v0.2.2 — Multi-dataset Streamlit UI support

Quaero v0.2.2

This patch release brings the Streamlit UI closer to Quaero's real framework capabilities by adding support for URL-based runs and multi-dataset project execution.

Added

• single-dataset runs from direct dataset URLs in the Streamlit UI
• multi-dataset project runs directly from the Streamlit UI
• per-dataset source selection in the UI:
  • uploaded file
  • direct URL
• a module-friendly config-driven wrapper:
  • run_project_pipeline(config_path, project_root)

Changed

• the Streamlit UI now builds a stable project_config.json for multi-dataset runs
• multi-dataset UI runs now delegate to Quaero's official config-driven pipeline path
• uploaded UI source files for multi-dataset runs are stored under the project folder and excluded from git
• generated marts in the UI are now rendered from mart_paths correctly

Testing

Added or extended coverage for:

• config-driven wrapper delegation
• Streamlit UI single-dataset URL mode
• Streamlit UI multi-dataset mode rendering

Current CI status:

• 89 passed

Notes

• no architecture changes
• no breaking changes to CLI or existing pipeline behavior
• this release closes the previous gap where Quaero supported multi-dataset projects in the framework but not yet in the local UI

Quaero v0.2.1

Quaero v0.2.1

This patch release fixes the post-release CI issue affecting the Streamlit UI test path.

Fixed

• corrected the Streamlit UI smoke test to use the proper AppTest access pattern
• restored clean CI status after the v0.2.0 release commit

Notes

• no framework architecture changes
• no pipeline behavior changes
• no changes to the public feature set introduced in v0.2.0

v0.2.0 — Streamlit UI, module wrapper, and optional LLM fallback

Quaero v0.2.0

This release introduces a local Streamlit UI, a module-friendly pipeline wrapper, and an optional LLM fallback for ambiguous inference.

Added

• Streamlit UI in app/ui.py
  • upload a CSV or Parquet file
  • enter a business question
  • run the pipeline locally from a simple interface
• Streamlit theme config in .streamlit/config.toml
• Importable wrapper in app/main.py:
  • run_pipeline(dataset_path, source_name, question, project_root)
• Optional Claude-backed inference fallback in:
  • kpi_engine/llm_inference.py

Changed

• kpi_engine/metric_inference.py
  • deterministic inference remains the default
  • when inference is ambiguous and QUAERO_ANTHROPIC_API_KEY is set, Quaero can use an LLM fallback for metric or dimension inference
  • returned columns are schema-validated before acceptance
  • deterministic error behavior is preserved if the fallback fails
• Documentation updated across:
  • README.md
  • REPRODUCIBILITY.md
  • ARCHITECTURE.md
  • PRD.md

Testing

Added coverage for:

• module wrapper behavior
• optional LLM fallback paths
• Streamlit UI smoke testing

Current test status:

• 85 passed, 1 skipped

Notes

• The LLM fallback is optional and disabled by default.
• The fallback uses Claude 3 Haiku via the Anthropic API.
• Enable it by setting:
  • Windows: set QUAERO_ANTHROPIC_API_KEY=your_key_here
  • macOS/Linux: export QUAERO_ANTHROPIC_API_KEY=your_key_here
• The Streamlit UI runs locally with:
  • streamlit run app/ui.py