Spaces:
Sleeping
Sleeping
File size: 10,338 Bytes
3194955 1651eb7 3194955 1651eb7 3194955 6329a68 3194955 1651eb7 3194955 1651eb7 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 |
"""
LangGraph orchestration nodes for retrieval and generation
NEEDS TO BE UPDATED
"""
import logging
logger = logging.getLogger(__name__)
from datetime import datetime
import json
from typing import TYPE_CHECKING
from langchain_core.documents import Document
from .telemetry import extract_retriever_telemetry
from components.ingestor.ingestor import process_document
# Assuming these Type definitions are available from state.py and retriever_orchestrator.py
if TYPE_CHECKING:
from components.retriever.retriever_orchestrator import ChaBoHFEndpointRetriever
from components.generator.generator_orchestrator import Generator
from components.orchestration.state import GraphState
async def retrieve_node(
state: 'GraphState',
retriever: 'ChaBoHFEndpointRetriever' # Injected LangChain BaseRetriever instance
) -> 'GraphState':
"""
Node to retrieve relevant context using the ChaBoHFEndpointRetriever.
The retriever performs Embed -> Search -> Rerank in one async call.
"""
start_time = datetime.now()
# 1. Extract Query and Filters
filters = state.get("metadata_filters")
metadata = state.get("metadata", {})
logger.info(f"Retrieval: {state['query'][:50]}...")
raw_documents: list[Document] = []
try:
retriever_kwargs = {}
if filters:
retriever_kwargs['filters'] = filters
raw_documents = await retriever.ainvoke(
input=state['query'],
**retriever_kwargs
)
duration = (datetime.now() - start_time).total_seconds()
retriever_config = {
"initial_k": retriever.initial_k,
"final_k": retriever.final_k,
"qdrant_mode": retriever.qdrant_mode,
}
retriever_telemetry = extract_retriever_telemetry(raw_documents, retriever_config)
metadata.update({
"retrieval_duration": duration,
"filters_applied": json.dumps(filters) if filters else "None",
"retriever_config": retriever_telemetry,
"retrieval_success": True
})
return {
"raw_documents": raw_documents,
"metadata": metadata
}
except Exception as e:
duration = (datetime.now() - start_time).total_seconds()
logger.error(f"Retrieval failed: {str(e)}", exc_info=True)
metadata.update({
"retrieval_duration": duration,
"retrieval_success": False,
"retrieval_error": str(e)
})
return {"raw_documents": [], "metadata": metadata}
async def generate_node_streaming(state: "GraphState", generator: "Generator", *, writer):
"""
Node to generate the final response with StreamWriter for LangGraph custom streaming.
Uses StreamWriter to emit events that LangGraph can capture with stream_mode="custom".
"""
start_time = datetime.now()
query = state.get("query")
raw_docs = state.get("raw_documents", [])
metadata = state.get("metadata", {})
ingestor_context = state.get("ingestor_context")
# If we have ingestor_context, prepend it to raw_docs as a Document
if ingestor_context:
ingestor_doc = Document(
page_content=ingestor_context,
metadata={"source": "uploaded_file", "filename": state.get("filename", "unknown")}
)
raw_docs = [ingestor_doc] + raw_docs
logger.info(f"Including ingestor context ({len(ingestor_context)} chars) with retrieved docs")
accumulated_text = ""
logger.info(f"Generation: {query[:50]}... ({len(raw_docs)} docs)")
conversation_context = state.get("conversation_context")
try:
async for event in generator.generate_streaming(
query=query,
context=raw_docs,
chatui_format=True,
conversation_context=conversation_context
):
# Track content to calculate metadata (length) at the end
if event.get("event") == "data":
accumulated_text += event.get("data", "")
# Use StreamWriter to emit custom events
writer(event)
# Final Telemetry Update
duration = (datetime.now() - start_time).total_seconds()
metadata.update({
"generation_duration": duration,
"generation_success": True,
"response_length": len(accumulated_text)
})
logger.info(f"Streaming complete in {duration:.2f}s. Length: {len(accumulated_text)}")
except Exception as e:
duration = (datetime.now() - start_time).total_seconds()
logger.error(f"Generation node failed: {e}", exc_info=True)
metadata.update({
"generation_duration": duration,
"generation_success": False,
"generation_error": str(e)
})
writer({"event": "error", "data": {"error": str(e)}})
async def ingest_node(state: 'GraphState') -> 'GraphState':
"""
Node to process uploaded documents (PDF, DOCX) and extract chunked context.
