v0.1.5

New Features & Enhancements

Terminal User Interface (TUI)

• Provider Search: Made the provider modal searchable for quicker access and navigation.
• Judge Model Selection: Added a dedicated judge model selection option to the permissions tab within the settings page.

Core & Providers

• Smart Router System: Introduced a smart router system to dynamically manage and route requests.
• Decision Manager: Implemented a decision manager alongside support for the Jev model.
• Memory Management: Introduced a new markdown-based memory management system to handle agent context.
• Notifications: Added a robust notification system powered by a native OS driver.
• Website Connect: Rolled out website connect management capabilities.

---

Bug Fixes

Core & Providers

• Go Provider Headers: Fixed an issue with the default headers in the Go provider.
• Router Prompting: Removed specific models from the decision router prompt to ensure correct routing behavior.
• Subagent Inheritance: Fixed a bug to ensure subagents correctly inherit and retain the parent model when the router is disabled.

---

Refactors & Under the Hood

Terminal User Interface (TUI)

• Logo Update: Updated the TUI logo to reflect the new Pegg/Pagg branding.

Core & Providers

• Judge Execution: Transitioned the judge to utilize the newly implemented decision model.
• Judge Prompts: Refactored and improved the internal prompt structures used by the judge.

Architecture & Cleanup

• Brand & Repository Rename: Executed a comprehensive global refactor to update the brand and repository names.

---

Documentation & Chores

CI & Releases

• Version Bumps: 0.1.5

v0.1.4

New Features & Enhancements

Terminal User Interface (TUI)

• Hint Texts: Added hint texts to improve navigation and user guidance within the TUI.

Core & Providers

• Local Skills: Enabled loading skills directly from a project-local vesvai directory.
• Provider CLI: Added a new provider CLI command for better configuration management.
• Coding Rules: Introduced support for defining and applying coding rules.
• Built-in Rule Skill: Added a dedicated built-in skill for managing rules.
• Built-in Skillify Skill: Added a dedicated built-in skillify capability.

---

Bug Fixes

Terminal User Interface (TUI)

• Terminal Drift: Reconciled the screen rendering against terminal drift to prevent layout corruption.
• Navigation Interrupt: Fixed an issue where using ESC for navigation was incorrectly treated as an agent interrupt.
• Viewport Clipping: Ensured long chat lines are properly clipped to the viewport width.
• Chip Spacing: Added automatic spacing after rendering mentions and chips.
• Search Sorting: Fixed search result ordering so items are properly sorted based on score.

Core & Providers

• VFS Permissions: Fixed virtual file system write access for global skills and allowed write access to the project's skills directory.
• Grep Parameters: Updated the grep tool to properly accept both strings and arrays for the include parameter.
• Tool Output: Removed overly aggressive redaction on LLM tool outputs.
• Task Tool Listing: Fixed the available agents listing within the task tool.
• Error Logging: Correctly routed LSP and MCP stderr streams to the logger instead of standard output.

---

Refactors & Under the Hood

Architecture & Cleanup

• Write Access Policy: Refactored file system access to include the rules directory in the permitted write paths.

---

Documentation & Chores

Architecture & Cleanup

• README Files: Created the initial README file and performed cleanup by removing the footer and ASCII art.
• Contributing Guidelines: Added the initial contributing documentation.
• Agent Documentation: Updated the project overview and repository structure in AGENTS.md.

CI & Releases

• Vesvabot CI: Integrated vesvabot into the Continuous Integration workflow.

v0.1.3

New Features & Enhancements

Terminal User Interface (TUI)

• Rendering Optimization: Introduced incremental chat rendering and coalesced redraws for a smoother and more performant chat experience.
• Settings System Tab: Added a dedicated System tab to the settings page for easier configuration management.
• Update Modal Enhancements: Upgraded the update modal with complete state management and robust error handling.

---

Bug Fixes

Core & Providers

• Streaming: Fixed an issue regarding the agent streaming deadline.
• Version Comparison: Corrected the logic used for comparing application versions.
• Zen Provider: Updated the provider to use dynamic random headers and resolved tier gate issues by correctly passing gate tools and forcing streams.

---

Refactors & Under the Hood

Architecture & Cleanup

• TUI Cache: Refactored the internal caching mechanism for the Terminal User Interface to improve stability.
• Codebase Cleanup: Removed unused and obsolete comments across the codebase for better maintainability.

---

Documentation & Chores

CI & Releases

• Documentation Updates: Fixed the site name configuration in the docs and repaired the documentation builder in the CI pipeline.
• Version Bumps: Tagged and released application versions 0.1.2.1 and 0.1.3.

v0.1.2

New Features

• Terminal User Interface (TUI): Introduced a new permission settings page.
• Context Management: Implemented session context compaction to optimize memory and context limits.
• System Reminders: Added comprehensive system reminders covering task tracking, todowrite actions, and general usage.

Bug Fixes

• Chat Interface: Corrected the "load more" functionality within the TUI so that historical items are properly prepended to the chat view.
• Tool Execution: Fixed an issue to ensure tool call arguments are strictly preserved when handling multi-part content.

