eosphoros-ai/DB-GPT

This system is an open-source AI agent framework that enables the creation, orchestration, and management of multi-agent workflows and data analysis applications. It provides a modular infrastructure for building agents with specialized skills, integrating with diverse data sources via connectors, and executing complex reasoning tasks through a visual workflow language. The platform supports local and cloud deployment with built-in database management, knowledge base retrieval, and a web-based interface for application configuration and interaction.

23 changes.

The project was restructured into a Python monorepo using uv, accompanied by the introduction of a standardized development environment with VS Code Dev Containers and comprehensive linting tools. Significant effort was directed toward Docker automation, creating all-in-one images with bundled databases and utility scripts for streamlined local and cloud deployment. The period also saw the initial release of the DB-GPT-Web frontend and the expansion of example suites for AWEL workflows, agents, and data processing pipelines.

16 changes.

This period focused on establishing the foundational database schema and expanding the RAG SDK with comprehensive examples, integration tests, and GraphRAG capabilities. It also introduced a native AI application framework featuring visual DAG flows and agent-based app creation, supported by new UI components and internationalization tooling.

19 changes.

The project restructured its codebase into distinct modular packages (core, serve, client, app, ext) to support a new SKILL mechanism for agents and multi-agent orchestration. This period also introduced comprehensive database schema migrations for v0.6.1 through v0.8.2, adding support for code graph analysis, scheduled tasks, and enhanced data storage. Additionally, developer experience was improved with a new CLI, SDK, and containerized development environment.

Introduces a new-task module that coordinates the reset and navigation logic for starting a new task. The \NewTaskCoordinator\ ensures that when a new task is initiated, the current task session is reset before navigating to the canonical home, handling edge cases like concurrent commands, navigation failures, and guard-based cancellations. React hooks (\NewTaskProvider\, \useStartNewTask\, \useNewTaskOwner\, \useNewTaskGuard\) provide the integration layer for the UI to trigger resets and control navigation behavior.

web/modules/new-task · high confidence

The \pilot/meta\_data\ area now includes the standard Alembic configuration files (\alembic.ini\, \env.py\, \script.py.mako\, and \README\) required to manage database schema migrations. This setup initializes the migration environment for the project's metadata storage, enabling version-controlled database changes.

_pilot/meta\data · high confidence

The Docker directory now includes a build\_all\_images.sh script that automates the creation of Docker images by sequentially invoking base and allinone build scripts, and provides an example my.cnf file for MySQL configuration with UTF-8 support and specific security settings.

docker · high confidence

New utility scripts have been added to the metadata examples to migrate chat history and connection configuration data from DuckDB databases to MySQL or SQLite. Users can now easily transfer their existing local metadata stored in DuckDB (specifically \chat\_history.db\ and \connect\_config.db\) to more persistent or compatible relational databases, facilitating smoother data management and persistence upgrades.

docker/examples/metadata · high confidence

Added example test files for GraphRAG and Transformers story content to the examples/test\_files directory. The new graphrag-mini.md file contains structured entity and relationship data for TuGraph and DB-GPT project ecosystems, while graphrag-test.md provides a comprehensive overview of the DB-GPT framework, including its architecture, features, and supported models. Additionally, tranformers\_story.md was added to provide a detailed chronological narrative of the Transformers universe, serving as test data for text processing or generation capabilities.

_examples/test\files · high confidence

Users can now load example SQL data into a local SQLite database on both Windows and Linux environments. New batch (load\_examples.bat) and shell (load\_examples.sh) scripts in the examples directory automate the process of executing SQL files against a default or custom database file, ensuring cross-platform support for setting up example datasets.

scripts/examples · high confidence

The \docker/examples/excel\ directory now includes two sample CSV files, \Walmart\_Sales.csv\ and \zx.csv\, which serve as data sources for the Excel-related examples. \Walmart\_Sales.csv\ provides weekly sales data for multiple stores, while \zx.csv\ contains a renovation budget breakdown.

