Mod example to use run_with_services

examples/ordinance_gpt/parse_pdf.py

@@ -11,34 +11,13 @@
 from elm.ords.llm import LLMCaller
 from elm.ords.extraction.ordinance import OrdinanceExtractor
 from elm.ords.extraction.apply import extract_ordinance_values
-from elm.ords.services.provider import RunningAsyncServices
+from elm.ords.services.provider import run_with_services
 from rex import init_logger
 
 from elm.ords.extraction.apply import (check_for_ordinance_info,
                                        extract_ordinance_text_with_llm)
 
 
-async def clean_text(services, doc, **kwargs):
-    text_splitter = RecursiveCharacterTextSplitter(
-        RTS_SEPARATORS,
-        chunk_size=3000,
-        chunk_overlap=300,
-        length_function=partial(ApiBase.count_tokens, model='gpt-4'),
-    )
-    async with RunningAsyncServices(services):
-        extractor = OrdinanceExtractor(LLMCaller(**kwargs))
-        doc = await check_for_ordinance_info(doc, text_splitter, **kwargs)
-        doc = await extract_ordinance_text_with_llm(doc, text_splitter,
-                                                    extractor)
-    return doc
-
-
-async def extract_ordinances(services, doc, **kwargs):
-    async with RunningAsyncServices(services):
-        doc = await extract_ordinance_values(doc, **kwargs)
-    return doc
-
-
 if __name__ == '__main__':
     init_logger('elm', log_level='INFO')
 
@@ -50,6 +29,13 @@ async def extract_ordinances(services, doc, **kwargs):
 
     doc = PDFDocument.from_file(fp_pdf)
 
+    text_splitter = RecursiveCharacterTextSplitter(
+        RTS_SEPARATORS,
+        chunk_size=3000,
+        chunk_overlap=300,
+        length_function=partial(ApiBase.count_tokens, model='gpt-4'),
+    )
+
     azure_api_key, azure_version, azure_endpoint = validate_api_params()
     client = openai.AsyncAzureOpenAI(api_key=azure_api_key,
                                      api_version=azure_version,
@@ -58,9 +44,16 @@ async def extract_ordinances(services, doc, **kwargs):
     initialize_service_queue(llm_service.__class__.__name__)
     services = [llm_service]
     kwargs = dict(llm_service=llm_service, model='gpt-4', temperature=0)
+    extractor = OrdinanceExtractor(LLMCaller(**kwargs))
+
+    check_for_ords = check_for_ordinance_info(doc, text_splitter, **kwargs)
+    doc = run_with_services(services, check_for_ords)
+
+    extract_text = extract_ordinance_text_with_llm(doc, text_splitter,
+                                                   extractor)
+    doc = run_with_services(services, extract_text)
 
-    doc = asyncio.run(clean_text(services, doc, **kwargs))
-    doc = asyncio.run(extract_ordinances(services, doc, **kwargs))
+    doc = run_with_services(services, extract_ordinance_values(doc, **kwargs))
 
     doc.metadata['ordinance_values'].to_csv(fp_ords)
     with open(fp_txt_all, 'w') as f: