LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
[FIX] Prompt service v2 regression fixes (#1257)
Deepak K034ae0eabf0b
diff --git a/backend/prompt_studio/prompt_studio_core_v2/prompt_studio_helper.py b/backend/prompt_studio/prompt_studio_core_v2/prompt_studio_helper.pyindex f36a08500..e55e1cd62 100644--- a/backend/prompt_studio/prompt_studio_core_v2/prompt_studio_helper.py+++ b/backend/prompt_studio/prompt_studio_core_v2/prompt_studio_helper.py@@ -378,13 +378,6 @@ class PromptStudioHelper:fs=fs_instance,tool=util,)- if DocumentIndexingService.is_document_indexing(- org_id=org_id, user_id=user_id, doc_id_key=doc_id- ):- return {- "status": IndexingStatus.PENDING_STATUS.value,- "output": IndexingStatus.DOCUMENT_BEING_INDEXED.value,- }extracted_text = PromptStudioHelper.dynamic_extractor(profile_manager=default_profile,file_path=file_path,@@ -1360,7 +1353,8 @@ class PromptStudioHelper:status_code = response.get("status_code")if status_code == 200:response_data = response.get("structure_output")- extracted_text = json.loads(response_data)+ structure_output = json.loads(response_data)+ extracted_text = structure_output.get("extracted_text")PromptStudioIndexHelper.mark_extraction_status(document_id=document_id,profile_manager=profile_manager,diff --git a/backend/sample.env b/backend/sample.envindex 102fb92d8..c149e73e6 100644--- a/backend/sample.env+++ b/backend/sample.env@@ -75,9 +75,9 @@ PROMPT_STUDIO_FILE_PATH=/app/prompt-studio-data# Structure Tool Image (Runs prompt studio exported tools)# https://hub.docker.com/r/unstract/tool-structure-STRUCTURE_TOOL_IMAGE_URL="docker:unstract/tool-structure:0.0.74"+STRUCTURE_TOOL_IMAGE_URL="docker:unstract/tool-structure:0.0.75"STRUCTURE_TOOL_IMAGE_NAME="unstract/tool-structure"-STRUCTURE_TOOL_IMAGE_TAG="0.0.74"+STRUCTURE_TOOL_IMAGE_TAG="0.0.75"# Feature FlagsEVALUATION_SERVER_IP=unstract-fliptdiff --git a/prompt-service/src/unstract/prompt_service/core/index_v2.py b/prompt-service/src/unstract/prompt_service/core/index_v2.pyindex e3c1342ad..65d67d214 100644--- a/prompt-service/src/unstract/prompt_service/core/index_v2.py+++ b/prompt-service/src/unstract/prompt_service/core/index_v2.py@@ -153,7 +153,7 @@ class Index:vector_db: VectorDB,doc_id: str,extracted_text: str,- doc_id_found: str,+ doc_id_found: bool,):if isinstance(vector_db.get_vector_db(@@ -183,7 +183,7 @@ class Index:try:vector_db.index_document(documents,- chunk_size=self.chunking_config.chunk_overlap,+ chunk_size=self.chunking_config.chunk_size,chunk_overlap=self.chunking_config.chunk_overlap,show_progress=True,)diff --git a/prompt-service/src/unstract/prompt_service/helpers/usage.py b/prompt-service/src/unstract/prompt_service/helpers/usage.pyindex eecade1bd..8fd5ea010 100644--- a/prompt-service/src/unstract/prompt_service/helpers/usage.py+++ b/prompt-service/src/unstract/prompt_service/helpers/usage.py@@ -40,11 +40,11 @@ class UsageHelper:# Process results as neededfor row in results:key, item = UsageHelper._get_key_and_item(row)- # Initialize the key as an empty list if it doesn't exist- if key not in metadata:- metadata[key] = []- # Append the item to the list associated with the key- metadata[key].append(item)+ # Initialize the key as an empty list if it doesn't exist+ if key not in metadata:+ metadata[key] = []+ # Append the item to the list associated with the key+ metadata[key].append(item)except Exception as e:logger.error(f"Error while querying usage metadata: {e}")return metadatadiff --git a/prompt-service/src/unstract/prompt_service/services/indexing.py b/prompt-service/src/unstract/prompt_service/services/indexing.pyindex 9205294c8..d4090facf 100644--- a/prompt-service/src/unstract/prompt_service/services/indexing.py+++ b/prompt-service/src/unstract/prompt_service/services/indexing.py@@ -1,6 +1,5 @@import logging-from flask import current_app as appfrom unstract.prompt_service.core.index_v2 import Indexfrom unstract.prompt_service.dto import (ChunkingConfig,@@ -70,9 +69,6 @@ class IndexingService:embedding=embedding,vector_db=vector_db,)- if doc_id_found:- app.logger.info("Doc ID found: %s", doc_id_found)- return doc_id# Index and return doc_idindex.perform_indexing(diff --git a/tools/structure/src/config/properties.json b/tools/structure/src/config/properties.jsonindex d3269fe2f..9b618c952 100644--- a/tools/structure/src/config/properties.json+++ b/tools/structure/src/config/properties.json@@ -2,7 +2,7 @@"schemaVersion": "0.0.1","displayName": "Structure Tool","functionName": "structure_tool",- "toolVersion": "0.0.74",+ "toolVersion": "0.0.75","description": "This is a template tool which can answer set of input prompts designed in the Prompt Studio","input": {"description": "File that needs to be indexed and parsed for answers"diff --git a/tools/structure/src/helpers.py b/tools/structure/src/helpers.pyindex aac6a4293..8855c477c 100644--- a/tools/structure/src/helpers.py+++ b/tools/structure/src/helpers.py@@ -1,4 +1,5 @@import datetime+import jsonfrom typing import Any, Optionalfrom constants import IndexingConstants as IKeys@@ -42,8 +43,10 @@ class StructureToolHelper:prompt_port=tool.get_env_or_die(SettingsKeys.PROMPT_PORT),)tool.stream_log(f"responder : {responder}")- extracted_text = responder.extract(payload=payload)-+ response = responder.extract(payload=payload)+ response_data = response.get("structure_output")+ structure_output = json.loads(response_data)+ extracted_text = structure_output.get("extracted_text")return extracted_text@staticmethod@@ -59,8 +62,8 @@ class StructureToolHelper:chunk_size: int,chunk_overlap: int,file_hash: Optional[str] = None,- tool_id: str = None,- extracted_text: str = None,+ tool_id: Optional[str] = None,+ extracted_text: Optional[str] = None,) -> str:x2text = tool_settings[SettingsKeys.X2TEXT_ADAPTER]