Skip to content

LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows

UN-2444 [FEAT] Adding support to Azure OpenAI (#1393)

harini-venkataraman2fba9503d5de
diff --git a/prompt-service/src/unstract/prompt_service/controllers/answer_prompt.py b/prompt-service/src/unstract/prompt_service/controllers/answer_prompt.py
index 5a0b5d289..db501980b 100644
--- a/prompt-service/src/unstract/prompt_service/controllers/answer_prompt.py
+++ b/prompt-service/src/unstract/prompt_service/controllers/answer_prompt.py
@@ -177,11 +177,28 @@ def prompt_processor() -> Any:
raise APIError(message=msg)
if output[PSKeys.TYPE] == PSKeys.TABLE:
- llm_config = (
- ToolAdapter.get_adapter_config(util, output[PSKeys.LLM])
- .get("adapter_metadata")
- .to_dict()
- )
+ adapter_parent_data = ToolAdapter.get_adapter_config(util, output[PSKeys.LLM])
+ llm_config = adapter_parent_data.get("adapter_metadata")
+ adapter_id = adapter_parent_data.get("adapter_id")
+ adapter_prefix = adapter_id.split("|")[0]
+ llm_adapter_config = {"adapter_id": adapter_prefix}
+ if adapter_prefix == "azureopenai":
+ llm_adapter_config["model"] = llm_config.get("model")
+ llm_adapter_config["api_key"] = llm_config.get("api_key")
+ llm_adapter_config["api_base"] = llm_config.get("azure_endpoint")
+ llm_adapter_config["max_retries"] = llm_config.get("max_retries")
+ llm_adapter_config["timeout"] = llm_config.get("timeout")
+ if adapter_prefix == "openai":
+ llm_adapter_config["model"] = llm_config.get("model")
+ llm_adapter_config["api_key"] = llm_config.get("api_key")
+ llm_adapter_config["api_base"] = llm_config.get("api_base")
+ llm_adapter_config["max_retries"] = llm_config.get("max_retries")
+ llm_adapter_config["timeout"] = llm_config.get("timeout")
+ if adapter_prefix == "anthropic":
+ llm_adapter_config["model"] = llm_config.get("model")
+ llm_adapter_config["api_key"] = llm_config.get("api_key")
+ llm_adapter_config["max_retries"] = llm_config.get("max_retries")
+ llm_adapter_config["timeout"] = llm_config.get("timeout")
table_settings = output[PSKeys.TABLE_SETTINGS]
document_type: str = table_settings.get(PSKeys.DOCUMENT_TYPE)
app.logger.info("Document type: %s", document_type)
@@ -200,7 +217,7 @@ def prompt_processor() -> Any:
extractor.process(
extractor_settings=output,
extracted_data=extracted_data,
- llm_config=llm_config,
+ llm_config=llm_adapter_config,
schema=prompt_text,
)
)
diff --git a/prompt-service/src/unstract/prompt_service/services/rentrolls_extractor/interface.py b/prompt-service/src/unstract/prompt_service/services/rentrolls_extractor/interface.py
index fdc215234..91c073e4c 100644
--- a/prompt-service/src/unstract/prompt_service/services/rentrolls_extractor/interface.py
+++ b/prompt-service/src/unstract/prompt_service/services/rentrolls_extractor/interface.py
@@ -37,7 +37,7 @@ class RentRollExtractor:
try:
# Load environment variables for rent roll service
rentroll_host = os.environ.get("RENTROLL_SERVICE_HOST", "http://localhost")
- rentroll_port = os.environ.get("RENTROLL_SERVICE_PORT", "5001")
+ rentroll_port = os.environ.get("RENTROLL_SERVICE_PORT", "5003")
rentroll_url = f"{rentroll_host}:{rentroll_port}/api/extract-rentrolls"
# Prepare the request payload