Spaces:
Sleeping
Sleeping
File size: 6,281 Bytes
42cd5f6 |
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 |
from rag.agents.interface import Pipeline
from llama_index.core.program import LLMTextCompletionProgram
import json
from llama_index.llms.ollama import Ollama
from typing import List
from pydantic import create_model
from rich.progress import Progress, SpinnerColumn, TextColumn
import requests
import warnings
import box
import yaml
import timeit
from rich import print
from typing import Any
warnings.filterwarnings("ignore", category=DeprecationWarning)
warnings.filterwarnings("ignore", category=UserWarning)
# Import config vars
with open('config.yml', 'r', encoding='utf8') as ymlfile:
cfg = box.Box(yaml.safe_load(ymlfile))
class VProcessorPipeline(Pipeline):
def run_pipeline(self,
payload: str,
query_inputs: [str],
query_types: [str],
keywords: [str],
query: str,
file_path: str,
index_name: str,
options: List[str] = None,
group_by_rows: bool = True,
update_targets: bool = True,
debug: bool = False,
local: bool = True) -> Any:
print(f"\nRunning pipeline with {payload}\n")
start = timeit.default_timer()
if file_path is None:
raise ValueError("File path is required for vprocessor pipeline")
with open(file_path, "rb") as file:
files = {'file': (file_path, file, 'image/jpeg')}
data = {
'image_url': ''
}
response = self.invoke_pipeline_step(lambda: requests.post(cfg.OCR_ENDPOINT_VPROCESSOR,
data=data,
files=files,
timeout=180),
"Running OCR...",
local)
if response.status_code != 200:
print('Request failed with status code:', response.status_code)
print('Response:', response.text)
return "Failed to process file. Please try again."
end = timeit.default_timer()
print(f"Time to run OCR: {end - start}")
start = timeit.default_timer()
data = response.json()
ResponseModel = self.invoke_pipeline_step(lambda: self.build_response_class(query_inputs, query_types),
"Building dynamic response class...",
local)
prompt_template_str = """\
""" + query + """\
using this structured data, coming from OCR {document_data}.\
"""
llm_ollama = self.invoke_pipeline_step(lambda: Ollama(model=cfg.LLM_VPROCESSOR,
base_url=cfg.OLLAMA_BASE_URL_VPROCESSOR,
temperature=0,
request_timeout=900),
"Loading Ollama...",
local)
program = LLMTextCompletionProgram.from_defaults(
output_cls=ResponseModel,
prompt_template_str=prompt_template_str,
llm=llm_ollama,
verbose=True,
)
output = self.invoke_pipeline_step(lambda: program(document_data=data),
"Running inference...",
local)
answer = self.beautify_json(output.model_dump_json())
end = timeit.default_timer()
print(f"\nJSON response:\n")
print(answer + '\n')
print('=' * 50)
print(f"Time to retrieve answer: {end - start}")
return answer
def prepare_files(self, file_path, file):
if file_path is not None:
with open(file_path, "rb") as file:
files = {'file': (file_path, file, 'image/jpeg')}
data = {
'image_url': ''
}
else:
files = {'file': (file.filename, file.file, file.content_type)}
data = {
'image_url': ''
}
return data, files
# Function to safely evaluate type strings
def safe_eval_type(self, type_str, context):
try:
return eval(type_str, {}, context)
except NameError:
raise ValueError(f"Type '{type_str}' is not recognized")
def build_response_class(self, query_inputs, query_types_as_strings):
# Controlled context for eval
context = {
'List': List,
'str': str,
'int': int,
'float': float
# Include other necessary types or typing constructs here
}
# Convert string representations to actual types
query_types = [self.safe_eval_type(type_str, context) for type_str in query_types_as_strings]
# Create fields dictionary
fields = {name: (type_, ...) for name, type_ in zip(query_inputs, query_types)}
DynamicModel = create_model('DynamicModel', **fields)
return DynamicModel
def invoke_pipeline_step(self, task_call, task_description, local):
if local:
with Progress(
SpinnerColumn(),
TextColumn("[progress.description]{task.description}"),
transient=False,
) as progress:
progress.add_task(description=task_description, total=None)
ret = task_call()
else:
print(task_description)
ret = task_call()
return ret
def beautify_json(self, result):
try:
# Convert and pretty print
data = json.loads(str(result))
data = json.dumps(data, indent=4)
return data
except (json.decoder.JSONDecodeError, TypeError):
print("The response is not in JSON format:\n")
print(result)
return {} |