v0.7.1

Added

• Targeted pandas 2.x/3.x compatibility tests in CI.

Changed

• pandas requirement updated to >=2,<4 (supports 2.x and 3.x).
• Dropped direct pytz dependency; now only transitive via pandas 2.x.

Fixed

• Secret detection CI no longer rewrites .secrets.baseline on check-only runs.

v0.7.0

Added

• Distributed tracing: W3C traceparent propagation for correlating requests with OpenTelemetry
  • Automatic OTel context propagation when available
  • Session-level trace ID for correlation when OTel is not configured
  • Cursor.new() children inherit parent trace ID for session-wide correlation
  • Zero configuration required - works automatically

Changed

• LouieClient._get_headers() now accepts session_trace_id and traceparent parameters
• All HTTP methods (add_cell, upload_*, streaming) propagate trace context

---

Full Changelog: v0.6.2...v0.7.0

v0.6.2

[0.6.2] - 2025-12-22

Added

• Dthread naming + folders: Thread name/folder are supported across create, add, list, and upload helpers, plus notebook cursor flows.
• Thread lookup by name: get_thread_by_name() resolves threads by name via the unified identifier endpoint.
• Anonymous desktop auth (optional): Anonymous token flow for local desktop servers via /auth/anonymous, with anonymous/token client options.

Changed

• Thread listing: list_threads() now parses both data and items response shapes and supports client-side folder filtering.
• 🔥 Breaking: server and anonymous_token were removed; use graphistry_server and token instead.
• Secret detection: CI now uses a temp baseline to avoid rewriting .secrets.baseline timestamps when nothing changes.

v0.6.1

Summary

• Add structured Table AI override plumbing for LouieClient.add_cell()/create_thread() and UploadClient.upload_dataframe(), including automatic singleshot routing + dataframe hydration
• Expose new TableAIOverrides dataclass via public API and update docs/tests to promote the dataclass-first flow

Contributors

• @graphistry (squash merge)

v0.6.0 - Upload Support

Summary

• Adds complete support for uploading DataFrames, images, and binary files
• Enables natural language analysis of user data through lui() interface

Changes

New Features

• ✨ UploadClient class for handling all upload operations
• 📊 DataFrame upload with multiple formats (parquet, csv, json, arrow)
• 🖼️ Image upload support (PNG, JPEG, etc.)
• 📄 Binary file upload support (PDF, Office docs, etc.)
• 🔄 Streaming display support for uploaded content in notebooks

API Additions

• client.upload_dataframe(prompt, df) - Upload and analyze DataFrames
• client.upload_image(prompt, image) - Upload and analyze images
• client.upload_binary(prompt, file) - Upload and analyze files
• lui(prompt, df) - Natural notebook interface for uploads
• lui(df, prompt) - Reversed syntax for convenience

Files Changed

• New: src/louieai/_upload.py (core upload implementation)
• Modified: _client.py, cursor.py, streaming.py (integration)
• Added: 8 test files with comprehensive coverage
• Updated: Documentation and notebooks with upload examples

Testing

• ✅ All upload tests passing (12 tests in test_upload.py)
• ✅ Linting and type checking pass
• ✅ Backwards compatible - no breaking changes

Related

• Original implementation by @author from dev/artifact-upload branch

v0.5.7 - Fix double display in Jupyter notebooks

Fixed

• Double display in Jupyter notebooks: Fixed issue where responses were displayed twice when calling lui("query") in notebooks
  • Modified _repr_html_ to show only session metadata instead of response content
  • Response content is now displayed only once via streaming or _display() method
  • Prevents duplicate display while maintaining useful session information

v0.5.6 - Fix cursor properties to return last result

What's Changed

Fixed

• Cursor properties: lui.df, lui.text, and lui.g now correctly return the last (most recent) result instead of the first
  • This matches user expectations where the most recent result is typically most relevant
  • Affects both Cursor and ResponseProxy classes

Added

• Thread management documentation: Added comprehensive documentation for .url and .thread_id properties
• Thread Management section in notebook API docs with examples for sharing analysis results

Improved

• Security documentation: Enhanced credential storage guidelines for AI assistants
• Cursor.url property: Enhanced docstring with examples and use cases

Full Changelog

v0.5.5...v0.5.6

v0.5.5

What's Changed

Fixed

• Vibes investigation notebook: Fixed remaining table references to use underscored o365_management_activity_flat_tcook instead of hyphenated names
• Notebook output cleaning: Enhanced to handle display_data output type in addition to execute_result

Details

• All Databricks table references now use the correct underscored format
• Notebook cleaning script properly handles all output types
• Improved notebook execution reliability

Full Changelog: v0.5.4...v0.5.5

v0.5.4

What's Changed

Fixed

• Vibes investigation notebook: Fixed Databricks table references to use correct o365_management_activity_flat_tcook table name instead of incorrect hyphenated names

Full Changelog: v0.5.3...v0.5.4

v0.5.3

What's Changed

New Features

• Add Cursor.new() method for creating fresh conversation threads while preserving authentication
• Thread naming support via name parameter to organize analyses
• Share mode validation with clear error messages
• DataFrame metadata preservation in elements (id, df_id, block_id)

Bug Fixes

• Fix empty text elements issue with DatabricksAgent - ResponseProxy now checks content/text/value fields
• Fix notebook cleaning to redact sensitive data instead of removing entire outputs
• Fix pre-commit hook to recognize **** as legitimate redaction pattern

Documentation & Examples

• Update vibes tutorial notebook to use lui.new() pattern
• Add DataFrame ID extraction example
• Add demo video to documentation homepage

Infrastructure

• Add notebook build utilities (build-notebook.sh, validate-notebook.py, clean-notebook-outputs.py)
• Improve CI with concurrency control and light smoke tests
• Enhance pre-commit hook for security tools

Example Usage

# Create fresh thread with same auth
lui = louie(g, server_url="https://louie.ai")
lui("Analyze sales data")

# Start new thread for different topic  
lui_new = lui.new(name="Customer Analysis")
lui_new("Show customer segments")

# Override share mode
private_lui = lui.new(share_mode="Private")

Full Changelog: v0.5.2...v0.5.3