docker/examples/excel · high confidence

New Python scripts (\test\_case\_mysql\_data.py\ and \test\_case\_sqlite\_data.py\) have been added to the dashboard example directory to generate and populate sample data. These scripts create \user\ and \transaction\_order\ tables and insert randomized records (including user demographics and order transactions) into MySQL or SQLite databases, enabling users to quickly set up test environments for the dashboard without needing external data sources.

docker/examples/dashboard · high confidence

The dbgpt-core package has been restructured into a modular layout (v0.8.2), introducing lazy loading for core sub-libraries (core, rag, model, agent, datasource, vis, storage, train, serve) via the main dbgpt namespace. This release includes a new Claude-style SKILL mechanism for agents, allowing skills to be defined in Markdown files and automatically applied to prompts. It also adds a new AgentManager for registering and retrieving agents, and introduces a Config class that parses environment variables for various proxy API keys (e.g., Zhipu, Wenxin, Gemini) and local database SSL verification. The package now requires Pydantic 2.x and includes a new dbgpt-accelerator module for hardware compatibility handling.

packages/dbgpt-core · high confidence

Users can now see interface text in French. This change adds automatically generated French translation files (.po) for the core platform, CLI, app configuration, knowledge, operators, scenes, client, and datasource modules, covering UI strings for AWEL flows, knowledge graph operations, RAG operators, and database connection settings.

i18n/locales · high confidence

The \assets/schema/dbgpt.sql\ file has been added, establishing the foundational database structure for the application. This schema defines the core tables required for operation, including \knowledge\_space\ and \knowledge\_document\ for managing knowledge bases, \chat\_history\ and \chat\_history\_message\ for storing conversation logs, \connect\_config\ for database connection settings, and tables for plugin management (\my\_plugin\, \plugin\_hub\) and prompt handling (\prompt\_manage\). It also includes \alembic\_version\ to support database migrations.

assets/schema · high confidence

Introduces the standalone DB-GPT-Web frontend, a Next.js-based chat UI for DB-GPT. This change adds the complete source code structure including configuration files (ESLint, Prettier, environment templates), internationalization setup, and a comprehensive API client layer for interacting with backend services (chat, knowledge, flows, observability, and evaluation). It also includes core React components for the user interface, such as agent plugin management and context providers.

web · high confidence

The docker/allinone directory now provides a complete, self-contained deployment unit. The new Dockerfile builds an image based on the base dbgpt image, installing Dashscope, MySQL server, and necessary dependencies, while configuring MySQL for remote access and initializing the database schema via entrypoint scripts. This allows users to run the entire application stack with a single image, including the database, rather than relying on external database services.

docker/allinone · high confidence

This change introduces the new \dbgpt-app\ Python package (version 0.8.2), which provides a dedicated CLI entry point (\dbgpt webserver\) for starting the DB-GPT application. The package includes a custom Hatch build hook to bundle built-in skill templates (such as \csv-data-analysis\ and \financial-report-analyzer\) and example data files into the wheel, ensuring they are available to users after installation. On first startup, the application automatically provisions these built-in skills into the user's \\~/.dbgpt/skills/\ directory and initializes the database storage.

packages/dbgpt-app · high confidence

This release adds the \dbgpt-client\ Python package (version 0.8.2), providing a new asynchronous HTTP client for the DB-GPT API v2. The SDK includes a command-line interface (CLI) for running AWEL flows and a comprehensive set of client modules for managing applications, datasources, evaluations, flows, and knowledge spaces. It also introduces a \ChatCompletionRequestBody\ schema that supports chat modes like \chat\_dashboard\ and handles reasoning parameters, along with tests to ensure correct HTTP method usage (e.g., POST for creation, PUT for updates with UIDs).

packages/dbgpt-client · high confidence

The new dbgpt-ext package (version 0.8.2) provides a centralized location for external integrations, starting with a catalog of MCP server connectors for services like Feishu, DingTalk, GitHub, and Notion. It also introduces new data source connectors for graph databases (Neo4j, TuGraph), analytical engines (Apache Spark, Apache Hive), and cloud data warehouses (Alibaba Cloud MaxCompute, GaussDB), alongside updated connectors for Clickhouse, Doris, and DuckDB.

