Version 0.1.11-alpha: Gemini & Groq Integration

This release introduces support for two new major LLM providers, Google Gemini and Groq, significantly expanding the range of available models. The developer experience is also improved with the introduction of an automated CI/CD testing pipeline and a comprehensive overhaul of the test suite.

New Features & Enhancements

• Google Gemini & Groq Support: The application now integrates with the Google Gemini and Groq APIs. Users can configure their API keys to access models such as gemini-2.5-pro, gemini-2.5-flash, and Groq's moonshotai/kimi-k2-instruct. The venice provider has been removed.
• New Dependencies: To support these new features, several packages have been added, including google-genai, textual[syntax], tiktoken, pyperclip, ddgs, and markdownify.

A Closer Look at the Changes

• LLM Provider Integration:

  • Configuration for Gemini and Groq providers has been added to api_providers.json, along with corresponding updates to the default model_list.json.
  • Client initialization for the Gemini API has been implemented, with the google-genai dependency handled gracefully if not installed.
  • A new service was created to fetch available models directly from the Gemini API, supplementing the data with cost and context details from external sources.

• Streaming Logic Refactoring:

  • The core LLM streaming logic in llm_requests.py has been refactored for better modularity.
  • Provider-specific streaming implementations for OpenAI, Anthropic, and the new Gemini service are now organized in the src/jrdev/services/streaming/ directory. This architecture simplifies maintenance and makes it easier to add new providers in the future.

• CI/CD and Testing Overhaul:

  • A new GitHub Actions workflow (pytest.yml) has been added to automatically run the test suite on every push and pull request across Python versions 3.9 through 3.12. This ensures code quality and compatibility.
  • The test suite has been significantly improved by removing several outdated test scripts and adding new, robust unit tests for configuration (test_config.py), Git utilities (test_git_utils.py), and string manipulation functions (test_string_utils.py).

• Code Refinements & Bug Fixes:

  • Improved the logic for adding entries to .gitignore to ensure correct newline handling.
  • Enhanced the accuracy of token count estimation displayed in the terminal UI by rebuilding the full prompt context.
  • Updated various type hints for better code clarity and compatibility with older Python versions.

• Documentation:

  • A new AGENTS.md file provides a technical guide to the codebase architecture and development conventions, intended for both human developers and AI agents interacting with the project.
    Generated by JrDev AI using gemini-2.5-pro

Version 0.1.10-alpha: Model & Provider Management

Overview

This release introduces a major overhaul of model and provider management, centered around a redesigned settings screen that allows for dynamic fetching and importing of models directly from API providers. This version also brings significant user experience improvements, including an enhanced model selection interface, a more informative terminal that tracks context usage, and smarter Git integration.

What This Means for Users

New Model & Provider Management UI

The "Settings" screen has been completely redesigned with a new "Manage Models" view. This unified interface provides a comprehensive toolkit for managing AI models and their providers.

• Centralized Management: View all configured models in a sortable, filterable table and manage the list of API providers.
• Intuitive CRUD Operations: Add, edit, and remove both models and providers through new, easy-to-use modal dialogs.

Fetch and Import Models from Providers

A key feature of this release is the ability to dynamically update the model list from providers.

• Automatic Discovery: Use the "Fetch Models" button to retrieve an up-to-date list of available models from providers like OpenRouter, Anthropic, and other OpenAI-compatible services.
• Interactive Import: An import screen displays the fetched models, allowing users to select which ones to add to their local configuration. The system automatically filters out models that are already configured.

Enhanced Model Selection

The model selection dropdown, accessible from the Chat and Git views, has been significantly improved for a better user experience.

• Search and Filter: Quickly find models using a new search input.
• Grouped by Provider: Models are now grouped by their provider, making it easier to navigate the list.
• Improved UI: The dropdown's layout, focus handling, and overall responsiveness have been refined.

A Smarter Terminal

The main terminal interface now provides more insight and control over the agent's conversation.

• Context Usage Tracking: A new label displays the current context window usage as a percentage and token count, helping to manage long conversations.
• Conversation Management: "Compact" (🗜️) and "Clear" (🗑️) buttons have been added to condense or reset the conversation history, providing tools to control token consumption.

