The Data Discovery feature includes the Backend discovery service. The Backend discovery service requires the Entity extraction capability to process Danish questions. The Backend discovery service requires Table ranking to prioritize relevant tables. The Backend discovery service depends on Join discovery mechanisms to find table relationships. The backend app api discovery router is part of the Backend discovery service to provide API integration. The backend service discovery.py implements the API endpoints for the Data Discovery system.

The file backend/app/services/duckdb_service.py is part of the DUCKDB Service. The DuckDBService class is defined within backend/app/services/duckdb_service.py. Vectorize Progress Tracking obtains chunk counts from DuckDB service to compute embedding progress accurately. The extraction coordinator service depends on the DuckDB service for managing extracted text chunks and related data. The text to SQL service depends on the DuckDB service to execute generated SQL queries against the extracted data. DuckDB Service is defined in backend/app/services/duckdb_service.py.

Embedding service uses Ollama running on elin GPU for batch embedding of document chunks. Embedding service uses the nomic-embed-text 768-dimensional model via Ollama for GPU accelerated vector embedding. EmbeddingService is used by the Docling extraction system to produce GPU-accelerated embeddings for semantic chunk vectors. batch_vectorize_job depends on EmbeddingService to perform batch vectorization of extracted document chunks. BatchProcessor uses EmbeddingService to vectorize extracted data in the processing pipeline. EmbeddingService integrates with OpenClawHttpClient to use Ollama on elin GPU for embedding computations. The embedding_service.py provides centralized embedding generation by communicating with Ollama running on elin GPU. GPU-first document extraction uses the embedding service in backend/app/services/embedding_service.py which communicates with Ollama on the GPU for embeddings. The embedding service performs batch embeddings using Ollama's nomic-embed-text model on GPU, supporting GPU utilization monitoring and automatic retries. The extraction worker chains extraction results to the embedding service for batch vectorization on GPU after successful DOCX/PPTX extraction.

findings_generator logging is to be implemented in backend/app/services/findings_generator.py. backend/app/services/agent_skills.py depends on backend/app/services/findings_generator.py to generate findings after query execution in the agent workflow.

The GDPR detector service is a component of the IronClaw-powered Agent Mode feature. GdprDetector uses StorageService to scan project data for GDPR-related personal data indicators.

The IronClaw client service supports the IronClaw-powered Agent Mode implementation. The backend/app/services/ironclaw_client.py is a component of the IronClaw-powered Agent Mode. backend/app/services/ironclaw_client.py implements the IronClawClient used for remote agent communication via IronClaw Gateway API.

The file backend/app/services/qdrant_service.py is part of the QDRANT SERVICE. The QdrantService class is defined within backend/app/services/qdrant_service.py. Qdrant Service is defined in backend/app/services/qdrant_service.py.

The file backend/app/services/question_router.py is part of the QUESTION ROUTER. The QuestionRouter class is defined within backend/app/services/question_router.py. question_router logging is a specific logging to be added after SQL query execution to provide diagnostics. question_router logging is to be implemented in backend/app/services/question_router.py. question_router logging is to be implemented in backend/app/services/question_router.py. question_router logging is a specific logging to be added after SQL query execution to provide diagnostics. question_router logging is to be implemented in backend/app/services/question_router.py. question_router logging is to be implemented in backend/app/services/question_router.py. backend/app/services/question_router.py implements core logic for table schema limiting, embedding search integration, and Qwen3 re-ranking in the Multi-Stage Text-to-SQL Architecture. Integration requires modifying backend/app/services/question_router.py to incorporate consolidated views in query routing. QuestionRouter is defined in backend/app/services/question_router.py.

backend/app/services/table_catalog.py implements the TableCatalog use case with Danish translations and join hints for the Multi-Stage Text-to-SQL Architecture.

The Data Discovery feature contains the TableIndex entity. The discovery.py service uses the TableIndex for semantic table matching. Intelligent consolidation uses TableIndex for semantic table matching. The Data Discovery feature contains the TableIndex entity. The discovery.py service uses the TableIndex for semantic table matching. Intelligent consolidation uses TableIndex for semantic table matching. backend/app/services/table_index.py implements the TableEmbeddingIndex use case as part of the Multi-Stage Text-to-SQL Architecture.

The file backend/app/services/text_to_sql.py is part of the TEXT-TO-SQL Service. The TextToSQLService class is defined within backend/app/services/text_to_sql.py. The streaming responses via the /ask-stream endpoint realize the Text-to-SQL analysis query endpoint by enabling streaming of query results to avoid client timeouts. The backend service text_to_sql.py uses SQLCoder-7B as the configured default model for Text-to-SQL queries after the code change. Coolify rebuilds the backend container to deploy updated Python code including changes in text_to_sql.py. backend/app/services/text_to_sql.py implements schema compression and SQL generation with error retry for the Multi-Stage Text-to-SQL Architecture. The text to SQL service depends on the DuckDB service to execute generated SQL queries against the extracted data. The code in backend/app/services/agent_skills.py depends on backend/app/services/text_to_sql.py for SQL extraction functionality within agent workflows. Text-to-SQL Service is defined in backend/app/services/text_to_sql.py.

Visualization service is included as part of the IronClaw-powered Agent Mode feature.

The scope field replaces hardcoded text in backend/app/workers/catalog.py to generate more accurate file summaries. The LLM Prompt Injection use case modifies prompts in catalog.py to include language parameter for file summaries.

The extraction worker uses the DOCX extractor for GPU-first document extraction of DOCX files. The extraction worker uses the PPTX extractor for GPU-first extraction of PPTX files. The extraction worker chains to the batch vectorize job to process GPU embeddings after extraction. The extraction worker uses Docling as mandatory extractor for DOCX/PPTX files and fails extraction if Docling fails. The extract.py worker is modified to write extracted data using pg_data_service.py instead of DuckDBService. The extract.py worker invokes Docling-based extractors for DOCX and PPTX files and enforces a no-fallback failure policy if Docling fails. The extract.py worker runs as part of the RQ workers to process extraction jobs asynchronously. The RQ worker calls the DOCX extractor for extraction using Docling and fails hard if extraction fails, enforcing the no fallback policy. The RQ worker calls the PPTX extractor for extraction using Docling, failing hard on extraction errors without fallback. The extraction worker chains extraction results to the embedding service for batch vectorization on GPU after successful DOCX/PPTX extraction.

Database migration scripts add tables for the IronClaw-powered Agent Mode. Agent database migration corresponds to the migration script backend/migrations/003_agent_tables.sql which was applied.

The backend/requirements.txt has been updated to include Docling version 2.0.0 or higher as a mandatory dependency, while python-docx and python-pptx have been removed.

Test fixture CSV for agent personnel data, no details provided.

Test fixture CSV for sales data, no details provided.

Test fixture CSV for timeseries data, no details provided.

Backend test for agent gateway, no additional details provided.

Backend test for agent orchestration, no details provided.

Backend test for agent skills, no additional details provided.

Test suite verifying Docling installation and extraction quality on elin GPU, ensuring no fallback errors and GPU readiness. The test_docling_extractors.py test suite verifies Docling installation and extraction quality on elin, including DOCX and PPTX extraction and the no fallback behavior.

Backend test for GDPR detector, no details provided.

Documents backend crash analysis; discusses lazy-loading Qdrant in QuestionRouter to prevent startup timeouts, with fallbacks if Qdrant is unavailable.

Docker container with no exposed ports, used for backend data storage and application services. Backend service requires a volume for backend_storage to persist user uploads and DuckDB files. The backend_storage (Docker) service is defined in docker-compose.yml. The backend_storage (Docker) service is defined in docker-compose.coolify.yml.