LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
Front end and tool helper changes for OCR support
gayathrivijayakumar05aceec82853
diff --git a/backend/adapter_processor/constants.py b/backend/adapter_processor/constants.pyindex 429d67b3a..5a6296f53 100644--- a/backend/adapter_processor/constants.py+++ b/backend/adapter_processor/constants.py@@ -4,6 +4,7 @@ class AdapterKeys:IS_DEFAULT = "is_default"LLM = "LLM"X2TEXT = "X2TEXT"+ OCR = "OCR"VECTOR_DB = "VECTOR_DB"EMBEDDING = "EMBEDDING"NAME = "name"diff --git a/backend/adapter_processor/views.py b/backend/adapter_processor/views.pyindex 2009527af..8b7144371 100644--- a/backend/adapter_processor/views.py+++ b/backend/adapter_processor/views.py@@ -63,6 +63,7 @@ class AdapterViewSet(GenericViewSet):or adapter_type == AdapterKeys.EMBEDDINGor adapter_type == AdapterKeys.VECTOR_DBor adapter_type == AdapterKeys.X2TEXT+ or adapter_type == AdapterKeys.OCR):json_schema = AdapterProcessor.get_all_supported_adapters(type=adapter_typediff --git a/backend/prompt_studio/prompt_profile_manager/migrations/0006_alter_profilemanager_x2text.py b/backend/prompt_studio/prompt_profile_manager/migrations/0006_alter_profilemanager_x2text.pydeleted file mode 100644index e34d26796..000000000--- a/backend/prompt_studio/prompt_profile_manager/migrations/0006_alter_profilemanager_x2text.py+++ /dev/null@@ -1,27 +0,0 @@-# Generated by Django 4.2.1 on 2024-02-23 17:02--import django.db.models.deletion-from django.db import migrations, models---class Migration(migrations.Migration):- dependencies = [- ("adapter_processor", "0004_alter_adapterinstance_adapter_type"),- (- "prompt_profile_manager",- "0005_removed_converter_and_added_x2text_foreign_key",- ),- ]-- operations = [- migrations.AlterField(- model_name="profilemanager",- name="x2text",- field=models.ForeignKey(- db_comment="Field to store the X2Text Adapter chosen by the user",- on_delete=django.db.models.deletion.PROTECT,- related_name="profile_manager_x2text",- to="adapter_processor.adapterinstance",- ),- ),- ]diff --git a/backend/tool_instance/tool_instance_helper.py b/backend/tool_instance/tool_instance_helper.pyindex 356d5374d..febbdc33a 100644--- a/backend/tool_instance/tool_instance_helper.py+++ b/backend/tool_instance/tool_instance_helper.py@@ -140,6 +140,7 @@ class ToolInstanceHelper:embedding_properties = schema.get_embedding_adapter_properties()vector_db_properties = schema.get_vector_db_adapter_properties()x2text_properties = schema.get_text_extractor_adapter_properties()+ ocr_properties = schema.get_ocr_adapter_properties()for adapter_key, adapter_property in llm_properties.items():ToolInstanceHelper.update_metadata_with_adapter_properties(@@ -173,6 +174,13 @@ class ToolInstanceHelper:adapter_type=AdapterTypes.X2TEXT,)+ for adapter_key, adapter_property in ocr_properties.items():+ ToolInstanceHelper.update_metadata_with_adapter_properties(+ metadata=metadata,+ adapter_key=adapter_key,+ adapter_property=adapter_property,+ adapter_type=AdapterTypes.OCR,+ )@staticmethoddef get_altered_metadata(tool_instance: ToolInstance,@@ -257,6 +265,8 @@ class ToolInstanceHelper:properties = schema_spec.get_vector_db_adapter_properties()if adapter_type == AdapterTypes.X2TEXT:properties = schema_spec.get_text_extractor_adapter_properties()+ if adapter_type == AdapterTypes.OCR:+ properties = schema_spec.get_ocr_adapter_properties()for adapter_key, adapter_property in properties.items():metadata_key_for_id = adapter_property.get(AdapterPropertyKey.ADAPTER_ID_KEY, AdapterPropertyKey.ADAPTER_IDdiff --git a/backend/tool_instance/tool_processor.py b/backend/tool_instance/tool_processor.pyindex 68c3dd10d..3d12d2055 100644--- a/backend/tool_instance/tool_processor.py+++ b/backend/tool_instance/tool_processor.py@@ -78,6 +78,7 @@ class ToolProcessor:embedding_keys = schema.get_embedding_adapter_properties_keys()vector_db_keys = schema.get_vector_db_adapter_properties_keys()x2text_keys = schema.get_text_extractor_adapter_properties_keys()+ ocr_keys = schema.get_ocr_adapter_properties_keys()if llm_keys:adapters = AdapterProcessor.get_adapters_by_type(@@ -119,6 +120,16 @@ class ToolProcessor:)schema.properties[key]["enum"] = list(adapter_names)+ if ocr_keys:+ adapters = AdapterProcessor.get_adapters_by_type(+ AdapterTypes.OCR, user=user+ )+ for key in ocr_keys:+ adapter_names = map(+ lambda adapter: str(adapter.adapter_name), adapters+ )+ schema.properties[key]["enum"] = list(adapter_names)+@staticmethoddef get_tool_list() -> list[dict[str, Any]]:"""Function to get a list of tools."""diff --git a/frontend/src/components/navigations/side-nav-bar/SideNavBar.jsx b/frontend/src/components/navigations/side-nav-bar/SideNavBar.jsxindex d8ccdca6a..d06f6bbc4 100644--- a/frontend/src/components/navigations/side-nav-bar/SideNavBar.jsx+++ b/frontend/src/components/navigations/side-nav-bar/SideNavBar.jsx@@ -132,6 +132,16 @@ const SideNavBar = ({ collapsed }) => {},{id: 3.5,+ title: "OCR",+ description: "Setup platform wide access to OCR services",+ image: EmbeddingIcon,+ path: `/${orgName}/settings/ocr`,+ active: window.location.pathname.startsWith(+ `/${orgName}/settings/ocr`+ ),+ },+ {+ id: 3.6,title: "Platform",description: "Settings for the platform",image: PlatformSettingsIcon,diff --git a/frontend/src/components/tool-settings/tool-settings/ToolSettings.jsx b/frontend/src/components/tool-settings/tool-settings/ToolSettings.jsxindex 715a12760..7799663f4 100644--- a/frontend/src/components/tool-settings/tool-settings/ToolSettings.jsx+++ b/frontend/src/components/tool-settings/tool-settings/ToolSettings.jsx@@ -21,6 +21,7 @@ const titles = {vector_db: "Vector DBs",embedding: "Embeddings",x2text: "Text Extractor",+ ocr: "OCR",};const btnText = {@@ -28,6 +29,7 @@ const btnText = {vector_db: "New Vector DB Profile",embedding: "New Embedding Profile",x2text: "New Text Extractor",+ ocr: "New OCR",};function ToolSettings({ type }) {diff --git a/frontend/src/routes/Router.jsx b/frontend/src/routes/Router.jsxindex dfe29c798..f52102ba2 100644--- a/frontend/src/routes/Router.jsx+++ b/frontend/src/routes/Router.jsx@@ -85,6 +85,10 @@ function Router() {path="settings/textExtractor"element={<ToolsSettingsPage type="x2text" />}/>+ <Route+ path="settings/ocr"+ element={<ToolsSettingsPage type="ocr" />}+ /><Route path="settings" element={<SettingsPage />} /><Route path="settings/platform" element={<PlatformSettings />} /><Route path="settings/triad" element={<DefaultTriad />} />diff --git a/unstract/tool-registry/src/unstract/tool_registry/dto.py b/unstract/tool-registry/src/unstract/tool_registry/dto.pyindex 54707a493..7479807d5 100644--- a/unstract/tool-registry/src/unstract/tool_registry/dto.py+++ b/unstract/tool-registry/src/unstract/tool_registry/dto.py@@ -91,6 +91,9 @@ class Spec:def get_text_extractor_adapter_properties_keys(self) -> set[str]:return self.get_adapter_properties_keys(AdapterTypes.X2TEXT)+ def get_ocr_adapter_properties_keys(self) -> set[str]:+ return self.get_adapter_properties_keys(AdapterTypes.OCR)+def get_adapter_properties(self, adapter_type: AdapterTypes) -> dict[str, dict[str, Any]]:@@ -118,6 +121,11 @@ class Spec:) -> dict[str, dict[str, Any]]:return self.get_adapter_properties(AdapterTypes.X2TEXT)+ def get_ocr_adapter_properties(+ self,+ ) -> dict[str, dict[str, Any]]:+ return self.get_adapter_properties(AdapterTypes.OCR)+@classmethoddef from_dict(cls, data_dict: dict[str, Any]) -> "Spec":title = data_dict.get(cls.TITLE_KEY, "")@@ -214,6 +222,7 @@ class Adapter:EMBEDDING_SERVICES_KEY (str): Constant for the key "embeddingServices".VECTOR_STORES_KEY (str): Constant for the key "vectorStores".TEXT_EXTRACTORS_KEY (str): Constant for the key "textExtractors".+ OCR_KEY (str): Constant for the key "ocr".language_model (AdapterProperties): The properties of the language modeladapter.@@ -232,6 +241,8 @@ class Adapter:EMBEDDING_SERVICES_KEY = "embeddingServices"VECTOR_STORES_KEY = "vectorStores"TEXT_EXTRACTORS_KEY = "textExtractors"+ OCRS_KEY = "ocrs"+language_models: list[AdapterProperties] = field(default_factory=list[AdapterProperties]@@ -245,6 +256,9 @@ class Adapter:text_extractors: list[AdapterProperties] = field(default_factory=list[AdapterProperties])+ ocrs: list[AdapterProperties] = field(+ default_factory=list[AdapterProperties]+ )@classmethoddef from_dict(cls, data_dict: dict[str, Any]) -> "Adapter":@@ -252,6 +266,7 @@ class Adapter:embedding_services_list = data_dict.get(cls.EMBEDDING_SERVICES_KEY, [])vector_stores_list = data_dict.get(cls.VECTOR_STORES_KEY, [])text_extractors_list = data_dict.get(cls.TEXT_EXTRACTORS_KEY, [])+ ocrs_list = data_dict.get(cls.OCRS_KEY, [])language_models = [AdapterProperties.from_dict(language_model_dict)@@ -266,8 +281,12 @@ class Adapter:for vector_store_dict in vector_stores_list]text_extractors = [- AdapterProperties.from_dict(vector_store_dict)- for vector_store_dict in text_extractors_list+ AdapterProperties.from_dict(text_extractor_dict)+ for text_extractor_dict in text_extractors_list+ ]+ ocrs = [+ AdapterProperties.from_dict(ocr_dict)+ for ocr_dict in ocrs_list]return cls(@@ -275,6 +294,7 @@ class Adapter:embedding_services=embedding_services,vector_stores=vector_stores,text_extractors=text_extractors,+ ocrs=ocrs,)def to_dict(self) -> dict[str, Any]:diff --git a/unstract/tool-registry/src/unstract/tool_registry/tool_utils.py b/unstract/tool-registry/src/unstract/tool_registry/tool_utils.pyindex e0f0f312d..22114ecf3 100644--- a/unstract/tool-registry/src/unstract/tool_registry/tool_utils.py+++ b/unstract/tool-registry/src/unstract/tool_registry/tool_utils.py@@ -248,6 +248,9 @@ class ToolUtils:text_extractors = ToolUtils.get_enabled_adapters(adapter.text_extractors)+ ocrs = ToolUtils.get_enabled_adapters(+ adapter.ocrs+ )ToolUtils.process_adapter_models(models=language_models,@@ -272,6 +275,11 @@ class ToolUtils:adapter_type=AdapterTypes.X2TEXT,schema=schema,)+ ToolUtils.process_adapter_models(+ models=ocrs,+ adapter_type=AdapterTypes.OCR,+ schema=schema,+ )return schema@staticmethod