Improved Git Tools

The Git Tools screen has been updated for a more streamlined workflow.

• Branch & Tag Selection: The "Base Branch" field is now a dropdown menu automatically populated with all local branches, remote branches, and tags from the current repository. This eliminates the need for manual entry and reduces the chance of errors.

A Closer Look at the Changes

• Code Refinements & Improvements:

  • UI Code Reorganization: The Textual UI source code has been restructured into a more modular and maintainable directory layout (e.g., tui/chat, tui/settings, tui/git). This architectural improvement facilitates future development.
  • New Model Fetching Service: A ModelFetchService was implemented to handle asynchronous API requests to providers for model discovery. This service is extensible to support various provider endpoints.
  • New Documentation: Comprehensive documentation for the Project Context system (/init) has been added, detailing its function and its integration into the agent pipeline.

• Dependency Updates:

  • The httpx library has been added as a dependency to support asynchronous HTTP requests for the new model fetching service.
    Generated by JrDev AI using google/gemini-2.5-pro

Version 0.1.9-alpha

Version 0.1.9-alpha Release Notes

Overview

This release introduces significant enhancements to the AI's planning capabilities, major user interface improvements for better usability, and expanded functionality for agent tools. The core focus is on enabling the AI to make more informed decisions about the context it needs, while also refining the user experience with features like searchable model lists and multi-line input.

New Features & Enhancements

Searchable Model Selection Lists

The model selection dropdowns across the application have been upgraded. They now feature a search bar, allowing for quick filtering and selection of models. This is particularly useful in environments with a large number of available models, streamlining the process of switching between them in the Chat, Git, and Terminal views.

Multi-line Input in Command Bar

It is now possible to insert newlines into the main command input bar using the Shift+PageDown key combination. This allows for formatting more complex, multi-line prompts and commands directly in the UI.

Web Scrape to File

The web_scrape_url agent tool has been enhanced. It now accepts an optional second argument: a file path. This allows the agent to save the scraped content of a webpage directly to a markdown file within the project, which can be useful for documentation, analysis, or providing context for subsequent tasks.

AI & Agent Improvements

Intelligent Context Selection

The planning phase of the coding agent has been fundamentally improved. The agent can now analyze the user's request and generate a specific list of files (use_context) it determines are necessary to complete the task. This list is then used as the exclusive context for the code generation steps.

• What this means: The agent is no longer reliant on a broad, pre-selected set of files. It actively chooses its own context, leading to more focused and accurate code modifications, reduced token usage, and faster processing.
• Implementation: This was achieved by updating the create_steps prompt, instructing the AI to output a use_context list alongside the action plan. The planning pipeline now reads this list and adjusts the context accordingly.

Refined Agent Instructions

The core prompts guiding the agent's behavior have been updated for better clarity and more robust error handling. Instructions for handling user-rejected actions have been generalized, ensuring the agent clarifies the user's intent before proceeding.

Technical Refinements & Fixes

• UI Component Refactoring: The ModelListView has been refactored from a simple ListView into a composite Widget containing an Input for search and a ListView for results. A custom ModelSelected event was introduced for more robust and specific event handling across the UI.
• Tool Execution: The web_scrape_url tool is now correctly wired into the application's main tool execution loop.
• Message Builder Fix: A minor bug was corrected in the MessageBuilder where a project overview file was being added to the wrong internal set.
  Generated by JrDev AI using google/gemini-2.5-pro

What's Changed

• Fix web-scraping integration and improve context handling by @presstab in #119
• Add model search and improve model selection UX by @presstab in #120

Full Changelog: v0.1.8-alpha...v0.1.9-alpha

Version 0.1.8-alpha

Overview

This release introduces expanded capabilities for JrDev's AI agents, with a primary emphasis on enabling secure access to system tools and external resources. These enhancements allow agents to perform more complex tasks while maintaining user control and safety. Key additions include terminal access for shell commands, web search and scraping tools, and improved context management for project files. Reliability improvements in the router agent ensure more consistent interpretation of user requests, reducing errors and enhancing the overall development workflow.

