- All notable changes to this project will be documented in this file.
- "Breaking Changes" describes any changes that may break existing functionality or cause compatibility issues with previous versions.
- Add LLM management module to reuse connection across agents.
- Add ChatAPP configuration sub-module in Agent Management.
- Enhance dimension value management sub-module.
- Enhance memory management and term management sub-module.
- Enhance semantic translation of complex S2SQL.
- Enhance user experience in Chat UI.
- Introduce LLM-based semantic corrector and data interpreter.
- support multiple rounds of dialogue
- add term configuration and identification to help LLM learn private domain knowledge
- support configuring LLM parameters in the agent
- metric market supports searching in natural language
- introducing WorkFlow, Mapper, Parser, and Corrector support jump execution
- Introducing the concept of Model-Set to simplify Domain management
- overall optimization and upgrade of system pages
- optimize startup script
- add tag abstraction and enhance tag marketplace management.
- headless-server provides Chat API interface.
- migrate chat-core core component to headless-core.
- support view abstraction to Headless.
- add the Metric API to Headless and optimizing the Headless API.
- add integration tests to Headless.
- add TimeCorrector to Chat.
- support creating derived metrics.
- Support creating metrics using three methods: by measure, metric, and field expressions.
- added support for postgresql data source.
- code adjustment and abstract optimization for chat and headless.
- rewrite Python service with Java project, default to Java implementation.
- support setting the SQL generation method for large models in the interface.
- optimization of metric market experience.
- optimization of semantic modeling canvas experience.
- code structure adjustment and abstraction optimization for chat.
- add SQL generation improvement optimization, support LLM SQL, Logic SQL, and Physical SQL display.
- add showcase functionality to support recommending similar questions.
- add frontend modification of filtering conditions and re-querying feature.
- support nested query functionality in semantic.
- support switching queries between multiple parsers in the frontend.
- optimizing the build and deployment of the project.
- overall optimization of the SQL Corrector functionality.
- fix execute error on mysql <=5.7
- add llm parser config
- add datasource agg_time option
- add function name adaptor in clickhouse
- add dimension and metric show in dsl
- update user guide doc
- update query building of plugin in default model
- update some core API constructs to keep naming consistency
- update ConfigureDemo config
- update the association mechanism so that invisible dimensions and metrics will no longer be associated
- fix hasAggregateFunction logic in SqlParserSelectHelper
- meet checkstyle code requirements
- save parseInfo after parsing
- add time statistics
- add agent
- dsl where condition is used for front-end display
- dsl remove context inheritance
- Support asynchronous query - return parse information to user before executing result
- Add Model as the basic data structure of the semantic definitions - this will repalce the old conception of subdomain
- improve knowledge word similarity algorithm
- improve embedding plugin chooser
- improve DSLQuery field correction and parser
- Fix mapper error that detectWord text is shorter than word
- Fix MetricDomainQuery inherit context
- Add function call parser and embedding recall parser
- Add plugin management
- Add web page query and web service query
- Metric filter query support querying metrics and comparing them in different dimensions
- Support dimension value mapping
- Support dimension/metric invisible, chat filter related data
- Add user guide docs
- Fix the data problem of getDomainList interface in standalone mode
- Support llm parser and llm api server - users can query data through complex natural language.
- Support fuzzy query dimension and metric name - users can set the 'metric.dimension.threshold' parameter to control the fuzzy threshold.
- Support query filter and domain filter in query and search - users can specify domainId and query filter to filter the results in search and query.
- Support standalone mode - users can integrate semantic and chat services in one process for easy management and debugging.
- Support dsl query in semantic - users can specify DSL language to query data in Semantic. In the past, data querying was limited to struct language.
- Add unit and integration testing - add integration tests for single-turn and multi-turn conversations, to efficiently validate the code.
- Support dimension and metric alias - users can specify one or multiple aliases to expand search and query.
- Add scheduled semantic metadata update functionality in chat.
- Support create datasource by table name in the web page.
- Add the ability to set permissions for domain.
- Add a local/Remote implementation to the SemanticLayer interface.
- Code architecture adjustment in chat.
- Abstracting into three modules, namely api, core, and knowledge. Providing four core interfaces: SchemaMapper, SemanticLayer, SemanticParser, and SemanticQuery.
- Add RuleSemanticQuery and LLMSemanticQuery implement to SemanticQuery.
- Add all possible queries to the candidate queries, and then select the most suitable query from the candidate queries.
- Code architecture adjustment in semantic.
- Refactor semantic layer SQL parsing code through Calcite.
- Add QueryOptimizer interface.
- Chat config subdivided into detailed and metric scenarios - users can set different parameters in these two scenarios.
- Resolved last word not be recognized in SchemaMapper.
- Fix context inheritance problem.
- Fix the error of querying H2 database by month unit.
- Set faker user to context when authentication disable.
- Add the search and query feature in chat according to rules in an extensible way.
- Add semantic/chat independent service for users.
- Add Modeling Interface - users can visually define and maintain semantic models in the web page.
- Add a unified semantic parsing layer - user can query data by struct language.
- add data portal
- add metric trend chart
- add feedback component
- add tab component
- add page setting
- modify permission process
- optimize css style
- optimize filter
- delete view module