Arctic-Text2SQL-R1-7B is a production model backed by Snowflake integration.

DataLens uses the Sonnet model strategy as the default for data analysis work. Phase 2 Strategy Research & Decision Point involves the use of Sonnet 4.6 agent for generating PHASE2_IMPLEMENTATION_PLAN.md document The Sonnet subagent builds GPU-first extractors that make Docling mandatory with no fallback options. DataLens agent uses the Sonnet model as its primary LLM for data analysis work requiring high quality.

Sonnet 4.6 created the PHASE2_IMPLEMENTATION_PLAN.md containing the go/no-go recommendation and effort versus value analysis. Opus 4.6 and Sonnet 4.6 together were used to triangulate decision for the Phase 2 strategy due to GPT-5.2 unavailability.

SOUL.md contains the core configuration and design decisions for the DataLens platform; included in the project repository. Agent configuration includes the SOUL.md file. Phase 2 research outputs include SOUL.md in the research documents.

DataLens provides source attribution tracing results back to exact files and chunks.

Structured query language used for data manipulation and retrieval in DuckDB. QuestionRouter generates SQL statements when routing structured questions via TextToSQLService and DuckDBService.

Targeted to achieve greater than 40% correctness on DABStep benchmark.

Planned to create SQL CTEs for consolidating multiple related tables, supporting complex and multi-table queries with better accuracy.

SQL extraction regex fix addresses the problem in Qwen3 response format where the SQL query was not properly captured due to missing newline before closing backticks. Safety net cleanup enhances SQL extraction regex fix by stripping explanation text markers after extraction to ensure pure SQL before execution. SQL extraction regex fix reduces SQL syntax errors by properly capturing SQL code blocks which fixed previous syntax errors related to backticks. Logging is associated with SQL extraction regex fix as it will be added to trace extraction and execution results. SQL extraction regex fix was deployed on theo. Commit 408be74 fixes the SQL extraction issue where multiple code blocks in Qwen3 response cause incorrect matching by capturing only SELECT...; patterns in code blocks. The Qwen3 response format required the SQL extraction regex to be fixed to handle SQL code blocks without a newline before closing backticks. Commit 70f724f cleaned up safety nets by stripping explanation text markers from the extracted SQL, ensuring pure SQL before execution. The fixed SQL extraction regex and cleanup removed SQL syntax errors in the logs caused by improper SQL extraction. The generation and display of findings depend on correct SQL extraction, which was fixed by the regex updates.

Ongoing development to produce accurate SQL queries from natural language, integrating multi-stage reasoning and handling larger schemas.

The SQL migration execution script is used to apply the agent migration file 003_agent_tables.sql to the PostgreSQL database on theo.

The DataLens backend depends on SQL migration files to create and update necessary database tables, including those for agent features. Alembic can be used to manage SQL migration files for database schema evolution in DataLens.

System reviews indicate no specific summaries provided; current focus is on core functionality.

SQL response parsing has been improved to handle raw SQL output from SQLCoder-7B, with enhanced error handling and support for multi-line SQL statements, ensuring correct interpretation of generated queries.

SQL extraction regex fix reduces SQL syntax errors by properly capturing SQL code blocks which fixed previous syntax errors related to backticks. Decimal type support contributes to fixing SQL syntax errors by enabling proper support for decimal types in queries. The fixed SQL extraction regex and cleanup removed SQL syntax errors in the logs caused by improper SQL extraction.

The platform prioritizes SQL-first queries to allow direct, optimized SQL execution, enhancing flexibility and performance in data analysis. The design decision to accept raw SQL queries (SQL-first queries) is implemented in DataLens Development to prioritize working end-to-end flow.

The plan includes SQLAgent that uses Text-to-SQL capabilities to generate and execute SQL queries. SQLAgent integrates with Vanna.AI for SQL generation from natural language questions. SQLAgent executes queries using DuckDB as the analytics database SQLAgent is represented as the sql agent entity SQLAgent is represented as the sql agent entity SQLAgent executes queries using DuckDB as the analytics database SQLAgent is represented as the sql agent entity SQLAgent is represented as the sql agent entity SQLAgent executes queries using DuckDB as the analytics database SQLAgent is represented as the sql agent entity SQLAgent is represented as the sql agent entity The DataLens Project uses the SQLAgent component to convert natural language to SQL queries. The DataLens Project uses the SQLAgent component to convert natural language to SQL queries. The DataLens Project uses the SQLAgent component to convert natural language to SQL queries. The plan includes the Text-to-SQL Agent integration with Vanna.AI for natural language to SQL translation. The Text-to-SQL Agent integrates with Vanna.AI to generate SQL queries from natural language. The plan includes the Text-to-SQL Agent integration with Vanna.AI for natural language to SQL translation. The Text-to-SQL Agent integrates with Vanna.AI to generate SQL queries from natural language. SQLAgent realizes the Text-to-SQL capability by converting natural language to SQL executions. SQLAgent realizes the Natural Language Querying capability converting questions into SQL queries. Lightweight Implementation uses SQLAgent for text-to-SQL functionality. DataLens Project includes SQLAgent which converts natural language to SQL using Ollama. SQLAgent executes generated SQL queries on DuckDB.

The DataLens Platform uses SQLAlchemy to interact with the PostgreSQL schema for database operations. SQLAlchemy depends on psycopg2-binary as a PostgreSQL database driver. Alembic depends on SQLAlchemy for database migrations.

The FastAPI backend in the DataLens Project uses SQLAlchemy ORM for data access. The Backend uses SQLAlchemy ORM as a third-party component for database interaction.

SQLCoder-7B is deployed on elin at port 11434, replacing qwen3-coder-next for faster Text-to-SQL, reducing inference time to 2-3 seconds. Generates valid DuckDB SQL, used by backend on theo, and supports full SVGV analysis post-deployment. Hosted as Arctic-Text2SQL-R1-7B with 7B parameters, optimized for GPU (16GB VRAM).

SQLExecutor runs generated SQL queries against PostgreSQL with retry logic and error recovery. Multi-Stage Text-to-SQL Architecture realizes the SQLExecutor use case for executing SQL with error recovery and retry.