What This Means for Users

Users now have greater flexibility in how JrDev assists with development tasks, particularly through the new terminal access feature. This allows the AI agent to execute bash shell commands directly from the project environment, such as running git operations (e.g., checking out branches or committing changes), managing files, or verifying directory structures. For safety, every command requires explicit user confirmation via a prompt in the UI, giving you full control to review and approve (or cancel) before execution. This means you can delegate routine terminal tasks to the agent— like creating branches or listing files—without leaving the JrDev interface, streamlining workflows while preventing unintended actions.

Additionally, agents can now perform web searches and scrape content from URLs, enabling research-driven tasks like gathering documentation or examples from the internet. These tools return summarized results or markdown-formatted content, making it easier to incorporate external knowledge into your projects. For larger projects, the new indexed file context tool provides quick summaries of key files, helping agents (and you) understand codebases more efficiently without manual navigation.

These changes empower users to handle more dynamic tasks securely, reducing the need to switch between tools and improving productivity in coding, debugging, and exploration.

A Closer Look at the Changes

New Functionality

• Terminal Access Tool: A new "terminal" tool integrates bash shell execution into the agent toolkit. Commands are run via Python's subprocess module with a 30-second timeout, and results are returned as strings. User confirmation is mandatory for each command, ensuring secure usage. This supports tasks requiring file system interaction, with guidelines for verifying directories and quoting paths.
• Web Tools: Added "web_search" and "web_scrape_url" tools. Web search uses DuckDuckGo to return URLs and summaries for queries, while web scraping fetches and converts URL content to markdown. These enable agents to access online resources for informed decision-making.
• Indexed Files Context: The "get_indexed_files_context" tool retrieves compact summaries of project files, optionally filtered by paths. This aids agents in understanding large codebases efficiently.
• UI Enhancements: Introduced a command confirmation widget in the Textual UI for terminal prompts, allowing users to approve or cancel commands interactively. This integrates seamlessly into the terminal output area.

Code Refinements & Improvements

• Router Agent Reliability: Added a salvage mechanism to recover from malformed JSON responses, improving parsing robustness. Updated prompts emphasize clarifying user intent, especially after cancellations, and refined decision-making logic to prefer tools for questions over launching full coding agents unnecessarily.
• Plan Phase Handling: Improved retry logic for empty responses when generating step plans, ensuring more reliable agent pipelines.
• Message Builder: Added support for including project summaries in messages, enhancing context for agents.

Bug Fixes

• Addressed issues in router agent response parsing by implementing a fallback salvage prompt, reducing errors from invalid JSON.
• Enhanced UI scrollbar consistency across widgets for better readability.

Dependency Updates

• Added ddgs for web search, markdownify for HTML-to-markdown conversion, and httpx for asynchronous HTTP requests.
• Expanded Python version compatibility to include 3.11 and 3.12 in setup.py.

These updates collectively make JrDev a more versatile tool for AI-assisted development, with a strong focus on user-controlled automation. For installation, use pip install jrdev==0.1.8-alpha. Report any issues on the GitHub repository.
Generated by JrDev AI using x-ai/grok-4

What's Changed

• Expand agent capabilities with shell access, project context tools, and safer command execution by @presstab in #114
• Enhance router agent resilience and inject project overview context by @presstab in #116
• Add Web Search (DuckDuckGo) and Scraping Capabilities by @presstab in #117
• Bump v0.1.8 by @presstab in #118

Full Changelog: v0.1.7-alpha...v0.1.8-alpha

Version 0.1.7-alpha

v0.1.7-alpha Release

Overview

This release enhances JrDev’s AI-driven development workflows by adding per-context model selection, commit history browsing, and misc bug fixes. Core entrypoints have been refactored for consistency, error handling in the command router has been improved, documentation has been overhauled, and Git integration now includes commit diffs alongside staged/unstaged files.

What This Means for Users

• On-the-fly Model Switching
  Chat threads, terminal commands, PR summaries/reviews, and commit message generation each feature a dropdown to select the active AI model without restarting the session.
• Commit History & Diff Viewer
  The Git Tools panel lists recent commits; selecting one displays its full diff. Commit message generation can target a dedicated AI profile.
• Streamlined CLI Entry-Points
  python -m jrdev now launches the Textual UI by default; the jrdev-cli script has been refactored into its own module for clearer separation.
• Expanded Documentation
  The README has been condensed to key features and installation steps, and a new docs/code.md details the /code command’s six-phase pipeline: Analyze → Fetch Context → Plan → Execute → Review → Validate.