packages/dbgpt-ext · high confidence

The new dbgpt-serve package provides the core serving infrastructure for the application's multi-agent system. It introduces an AppManager and controller that handle the lifecycle of GPTS applications, including conversation management, memory persistence (using database-backed plans and messages), and agent execution. The package also includes a suite of specialized expandable agents—such as AnomalyDetection, VolatilityAnalysis, and ReportGeneration—along with their corresponding actions, enabling the system to perform complex, multi-step data analysis workflows and link to other applications.

packages/dbgpt-serve · high confidence

The application creation and editing interface now supports a new 'Native data AI application framework' (based on AWEL+AGENT). Users can create apps using either a visual DAG flow layout (via the new DagLayout component) or a single/auto-plan agent configuration (via the new AgentPanel component). The AgentPanel allows configuring LLM strategies and managing dynamic resources, while the AppModal orchestrates these modes. Additionally, a fix ensures that resource parameter values are correctly echoed when editing existing applications.

web/components · high confidence

The examples/awel directory now includes several new demonstration files that showcase AWEL capabilities. awel\_flow\_ui\_components.py adds UI-focused flow operators (select, cascader, checkbox) for building interactive flows. simple\_dag\_example.py and simple\_chat\_dag\_example.py provide basic HTTP-triggered DAGs for simple text processing and LLM chat. simple\_chat\_history\_example.py demonstrates multi-turn chat with conversation history management. simple\_llm\_client\_example.py shows LLM client usage including token counting. data\_analalyst\_assistant.py provides a code assistant example with multiple commands (fix, optimize, explain, comment, translate). simple\_nl\_schema\_sql\_chart\_example.py demonstrates schema linking for natural language to SQL. simple\_rag\_rewrite\_example.py and simple\_rag\_summary\_example.py show RAG operators for query rewriting and document summarization.

examples/awel · high confidence

The \docker/base\ directory now includes a new \Dockerfile\ and several shell scripts (\build\_image.sh\, \run\_sqlite.sh\, \run\_sqlite\_proxyllm.sh\) that streamline building and running the application. The Dockerfile supports configurable Python versions (default 3.11), optional Tsinghua mirrors for faster package installation, and specific extras for features like SQLite, ChromaDB, and various LLM proxies. The new scripts provide ready-to-run configurations for SQLite-based local databases and proxy LLM integrations (e.g., OpenAI-compatible APIs), simplifying initial setup and testing for users.

docker/base · high confidence

The examples/rag directory now includes a comprehensive suite of runnable scripts demonstrating the new RAG SDK capabilities and AWEL (Agentic Workflow Expression Language) operators. Users can explore various retrieval strategies through dedicated examples, including BM25, cross-encoder reranking, document tree retrieval, keyword search, metadata filtering, and GraphRAG with TuGraph integration. The collection also features database schema retrieval examples (both simple and AWEL-based), query rewriting, summary extraction, and retriever evaluation metrics (HitRate, MRR, Similarity). Additionally, simple RAG embedding and retriever examples showcase how to wire these components into HTTP-triggered AWEL DAGs for local development and debugging.

examples/rag · high confidence

Added three new example scripts in the SDK directory demonstrating how to build data processing pipelines using the AWEL (Agent Workflow Execution Language) framework. The examples include a simple LLM interaction, a SQL generation workflow that retrieves database schema and executes queries, and a more complex chat data pipeline featuring vector store integration, branching logic, and custom operators for data transformation.

examples/sdk · high confidence

The DB-GPT Agent framework now supports a modular SKILL system, allowing agents to load and manage predefined capability packs for modularity and reusability. This release introduces the core SKILL infrastructure (including Skill, SkillBuilder, SkillManager, and SkillLoader classes) and provides integration guides for enabling skills in existing agents. Additionally, it ships with example skills: a CSV/Excel data analysis skill that generates interactive ECharts reports, a financial report analyzer skill for extracting and visualizing key financial metrics from PDFs/text, and an agent-browser skill for headless browser automation with accessibility tree snapshots.