Only runs if file_content and filename are present in state.
"""
start_time = datetime.now()
file_content = state.get("file_content")
filename = state.get("filename")
metadata = state.get("metadata", {})
# Skip if no file uploaded
if not file_content or not filename:
logger.info("No file to ingest, skipping ingest_node")
return {}
logger.info(f"Ingesting document: {filename}")
try:
# Process document and get chunked context
ingestor_context = process_document(file_content, filename)
duration = (datetime.now() - start_time).total_seconds()
metadata.update({
"ingest_duration": duration,
"ingest_success": True,
"ingested_filename": filename,
"ingestor_context_length": len(ingestor_context)
})
logger.info(f"Document ingested successfully in {duration:.2f}s")
return {
"ingestor_context": ingestor_context,
"metadata": metadata
}
except Exception as e:
duration = (datetime.now() - start_time).total_seconds()
logger.error(f"Document ingestion failed: {str(e)}", exc_info=True)
metadata.update({
"ingest_duration": duration,
"ingest_success": False,
"ingest_error": str(e)
})
return {"ingestor_context": "", "metadata": metadata}
# from .state import GraphState
# if TYPE_CHECKING:
# from components.retriever.retriever_orchestrator import RetrieverOrchestrator
# from components.orchestration.state import GraphState
# async def retrieve_node(
# state: GraphState,
# retriever: 'RetrieverOrchestrator' # Injected service instance
# ) -> GraphState:
# """Retrieve relevant context using adapter"""
# start_time = datetime.now()
# logger.info(f"Retrieval: {state['query'][:50]}...")
# context = ""
# try:
# # Get filters from state (provided by ChatUI or LLM agent)
# filters = state.get("metadata_filters")
# # --- FILLED CODE START ---
# # Call the async method on the injected service instance
# # The retriever orchestrator handles the remote API call to the Reranker/Embedder service
# context_docs, retriever_meta = await retriever.aretrieve(
# query=latest_message,
# filters=filters
# )
# # Format the retrieved documents into a single context string
# # (This is commonly done here or inside the orchestrator)
# context = "\n---\n".join([doc.page_content for doc in context_docs])
# # --- FILLED CODE END ---
# duration = (datetime.now() - start_time).total_seconds()
# metadata = state.get("metadata", {})
# # Update metadata and append retriever-specific metadata
# metadata.update({
# "retrieval_duration": duration,
# "context_length": len(context) if context else 0,
# "retrieval_success": True,
# "filters_applied": filters,
# "retriever_config": retriever_meta, # Add metadata from retriever call
# })
# # Return the updated state
# return {"context": context, "metadata": metadata}
# except Exception as e:
# # ... (Error handling logic is good, no change needed) ...
# duration = (datetime.now() - start_time).total_seconds()
# logger.error(f"Retrieval failed: {str(e)}")
# metadata = state.get("metadata", {})
# metadata.update({
# "retrieval_duration": duration,
# "retrieval_success": False,
# "retrieval_error": str(e)
# })
# # Note: We return context as an empty string on failure to avoid cascading errors
# return {"context": "", "metadata": metadata}
# async def retrieve_node(state: GraphState) -> GraphState:
# """Retrieve relevant context using adapter"""
# start_time = datetime.now()
# logger.info(f"Retrieval: {state['query'][:50]}...")
# try:
# # Get filters from state (provided by ChatUI or LLM agent)
# filters = state.get("metadata_filters")
# # instantiate the retirever instance
# # get context using aysnc call
# duration = (datetime.now() - start_time).total_seconds()
# metadata = state.get("metadata", {})
# metadata.update({
# "retrieval_duration": duration,
# "context_length": len(context) if context else 0,
# "retrieval_success": True,
# "filters_applied": filters,
# "retriever_config": # get metadata from retirever
# })
# return {"context": context, "metadata": metadata}
# except Exception as e:
# duration = (datetime.now() - start_time).total_seconds()
# logger.error(f"Retrieval failed: {str(e)}")
# metadata = state.get("metadata", {})
# metadata.update({
# "retrieval_duration": duration,
# "retrieval_success": False,
# "retrieval_error": str(e)
# })
# return {"context": "", "metadata": metadata} |