Version 0.9.0

What's Changed

• Fix VegaLite retry by moving the log outside by @ahuang11 in #1069
• Propagate tool context by @ahuang11 in #1072
• Use model n ctx by @ahuang11 in #1083
• Stretch TableListAgent Tabulator by @ahuang11 in #1091
• Add create kwargs by @ahuang11 in #1089
• Add examples to analyst prompt by @ahuang11 in #1086
• Set temperature=1 for AzureOpenAI by @ahuang11 in #1087
• allow_None for SnowflakeSource.authenticator by @ahuang11 in #1092
• Add Snowflake conn_kwargs and private_key support by @ahuang11 in #1088
• Add HuggingFaceEmbeddings by @philippjfr in #1093
• Handle min max case by @ahuang11 in #1095
• Tweak prompt and increase enum count by @ahuang11 in #1094
• Update VegaLite prompting by @ahuang11 in #1078
• Refactor SQLTransforms internals with sqlglot and add SQLSample by @ahuang11 in #1085
• Various fixes by @ahuang11 in #1096
• Improve logging by @ahuang11 in #1101
• Add _llm_spec_key for granular LLM control by @ahuang11 in #1100
• Improve to work with Sonnet by @ahuang11 in #1102
• Fix parsing sql by @ahuang11 in #1106
• docs: haiku 3 to 3.5 by @ahuang11 in #1107
• Add missing length in schema by @ahuang11 in #1108
• Ensure memory callbacks and rx refs are cleaned up by @philippjfr in #1109
• Hover tabs CSS by @ahuang11 in #1112
• Refactor ListAgent and add pagination/filters by @ahuang11 in #1111
• Add Source.get_metadata by @ahuang11 in #1074
• Duplicate source prefixes by @ahuang11 in #1114
• Revert hyphen by @ahuang11 in #1113
• Prevent deadlock by @ahuang11 in #1116
• Add SQL table to dict by @ahuang11 in #1115
• SQLSelectFrom fix by @ahuang11 in #1117
• Do not normalize twice by @ahuang11 in #1119
• Refactor retry output by @ahuang11 in #1118
• Azure api version and default model by @ahuang11 in #1121
• Add missing model_kwargs by @amaloney in #1126
• Save tokens by @ahuang11 in #1124
• Add default favicon by @ahuang11 in #1122
• Pin gwalker by @ahuang11 in #1129
• Improve initial entry by @ahuang11 in #1123
• Add drop shadow by @ahuang11 in #1131
• Add status badges by @ahuang11 in #1132
• Order by closest tables by @ahuang11 in #1135
• Extract mode from model_kwargs by @ahuang11 in #1134
• Add BigQuerySource by @amaloney in #1062
• Explicit check for table by @ahuang11 in #1136
• Allow passing params to Planner thru UI by @ahuang11 in #1138
• Fix SQLSelectFrom by @ahuang11 in #1137
• Insert context by @ahuang11 in #1127
• Clean up my clean up induced bugs by @ahuang11 in #1139
• More fixes optimizations by @ahuang11 in #1140
• Add view definition to views by @ahuang11 in #1141
• Refactor lookup by @ahuang11 in #1142
• Gated Context Propagation and Dataclasses by @ahuang11 in #1143
• fix duplicate context by @ahuang11 in #1147
• Fix clean_sql by @ahuang11 in #1150
• Fix tool use and template rendering by @ahuang11 in #1151
• Fix the docs build by @maximlt in #1152
• Improve vega lite by @ahuang11 in #1148
• Add WebLLM based LLM provider by @philippjfr in #1153
• Small tweaks for prompt by @philippjfr in #1154