(repo-wide) · high confidence

Added SQL initialization scripts for \docker/examples/sqls\ covering three database engines: MySQL, SQLite, and Vertica. These files provide schema definitions and sample data for a student management system (students, courses, scores), an e-commerce system (users, products, orders), and a wide-table order structure, along with corresponding test case metadata and a MySQL root user configuration script.

docker/examples/sqls · high confidence

Developers can now use VS Code's Dev Containers extension to launch a pre-configured, containerized development environment based on the eosphorosai/dbgpt-full image. This setup automates the installation of Python 3.11, uv, and development dependencies, configures a Chinese locale (zh\_CN.UTF-8), and sets up a Zsh environment with Oh My Zsh, autojump, and SSH agent management. It also handles host-user permission mapping to prevent file ownership issues and provides a README with instructions for initialization and starting the web server.

.devcontainer · high confidence

The \examples/agents\ directory now includes a comprehensive set of runnable agent examples, orchestrated by a new \examples/\_\main\\_.py\ entry point. Users can now explore and test various agent capabilities, including single-agent interactions (Code, SQL, Summary), multi-agent workflows (AutoPlan, AWEL Layout), and specialized tool integrations (Custom Tools, MCP, AutoGPT Plugins, Sandbox Code Execution, and Claude-style Skills).

examples/agents · high confidence

Added a suite of new React components in the app construction extra configuration area to support specific team modes. The AutoPlan module (index.tsx, DetailsCard.tsx, ResourcesCardV2.tsx) enables selecting agents, configuring LLM strategies and prompts, and managing resources with a V2 API. The NativeApp component handles native app settings like chat scenes and model parameters. The AwelLayout component allows selecting and previewing AWEL workflows. Additionally, RecommendQuestions provides a dynamic list for suggested user queries, and config.tsx defines icons for agents and resource types.

web/pages/construct/app/extra · high confidence

Added new React components for the application workspace: AppCard and TabContent in the new-components directory handle displaying and interacting with app lists (including navigation to chat and collection features), while the create-app-modal in the construct/app/components directory provides a modal interface for creating and editing applications with a custom work-mode selector and updated styling.

web/new-components, web/pages/construct/app/components · high confidence

Added seven new Python scripts in the examples/client directory that demonstrate how to use the dbgpt client for common operations. These include CRUD examples for apps, datasources, flows, and knowledge spaces/documents, a chat example showing normal and streaming interactions, and an evaluation example for RAG recall and app answer metrics. The examples also feature an OpenAI-compatible chat client supporting various chat modes and reasoning content display.

examples/client · high confidence

This change introduces the tooling and documentation required to maintain internationalization across the project's packages. It adds a Makefile to automate the extraction of translatable strings (POT), generation and merging of translation files (PO), and compilation into runtime formats (MO) for supported languages (zh\_CN, ja, ko, fr, ru). It also includes a Python utility that uses an LLM to automatically generate or optimize translations for missing or existing strings, alongside comprehensive English and Chinese guides for developers to follow during code submissions.

i18n · high confidence

A new \scripts/install/install.sh\ script and its supporting libraries (\common.sh\, \profiles.sh\) provide a streamlined, one-line installation method for DB-GPT. Users can now install the application by piping the script into bash, which automatically detects the OS, installs the \uv\ package manager, clones the repository, and configures the environment based on selected deployment profiles (OpenAI, Kimi, Qwen, MiniMax, GLM, custom, or default). The installer supports non-interactive execution via flags, allows reusing existing local repositories, and handles API key injection for the chosen provider.

