v0.4.2

What's changed

• Automatically count table rows when previews open, reuse successful counts during the session, and clarify preview totals and pagination. #106
• Expand pasted text in the chat transcript after submission while preserving compact input-history references. #104
• Fix SOCKS proxy support by including the required socksio dependency. #105
• Update the documentation with a TabulaFlow/Codex visualization-trajectory comparison. #107

Full changelog: v0.4.1...v0.4.2

v0.4.1

What's Changed

Features

• Improve structured disambiguation guidance by @yanlinf in #97
• Improve provider request failure messages by @yanlinf in #98

Fixes

• Fix demo tab navigation under instant navigation by @yanlinf in #100

Full Changelog: v0.4.0...v0.4.1

v0.4.0

What's Changed

Features

• Add benchmark data-agent exploration and ARCS agent comparisons by @yanlinf in #89
• Add gold ambiguity evaluation to ARCS benchmark runs by @yanlinf in #90
• Support shell-aware path completion by @yanlinf in #88

Full Changelog: v0.3.0...v0.4.0

v0.3.0

What's Changed

Features

• Add managed ARCS benchmark support by @yanlinf in #82

Full Changelog: v0.2.2...v0.3.0

v0.2.2

What's Changed

Features

• Improve vLLM discovery failure messages by @yanlinf in #66

Fixes

• Silence expected PSJS evaluation failures by @yanlinf in #65
• Stabilize TUI tests under concurrent load by @yanlinf in #68
• Fix chart layout after view switching by @yanlinf in #67
• Stabilize TUI submission tests by @yanlinf in #69

Full Changelog: v0.2.1...v0.2.2

v0.2.1

What's Changed

Fixes

• Fix Neo4j schema topology introspection by @yanlinf in #63

Full Changelog: v0.2.0...v0.2.1

v0.2.0

What's Changed

Features

• Add a dynamic model catalog by @yanlinf in #1
• Drop Docker Compose from the CypherBench runtime by @yanlinf in #3
• Add end-to-end benchmark runs by @yanlinf in #6
• Enable Anthropic prompt caching by @yanlinf in #12
• Discover vLLM models in the model picker by @yanlinf in #33
• Select a single discovered vLLM model automatically by @yanlinf in #38
• Simplify LLM-off startup message by @yanlinf in #40
• Support selective CypherBench runtimes by @yanlinf in #45
• Show managed benchmark database endpoints by @yanlinf in #49
• Raise research pipeline batch default to 64 by @yanlinf in #50
• Set benchmark-specific query timeouts by @yanlinf in #53

Fixes

• Fix CypherBench runtime lifecycle and readiness probe by @yanlinf in #4
• Suppress Pydantic AI banner in CLI by @yanlinf in #7
• Flatten benchmark run output directories by @yanlinf in #8
• Fix AMBROSIA-S global IDs by @yanlinf in #10
• Wait for Beaver MySQL TCP readiness by @yanlinf in #11
• Stabilize asynchronous UI tests by @yanlinf in #44
• Stabilize model picker resize updates by @yanlinf in #46
• Fix CypherBench runtime evaluation by @yanlinf in #47
• Suppress Neo4j notifications in CypherBench runs by @yanlinf in #52
• Make Bash concurrency tests deterministic by @yanlinf in #56

New Contributors

• @yanlinf made their first contribution in #1
• @dependabot[bot] made their first contribution in #18

Full Changelog: v0.1.0...v0.2.0

v0.1.0

Initial public release of TabulaFlow, an open-source data agent built on a modular Python library.

Highlights

• Interactive data agent for querying, analyzing, and visualizing data
• Support for SQL and graph databases, files, Hugging Face datasets, Wikidata, and web pages
• Structured tables, charts, maps, and graph visualizations
• Python library for building custom data applications
• Research toolkit for text-to-SQL and data-agent experiments

Install

uv tool install tabulaflow

See the documentation to get started.