUI file adapter error: {str(e)}")
yield f"Error: {str(e)}" |