scripts/install · high confidence

This change introduces several new automation scripts to the \scripts/\ directory to streamline development and deployment workflows. \build\_web\_static.sh\ automates the compilation of web assets and copies them to the static directory, handling environment file preservation. \setup\_autodl\_env.sh\ provides a one-command setup for the AutoDL cloud platform, configuring Conda, cloning repositories, and installing dependencies. \llama\_cpp\_install.sh\ simplifies the installation of \llama-cpp-python\ by detecting CUDA and CPU instruction set support (AVX/AVX2/AVX512) to select the appropriate binary index. \run\_llm\_benchmarks.sh\ offers a quick way to execute LLM performance benchmarks with configurable input/output lengths and parallelism. Finally, \models\_dev\_sync.py\ automates the synchronization of the model catalog by fetching data from models.dev and merging it with hand-curated provider entries, while \update\_version\_all.py\ provides an interactive tool for updating version numbers across the monorepo's configuration files.

scripts · high confidence

Developers can now use a standardized, containerized development environment via VS Code Dev Containers, which automatically configures Python 3.11, Ruff for formatting and linting, and mypy for type checking. This change introduces a \.devcontainer.json\ configuration, a \Makefile\ for managing the virtual environment and running tests, and dedicated configuration files for \.flake8\, \.isort\, and \.mypy\. Additionally, pre-commit hooks are configured to enforce code quality standards before commits, and a \.dockerignore\ file is added to optimize Docker build contexts.

(repo-wide) · high confidence

A change to existing behaviour in pilot/examples — 1 commit, 1 file.

pilot/examples · low confidence · unverified

A change to existing behaviour in pilot/benchmark\_meta\_data — 3 commits, 1 file.

_pilot/benchmark\_meta\data · medium confidence · unverified

The upgrade script for version 0.5.2 now includes the full SQL definition for the v0.5.1 database schema (in \v0.5.1.sql\). This ensures that the upgrade process has a complete baseline of existing tables—such as \knowledge\_space\, \chat\_history\, and \plugin\_hub\—to correctly apply migrations or verify state, rather than relying on external or missing schema definitions.

_assets/schema/upgrade/v0\_5\2 · medium confidence

The upgrade script for version 0.5.10 modifies the \knowledge\_space\ table by adding a new \domain\_type\ column (varchar(50)) to support categorizing knowledge spaces, such as for financial report analysis. This change is part of the broader schema migration that also includes the full definition of the v0.5.9 database structure, ensuring the database is ready for the new feature requirements.

_assets/schema/upgrade/v0\_5\10 · high confidence

The upgrade script for version 0.5.1 adds an \error\_message\ column to the \dbgpt\_serve\_flow\ table to store error details. This location also includes the full baseline schema for v0.5.0, establishing the initial database structure including tables for knowledge spaces, documents, chat history, plugins, and prompts.

_assets/schema/upgrade/v0\_5\1 · high confidence

This release includes database migration scripts for versions 0.6.1, 0.6.2, and 0.7.0. The v0.6.1 migration adjusts column types and nullability in the \knowledge\_document\, \document\_chunk\, and \chat\_history\_message\ tables. The v0.6.2 migration modifies nullability constraints on \sys\_code\ and \app\_code\ columns in the \gpts\_app\_collection\, \recommend\_question\, and \user\_recent\_apps\ tables. The v0.7.0 migration adds timestamp and extended configuration columns to \connect\_config\, introduces a new \dbgpt\_serve\_model\ table for persisting model worker information, and updates the \user\_code\ column in \recommend\_question\.

(repo-wide) · high confidence

The database schema for version 0.7.4 has been updated to include new tables supporting LLM benchmarking (\evaluate\_manage\, \benchmark\_summary\), knowledge management (\knowledge\_space\, \knowledge\_document\, \document\_chunk\), chat history (\chat\_history\, \chat\_history\_message\), user feedback (\chat\_feed\_back\), plugin management (\my\_plugin\, \plugin\_hub\), and prompt configuration (\prompt\_manage\). These changes ensure the metadata database structure aligns with the new capabilities introduced in this release.