A Closer Look at the Changes

Core & Architecture

• Entry-Point Refactor
  • Moved CLI logic into src/jrdev/cli.py; updated setup.py accordingly.
  • Updated the module’s __main__.py so that python -m jrdev invokes the Textual UI.
• Model Discovery
  • Added Application.get_available_models() to expose only models for which API keys are configured.

Agents & Error Handling

• Router Agent
  • Enhanced JSON parsing error logs to include raw content for easier troubleshooting.

Documentation

• README Overhaul
  • Simplified project description, consolidated feature list, and updated installation instructions.
• New Code Agent Guide (docs/code.md)
  • Step-by-step explanation of the structured /code pipeline, including prompts, model profiles, and user-driven confirmations.

UI & TUI Enhancements

• ModelListView Widget
  • Introduced a generic dropdown component that dynamically sizes itself and positions relative to a trigger button.
• Chat View
  • Added a model-selection button with a popup per thread.
  • Integrated update_models() calls so that model changes reflect immediately in the UI.
• Terminal Output Widget
  • Added model-selection support for intent routing; the model button and popup appear alongside the copy button.
• Worker Lifecycle Management
  • The TUI now tracks launched workers to enable finer-grained cancellation and cleanup.

Git Tools Improvements

• Commit History & Diff
  • Added get_commit_history() and get_commit_diff() in git_utils.py.
  • The Git Tools screen now shows the most recent commits in a list, with diffs rendered upon selection.
• Commit Message Model Profile
  • A dedicated model dropdown lets users choose which AI profile generates the commit message.

CLI & Editor Fallback

• Curses Editor Simplification
  • Restructured imports and replaced the raw‐mode implementation with a portable, line-by-line input fallback for editing multi-line JSON.

Styling & Layout

• TCSS Adjustments
  • Introduced layering rules for toast notifications and model popups.
  • Refined borders, padding, and height constraints across various widgets for a more cohesive appearance.

Generated by JrDev AI using o4-mini-2025-04-16

What's Changed

• docs: add code agent overview documentation and link in README by @presstab in #107
• feat: add git commit history and diff functionality by @presstab in #108
• refactor: lazy import curses and simplify fallback CLI editor by @presstab in #109
• Fix DirectoryWidget freeze when cancelling workers by @presstab in #110
• feat: add model selection dropdown to git tools and commit views by @presstab in #111
• Integrate Model Selection into UI Components by @presstab in #112
• refactor: bump version, move CLI entry point to cli.py and update con… by @presstab in #113

Full Changelog: v0.1.6-alpha...v0.1.7-alpha

Version 0.1.6-alpha

Overview

This version bump introduces improvements to the application's initialization process by ensuring a default chat model is set if one isn't already configured. Additionally, it enhances the robustness of the Router Agent by restricting its ability to execute certain sensitive commands, preventing potential misuse or unintended operations. These changes contribute to a more stable and secure application environment.

A Closer Look at the Changes

Code Refinements & Improvements

• Default Chat Model Initialization: The Application class now includes logic to automatically set a default chat model during initialization if no model is currently configured in the application state. This ensures that the application always has a functional chat model available, improving startup reliability.
  • The system attempts to retrieve an "intermediate_reasoning" model from the ModelProfileManager.
  • If successful, this model is set as the active chat model. Error handling is included to gracefully manage scenarios where the default model cannot be set.
• Router Agent Command Restrictions: The Application class has been updated to prevent the Router Agent from executing a predefined list of sensitive commands. This adds a layer of security and control over the agent's actions.
  • A new list restricted_commands has been introduced, currently including /init, /migrate, and /keys.
  • Before executing a command suggested by the Router Agent, the system now checks if the command is in this restricted list.
  • If a restricted command is detected, an error message is displayed, and the command execution is prevented.
• Version Update: The application version has been incremented from 0.1.5-alpha to 0.1.6-alpha.