Full Changelog: v0.8.0...v0.9.0

v0.8.0

What's Changed

• Verify that LLM is working on load by @philippjfr in #965
• Switch to Nunito font by @philippjfr in #971
• Add docs and fixes for LLM Providers by @philippjfr in #970
• Remove LLM getting started material by @philippjfr in #972
• Add print method by @ahuang11 in #948
• Remove card layout by @ahuang11 in #975
• Fix suggestions showing by @ahuang11 in #974
• Add ability to add files to VectorStore easily by @philippjfr in #973
• Touch ups by @ahuang11 in #978
• Better sql error handling by @ahuang11 in #980
• Fixes to improve robustness by @philippjfr in #981
• Make ChatInterface configurable at UI level by @philippjfr in #982
• Small tweaks by @philippjfr in #984
• Separate context lookups in Planner by @philippjfr in #983
• Add docs list agent by @ahuang11 in #882
• Revert change to stream validity by @philippjfr in #985
• Fix regressions by @philippjfr in #986
• Fix async race condition when setting context by @philippjfr in #988
• Extract table name by @ahuang11 in #987
• Handle execution errors by @philippjfr in #990
• Fix missing arg in set_context by @ahuang11 in #991
• Remove check validity by @ahuang11 in #994
• Explicitly condense chat history by @philippjfr in #996
• ci: Add checkout to setup so it doesn't fail by @hoxbro in #995
• Fix retry by @ahuang11 in #992
• Implement generic mechanism to normalize table name by @philippjfr in #998
• Agent robustness by @philippjfr in #999
• Minor UI fixes by @philippjfr in #1000
• Ensure root conversation is synced if no exploration is created by @philippjfr in #1001
• Fix uploading only docs by @ahuang11 in #1002
• Add GraphicWalker view and integrate in AI ExplorerUI by @philippjfr in #1003
• Adjust tab location, move input row above, tweak VegaLite by @ahuang11 in #1005
• Add support for more Snowflake authenticators by @philippjfr in #1006
• Improve VegaLite and fix issue with removing count column by @ahuang11 in #1008
• Ensure extensions are loaded in notebook export by @philippjfr in #1007
• Support abs paths by @ahuang11 in #1009
• Do not show error traceback on lumen-ai serve by @philippjfr in #1014
• Share SQL in BaseViewAgent by @ahuang11 in #1013
• enh: Small enhancements to Lumen AI by @hoxbro in #1015
• UI size tweaks by @philippjfr in #1016
• Improve error reporting and load model on launch by @philippjfr in #1017
• Fix export issues by @ahuang11 in #1018
• Cherry pick changes by @ahuang11 in #1020
• Keep all steps open, and close steps card on complete by @ahuang11 in #1025
• doc: collect analytics with goatcounter by @maximlt in #1021
• Increase SQL max height by @ahuang11 in #1023
• Add help messages to SourceAgent by @ahuang11 in #1037
• Add ignore docs to gitignore by @ahuang11 in #1039
• Rename llama provider by @ahuang11 in #1038
• Add llama_url in cli by @ahuang11 in #1036
• Tweak Analyst and Planner by @ahuang11 in #1026
• Add chat loc by @ahuang11 in #1035
• Refactor SQL agent by @ahuang11 in #1010
• Rename llama url by @ahuang11 in #1040
• Move tabs to the right by @ahuang11 in #1041
• Avoid repeatedly looking up and streaming context by @philippjfr in #1043
• Fix repeat follow up queries, with no errors at all by @ahuang11 in #1046
• Simplify source/table separator logic by @philippjfr in #1047
• Update ui.py by @ahuang11 in #1048
• Fix validation error by @ahuang11 in #1050
• Save some tokens by disabling URLs in errors by @ahuang11 in #1049
• Fix for single table by @ahuang11 in #1056
• Update suggestions, tweaks to UX by @ahuang11 in #1057
• Add vega lite validation by @ahuang11 in #1055
• Fix indentation in notebook output by @ahuang11 in #1054
• Run sqlglot cleanup pass on generated SQL by @philippjfr in #1060
• ci: modify workflow to exclude forks by @Azaya89 in #1059
• Update planner rules by @ahuang11 in #1061
• Indent within step to surface VegaLite validation error by @ahuang11 in #1063
• Fix read_ regex on export by @ahuang11 in #1064
• Tweak vegalite by @ahuang11 in #1065
• Tweak analyst agent by @ahuang11 in #1066

New Contributors

• @Azaya89 made their first contribution in #1059

Full Changelog: v0.7.0...v0.8.0

v0.7.0

What's Changed

This release added the new lumen.ai subpackage to Lumen, which marks a shift away from Lumen being solely a declarative dashboarding and monitoring tool. Instead Lumen is now re-envisioned as an extensible agent based framework for chatting with data and for retrieval augmented generation (RAG). The declarative nature of Lumen's data model makes it possible for LLMs to easily generate entire data transformation pipelines, visualizations, and other many other types of output.

New Contributors

• @amaloney made their first contribution in #854

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

Version 0.6.0

This release updates Lumen for compatibility with latest Panel and Param releases. Additionally it adds a DuckDb source, focuses on a number of improvements for validation and serialization of Lumen specs.

Features

• Add DuckDb source (#366)
• Automatically support for automatically generating filters (#419, #501)
• Add Sum, Count and Eval transforms (#425)
• Add lifecycle callbacks to Config (#429, #441)

Enhancements

• Do not apply range and multi-select filters if it spans entire range (#420)
• Implement support for recursive layout specifications (#424)
• Add a shortcut for downloading a view (#427)
• Add View.title option (#430)
• Display loading indicator on View (#435)
• Improve Download with Index and Filename parameters (#438)
• Allow configuring a View data limit (#443)
• Add description to sidebar (#457)
• Sanitize string input in SQLFilter (#463)
• Add config.show_traceback (#465)
• Allow referencing Variable in Pipeline table definition (#478)
• Ensure valid key validation is enabled for all components (#493, #498)

Bug fixes

• Fix how Layouts resolve internal pipelines (#418)
• Correctly sync WidgetFilter with url (#491)
• Allow roundtrip of Layout to_spec and from_spec (#495)
• Ensure Lumen yaml paths can be correctly resolved on Windows (af86b04)

Compatibility

• Compatibility for Panel 1.0 (#449)
• Compatibility with Param 2 (#475)
• Add Python 3.12 and drop Python 3.7 and 3.8 (#497)

Version 0.5.1

v0.5.1

Version 0.5.1

Version 0.5.0

v0.5.0

Version 0.5.0

Version 0.4.1

Minor release:

• Fix filtering of Views
• Add precipitation data files (#133)

Version 0.4.0

(Relatively) major release:

New features:

• Handle errors while rendering dashboard (#131)
• Defer rendering of dashboard contents until page is rendered (#123)
• Add Melt transform (#122)
• Implement DerivedSource with ability to filter and transform existing sources (#121)
• Add caching to DerivedSource
• Use Datetime pickers (#119)

Bugfixes and minor improvements:

• Clear original source cache on DerivedSource
• Allow providing custom Download labels (#130)
• Fix handling of range filters (#129)
• Unpack panes correctly on Views (#128)
• Fixed dask kwarg on JSONSource (#127)
• Pin python3.8 in conda build env
• Ensure None on widget filter is handled (#120)
• Improve docs (#112)

Version 0.3.2

Minor release:

• Allow declaring Filter as shared (#111)
• Fix bug inserting Download selector

Version 0.3.1

Minor release:

• Updated dependencies
• Add Download options to targets (#110)
• Make editable a configurable option (#109)
• Improve docs (#107, #108)
• Gracefully handle missing dask and parquet libraries (#105)