Spaces:
Sleeping
Sleeping
File size: 9,862 Bytes
5d42805 |
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 |
import uuid
from fastapi import FastAPI
from fastapi.responses import StreamingResponse
from langchain_core.messages import (
BaseMessage,
HumanMessage,
trim_messages,
)
from langchain_core.tools import tool
from langchain_openai import ChatOpenAI
from langgraph.checkpoint.memory import MemorySaver
from langgraph.prebuilt import create_react_agent
from pydantic import BaseModel
import json
from typing import Optional, Annotated
from langchain_core.runnables import RunnableConfig
from langgraph.prebuilt import InjectedState
from document_rag_router import router as document_rag_router
from document_rag_router import QueryInput, query_collection, SearchResult
from fastapi import HTTPException
import requests
from sse_starlette.sse import EventSourceResponse
from fastapi.middleware.cors import CORSMiddleware
import re
app = FastAPI()
app.include_router(document_rag_router)
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
@tool
def get_user_age(name: str) -> str:
"""Use this tool to find the user's age."""
if "bob" in name.lower():
return "42 years old"
return "41 years old"
@tool
async def query_documents(
query: str,
config: RunnableConfig,
#state: Annotated[dict, InjectedState]
) -> str:
"""Use this tool to retrieve relevant data from the collection.
Args:
query: The search query to find relevant document passages
"""
# Get collection_id and user_id from config
thread_config = config.get("configurable", {})
collection_id = thread_config.get("collection_id")
user_id = thread_config.get("user_id")
if not collection_id or not user_id:
return "Error: collection_id and user_id are required in the config"
try:
# Create query input
input_data = QueryInput(
collection_id=collection_id,
query=query,
user_id=user_id,
top_k=6
)
response = await query_collection(input_data)
results = []
# Access response directly since it's a Pydantic model
for r in response.results:
result_dict = {
"text": r.text,
"distance": r.distance,
"metadata": {
"document_id": r.metadata.get("document_id"),
"chunk_index": r.metadata.get("location", {}).get("chunk_index")
}
}
results.append(result_dict)
return str(results)
except Exception as e:
print(e)
return f"Error querying documents: {e} PAUSE AND ASK USER FOR HELP"
async def query_documents_raw(
query: str,
config: RunnableConfig,
#state: Annotated[dict, InjectedState]
) -> SearchResult:
"""Use this tool to retrieve relevant data from the collection.
Args:
query: The search query to find relevant document passages
"""
# Get collection_id and user_id from config
thread_config = config.get("configurable", {})
collection_id = thread_config.get("collection_id")
user_id = thread_config.get("user_id")
if not collection_id or not user_id:
return "Error: collection_id and user_id are required in the config"
try:
# Create query input
input_data = QueryInput(
collection_id=collection_id,
query=query,
user_id=user_id,
top_k=6
)
response = await query_collection(input_data)
return response.results
except Exception as e:
print(e)
return f"Error querying documents: {e} PAUSE AND ASK USER FOR HELP"
memory = MemorySaver()
model = ChatOpenAI(model="gpt-4o-mini", streaming=True)
def state_modifier(state) -> list[BaseMessage]:
return trim_messages(
state["messages"],
token_counter=len,
max_tokens=16000,
strategy="last",
start_on="human",
include_system=True,
allow_partial=False,
)
agent = create_react_agent(
model,
tools=[query_documents],
checkpointer=memory,
state_modifier=state_modifier,
)
class ChatInput(BaseModel):
message: str
thread_id: Optional[str] = None
collection_id: Optional[str] = None
user_id: Optional[str] = None
@app.post("/chat")
async def chat(input_data: ChatInput):
thread_id = input_data.thread_id or str(uuid.uuid4())
config = {
"configurable": {
"thread_id": thread_id,
"collection_id": input_data.collection_id,
"user_id": input_data.user_id
}
}
input_message = HumanMessage(content=input_data.message)
async def generate():
async for event in agent.astream_events(
{"messages": [input_message]},
config,
version="v2"
):
kind = event["event"]
if kind == "on_chat_model_stream":
content = event["data"]["chunk"].content
if content:
yield f"{json.dumps({'type': 'token', 'content': content})}"
elif kind == "on_tool_start":
tool_input = str(event['data'].get('input', ''))
yield f"{json.dumps({'type': 'tool_start', 'tool': event['name'], 'input': tool_input})}"
elif kind == "on_tool_end":
tool_output = str(event['data'].get('output', ''))
yield f"{json.dumps({'type': 'tool_end', 'tool': event['name'], 'output': tool_output})}"
return EventSourceResponse(
generate(),
media_type="text/event-stream"
)
async def clean_tool_input(tool_input: str):
# Use regex to parse the first key and value
pattern = r"{\s*'([^']+)':\s*'([^']+)'"
match = re.search(pattern, tool_input)
if match:
key, value = match.groups()
return {key: value}
return [tool_input]
async def clean_tool_response(tool_output: str):
"""Clean and extract relevant information from tool response if it contains query_documents."""
if "query_documents" in tool_output:
try:
# First safely evaluate the string as a Python literal
import ast
print(tool_output)
# Extract the list string from the content
start = tool_output.find("[{")
end = tool_output.rfind("}]") + 2
if start >= 0 and end > 0:
list_str = tool_output[start:end]
# Convert string to Python object using ast.literal_eval
results = ast.literal_eval(list_str)
# Return only relevant fields
return [{"text": r["text"], "document_id": r["metadata"]["document_id"]}
for r in results]
except SyntaxError as e:
print(f"Syntax error in parsing: {e}")
return f"Error parsing document results: {str(e)}"
except Exception as e:
print(f"General error: {e}")
return f"Error processing results: {str(e)}"
return tool_output
@app.post("/chat2")
async def chat2(input_data: ChatInput):
thread_id = input_data.thread_id or str(uuid.uuid4())
config = {
"configurable": {
"thread_id": thread_id,
"collection_id": input_data.collection_id,
"user_id": input_data.user_id
}
}
input_message = HumanMessage(content=input_data.message)
async def generate():
async for event in agent.astream_events(
{"messages": [input_message]},
config,
version="v2"
):
kind = event["event"]
if kind == "on_chat_model_stream":
content = event["data"]["chunk"].content
if content:
yield f"{json.dumps({'type': 'token', 'content': content})}"
elif kind == "on_tool_start":
tool_name = event['name']
tool_input = event['data'].get('input', '')
clean_input = await clean_tool_input(str(tool_input))
yield f"{json.dumps({'type': 'tool_start', 'tool': tool_name, 'inputs': clean_input})}"
elif kind == "on_tool_end":
if "query_documents" in event['name']:
print(event)
raw_output = await query_documents_raw(str(event['data'].get('input', '')), config)
try:
serializable_output = [
{
"text": result.text,
"distance": result.distance,
"metadata": result.metadata
}
for result in raw_output
]
yield f"{json.dumps({'type': 'tool_end', 'tool': event['name'], 'output': json.dumps(serializable_output)})}"
except Exception as e:
print(e)
yield f"{json.dumps({'type': 'tool_end', 'tool': event['name'], 'output': str(raw_output)})}"
else:
tool_name = event['name']
raw_output = str(event['data'].get('output', ''))
clean_output = await clean_tool_response(raw_output)
yield f"{json.dumps({'type': 'tool_end', 'tool': tool_name, 'output': clean_output})}"
return EventSourceResponse(
generate(),
media_type="text/event-stream"
)
@app.get("/health")
async def health_check():
return {"status": "healthy"} |