_assets/schema/upgrade/v0\_5\_6, assets/schema/upgrade/v0\_5\_7, assets/schema/upgrade/v0\_5\_9, assets/schema/upgrade/v0\_7\4 · high confidence

The database schema for version 0.8.0 has been updated to support new features including conversation sharing, knowledge management, and plugin integration. The upgrade script introduces a new \share\_links\ table to store conversation share link tokens, and the full schema defines tables for knowledge spaces and documents (\knowledge\_space\, \knowledge\_document\, \document\_chunk\), database connection configurations (\connect\_config\), and a plugin ecosystem (\my\_plugin\, \plugin\_hub\). Additionally, chat history storage has been refactored into separate \chat\_history\ and \chat\_history\_message\ tables to better structure conversation records and details, alongside new tables for user feedback (\chat\_feed\_back\) and prompt management (\prompt\_manage\).

_assets/schema/upgrade/v0\_8\0 · high confidence

This release includes database schema changes to support new capabilities and fix existing limitations. It introduces three new tables to enable scheduled chat replay tasks (\dbgpt\_serve\_scheduled\_task\, \dbgpt\_serve\_scheduled\_run\) and persist Model Context Protocol (MCP) connector instances (\connector\_instance\). Additionally, it resolves a data truncation issue by expanding the \content\ column in the \gpts\_messages\ table to \longtext\. The full schema definition for v0.8.1 is also provided for fresh installations.

_assets/schema/upgrade/v0\_8\1 · high confidence

The v0.8.2 database schema adds support for multi-file upload analysis and code graph capabilities. A new \dbgpt\_session\_file\ table enables private persistence for owner-bound session and task files, including metadata like storage URIs and inspection results. The \knowledge\_space\ table gains an \index\_methods\ column to store JSON strings of selected index methods for agentic workflows. Additionally, three new tables—\code\_graph\_vertex\, \code\_graph\_edge\, and \code\_graph\_meta\—are introduced to store AST-extracted code nodes, their structural relationships, and per-knowledge-space graph metadata, enabling code intelligence features.

_assets/schema/upgrade/v0\_8\2 · high confidence

The upgrade path to version 0.6.0 now includes a comprehensive SQL migration script that modifies existing tables and introduces new ones to support enhanced application management, plugin systems, and user interaction tracking. Existing tables such as \chat\_history\, \gpts\_app\, \connect\_config\, and \prompt\_manage\ receive new columns to track app codes, publication status, user identifiers, and response schemas. The migration also creates new tables including \recommend\_question\ for AI-driven suggestions, \user\_recent\_apps\ for tracking usage history, \dbgpt\_serve\_file\ for file storage metadata, \dbgpt\_serve\_variables\ for managing flow and application variables, and dedicated tables for the new plugin hub (\dbgpt\_serve\_dbgpts\_hub\) and user-installed plugins (\dbgpt\_serve\_dbgpts\_my\).

_assets/schema/upgrade/v0\_6\0 · high confidence

The upgrade script for version 0.7.1 modifies the database schema to prevent data truncation issues. Specifically, it changes the \action\_report\ column in the \gpts\_messages\ table and the \flow\_data\ column in the \dbgpt\_serve\_flow\ table from \text\ to \longtext\, allowing them to store significantly larger amounts of JSON-formatted data.

_assets/schema/upgrade/v0\_7\1 · high confidence

The sandbox package now requires explicit opt-in to use the local (host-process) execution runtime. The \RuntimeFactory\ in \packages/dbgpt-sandbox/src/dbgpt\_sandbox/sandbox/execution\_layer/runtime\_factory.py\ checks the \SANDBOX\_ALLOW\_LOCAL\_RUNTIME\ environment variable and raises a \RuntimeError\ if \LocalRuntime\ is selected or auto-detected without this flag being set. This change ensures that containerized runtimes (Docker, Podman, Nerdctl) are preferred by default, and local execution is only enabled when explicitly configured via \SANDBOX\_RUNTIME=local\ and \SANDBOX\_ALLOW\_LOCAL\_RUNTIME=true\.

packages/dbgpt-sandbox · high confidence

Added empty test file for Wenxin LLM; Added empty test package init files for kbqa and vector\_store; Added empty tests package initialization; Added integration tests for datasource connectors; Added integration tests for graph store adapters and RAG extractors; Added tests for benchmark data manager query timeout handling; Added unit tests for EmbeddingEngine document and URL processing; Added unit tests for FileBasedSkill and SkillLoader; Added unit tests for graph store operations; Added unit tests for pgvector vector store imports; Added unit tests for plugin scanning and access control.