Refactoring & Improvements

• Syntax Highlighting: Re-engineered TUI syntax highlighting to utilize Chroma lexers along with caching for improved performance.
• Error Handling: Enhanced the TUI to display tool errors in an expandable format for better readability.
• Architecture: Refactored the internal handling and routing of system reminders.

v0.1.1

New Features

Core & Agents

• Agent Reasoning: Added reasoning effort support for agents and saved reasoning efforts directly into sessions.
• Subagents Enhancements: Introduced session-based subagents. Background agents can now trigger reminders, and parent agents will wake upon background subagent completion.
• Session Management: Implemented an automated session title generator that can inherit titles from parent sessions. Added support for saving todos based on the active session.
• Plan Mode: Added plan enter and exit tools. Restricted plan mode to allow operations only in the plans directory, and attached a plan mode reminder to the planner agent.
• Workspace: Added support for reading the .vesvai folder and writing to the plans folder. Introduced AGENTS.md file support.

Tools & Plugins

• New Tools: Added ask, batch, delete file, and loadskill tools.
• Tool Enhancements: The read tool now supports reading directories. Added fuzzy search and attachment folder mentions in the picker. Custom design implemented for write, edit, and todo tools.
• Plugin System: Introduced a complete plugin architecture, including a CLI toggle command and a dedicated settings page tab for plugins.
• Middleware: Implemented permission and retry middlewares. Enhanced command execution security with Windows support and added prompts to ask for permission before executing plan mode tools.

Terminal User Interface (TUI)

• Design & Styling: Added popular color themes, boxed user messages, and syntax highlighting for the bash tool. Styles are now saved directly in the configuration.
• Usability: Added the ability to interrupt the agent using the ESC key. Added a subagent tool card and display usage statistics when a session is loaded back.
• Settings Interface: Introduced dedicated settings tabs for Rules and Plugins.

API, SDK & Server

• Go SDK: Released the official Go SDK.
• Servers: Added an HTTP Server and an ACP (Agent Control Protocol) Server.
• LLM Integration: Made LLM provider sync asynchronous and introduced dynamic, provider-based system prompts. Updated models' data URLs and structure.

---

Bug Fixes

TUI Fixes

• Fixed rendering corruption during live streaming and theme changing.
• Merged fragmented assistant messages and tool errors to clean up the chat view.
• Fixed duplicate error messages and trimmed leading blank lines in rendered messages.
• Ensured the TUI honors the submit hook and unlocks theme application correctly.
• Reset the recorder session when switching sessions in the TUI.
• Fixed the display of written file contents and tool usages upon session load.

Core Fixes

• Session & History: Fixed session reload fragmentation and ensured partial history is returned on interrupted runs.
• Tool Execution: Gave the orchestrator agent access to the bash tool. Fixed a broken bash tool test and resolved the CLI web tool dependency.
• Providers: Fixed Zen provider default headers and updated them to include the User-Agent.
• Middleware & Plugins: Deduplicated agent middleware and correctly routed plugin logs through the central logger. Fixed usage token counts accumulation.
• Skills: Fixed built-in skills loading by materializing them directly to the skills directory.

---

Refactors & Under the Hood

Performance & Architecture

• Optimization: Improved application startup time, auto-selected session models, and optimized JSON unmarshaling.
• Parallelization: Implemented goroutines for parallel LLM provider syncing.
• Dependency Injection: Threaded cache dependency through the DI chain.
• Security: Delegated the escape policy entirely to the permission middleware via an access hook.

Prompts & Tool Improvements

• Refactored prompts and descriptions for core tools (bash, edit, glob, grep, write, ask, read, list, web) to improve agent understanding and accuracy.
• Standardized the structured output for tools: used XML tags for LSP diagnostics and web search outputs, and allowed the todo tool to take JSON directly.
• Updated the list tool with ignore support and a tree formatter.
• Removed the wait and message tools from subagents, extending the task tool instead.
• Expanded messages with tool calls instead of loading direct skill content, and converted embedded built-in skills to use the prompt builder.
• Provided history context to the judge and structured the judge output.

---

Documentation & Chores

• Added comprehensive vesvai and Go SDK documentation.
• Implemented a pre-commit hook for formatting in CI.
• Various code style fixes and normalized picker usage/tests.
• Made uniseg a direct library dependency.
• Ensured CLI commands are idempotent.

v0.1.0

New Features

• CI/CD Pipeline: Implemented a new GoReleaser workflow for automated releases.

Continuous Integration & Testing Fixes

• Test Stabilization: Fixed CI test failures caused by git detached HEAD states and bash timeouts.
• GoReleaser Configuration: Added a before hook to ensure go mod tidy runs prior to the build process.
• Formatting Automation: Fixed the fmt step during checkout and introduced auto-commit for code formatting changes. Added proper references to the fmt workflow.
• Dependency Tracking: Removed go.sum from .gitignore to ensure dependencies are properly tracked in the CI environment.
• Workflow Cleanup: Removed leftover workflow testing commands.

Chores & Maintenance

• Project Initialization: Initial commit and base repository setup.
• Configuration Updates: Replaced the deprecated format configuration with the formats array in the GoReleaser setup.
• Code Quality: Applied general code style fixes.