The project has been restructured into a Python monorepo using \uv\ with workspace members for \dbgpt-core\, \dbgpt-client\, \dbgpt-ext\, \dbgpt-serve\, \dbgpt-app\, \dbgpt-sandbox\, and accelerator packages, all versioned at 0.8.2. A new Docusaurus-based documentation site (\docs/\) has been added, relying on Docusaurus 3.4.0, React 18, and Algolia search. The web frontend (\web/\) dependencies have been updated to include Ant Design 5, Next.js 13.4.7, and various visualization libraries like G6 and G2.

(dependencies) · high confidence

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

• First survey — no prior run to compare against. CAI 37.

• Code Health 63
• Architecture 54
• Maturity 73
• Readiness 22
• Security 56
• Domain Modelling 100
• Accessibility 40

• 300 commits — 162 feature/other, 138 fixes

• packages/dbgpt-core — 79 commits
• packages/dbgpt-app — 66 commits
• packages/dbgpt-ext — 38 commits
• packages/dbgpt-serve — 22 commits
• (repo) — 21 commits
• docs/docs — 12 commits
• (root) — 11 commits
• web/components — 9 commits
• packages/dbgpt-accelerator — 5 commits
• docs/static — 4 commits
• assets/schema — 3 commits
• packages/dbgpt-client — 3 commits
• packages/dbgpt_serve — 3 commits
• assets/ding.jpg — 2 commits
• docs/patchs — 2 commits
• packages/dbgpt-sandbox — 2 commits
• web/new-components — 2 commits
• web/next.config.js — 2 commits
• .devcontainer/Dockerfile.dev — 1 commit
• configs/dbgpt-proxy-tongyi.toml — 1 commit

• fix: Bugfix(RAG):handle exceptions in aload_document_with_limit results (#2712)
• fix: Fix ChatWithDbQA param validate (#2569)
• fix: Fix Clickhouse SQL syntax error for RdbmsSummary (#2651)
• fix: Fix agent retriever resource (#2719)
• fix: Fix postgresql (#2601)
• fix: Fix v0.7.5 documentation build failure with patch file (#2979)
• fix: Fix：#2859 (#2944)
• fix: Merge remote-tracking branch 'upstream/main' into fix-修复提示词模板错别字
• fix: [BUG] RCE Vulnerability in DB-GPT Plugin Upload System (#2649)
• fix: [BUG] SQL Injection through CVE Bypass in DB-GPT 0.7.0 ([CVE redacted] & [CVE redacted]) (#2650)
• fix: [Fix] Fix DuckDB datasource creation from Web UI (#3009)
• fix: chore: Fix tongyi config example (#2884)
• fix: fix connect mssql embedding error (#2589)
• fix: fix(#2995): ignore leading vis-thinking block in react parser (#2996)
• fix: fix(2965): correct the number of arguments in BuiltinKnowledgeGraph.aload_document call (#2966)
• fix: fix(ChatKnowledge):Merge sheet cells problem in Excel Knowledge (#2907)
• fix: fix(RAG): fix url document rag mode (#2874)
• fix: fix(VectorStore) fix MilvusStore to use serialize function for metadata json encoding (#2672)
• fix: fix(VectorStore): fix task concurrency and batch processing
• fix: fix(VectorStore): fix task concurrency and batch processing issue (#2671)
• …and 280 more

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

eosphoros-ai/DB-GPT was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.