New Functionality

• Router:Ignore Annotation: A new Router:Ignore annotation has been added to the handle_init function in src/jrdev/commands/init.py. This annotation is likely intended to signal to the Router Agent or other automated processes that this specific command should be ignored or handled specially, aligning with the new command restriction logic.

Bug Fixes

• Initial Model State: The AppState class's initialization of self.model has been modified. Instead of defaulting to "deepseek-r1-671b" if no model is loaded, it now defaults to an empty string. This change supports the new default model setting logic in the Application class, allowing the application to explicitly set a preferred default rather than relying on a hardcoded fallback in the state.
  Generated by JrDev AI using google/gemini-2.5-flash

What's Changed

• Enhance Application First Run and Router Agent Restrictions. Bump v0.1.6-alpha by @presstab in #106

Full Changelog: v0.1.5-alpha...v0.1.6-alpha

Version 0.1.5-alpha

Overview

This release focuses on quality-of-life improvements and bug fixes across several components of the application. Key changes include enhanced API key management, improved model profile validation, better error handling for Git operations, and updates to the model list configuration. These changes collectively improve the stability and user experience of the application.

What This Means for Users

• Improved API Key Management: The key removal process now provides clearer validation messages and better handles edge cases when removing API keys.
• Model Profile Warnings: Users will now receive warnings when model profiles are configured to use models from providers without API keys, preventing potential runtime errors.
• Git Tools Enhancement: The Git tools screen now detects when Git is not installed and provides clear installation instructions.
• Model List Updates: Several models have been removed from the configuration, and new free-tier models have been added to Open Router's offerings.

A Closer Look at the Changes

API Key Management Improvements

• Refactored provider handling to use ApiProvider objects instead of dictionaries for better type safety
• Enhanced validation logic for key removal to prevent removing the last remaining key
• Improved error messages when attempting to remove keys from unknown services

Model Profile System

• Added new method get_profiles_with_missing_keys() to identify profiles using models without API keys
• Updated default model profiles to use more stable model selections
• Added warning system that checks profiles before executing key commands (/code, /init, /projectcontext)

Git Tools

• Added Git installation detection with helpful instructions when Git is not found
• Disabled Git-related buttons when Git is not installed to prevent errors
• Improved error handling for file list parsing operations

Model List Updates

• Removed several Venice provider models that are no longer available
• Added new free-tier models to Open Router:
  • deepseek/deepseek-chat-v3-0324:free
  • deepseek/deepseek-r1-0528:free
• Added new premium models to Open Router:
  • x-ai/grok-4
  • mistralai/devstral-small-2507
  • mistralai/devstral-medium-2507

Bug Fixes

• Fixed regex pattern in file list parsing to handle edge cases better
• Improved error logging for file operations
• Removed unnecessary provider-specific parameters for Open Router requests
• Changed default provider requirement from True to False when adding new providers
  Generated by JrDev AI using deepseek/deepseek-chat-v3-0324:free

v0.1.4-alpha

Release Version 0.1.4-alpha

Overview

This release introduces significant enhancements to the user interface and the intelligence of the agent system. Key updates include a fully customizable terminal color scheme, allowing users to tailor the look and feel of the application, and a more sophisticated Router Agent with improved reasoning and command execution transparency. Additionally, numerous under-the-hood improvements have been made to increase agent reliability, refine prompt engineering, and enhance error handling.

✨ New Features & Enhancements

Customizable Terminal UI

The terminal interface has been upgraded to support fully customizable text styles. This allows users to personalize the application's appearance for better readability and a more pleasant user experience.

• New Settings Panel: A "Terminal Styles" tab has been added to the settings screen (/settings). From here, users can define the color and style (e.g., bold, italic) for various types of output, including errors, successes, processing messages, and agent responses.
• Color-Coded Output: By default, terminal output is now color-coded to make it easier to distinguish between different message types at a glance.
• Improved Auto-Scrolling: The terminal's auto-scrolling behavior has been refined to be more intuitive, ensuring the view follows new output correctly but stops when the user manually scrolls up.

Smarter & More Transparent Router Agent

The Router Agent, which interprets user requests, has been substantially improved to be more effective and transparent.

• Command Reasoning: The agent now provides a brief "reasoning" statement before executing a command, explaining its purpose. This gives users clear insight into the agent's decision-making process.
• Clear Command History: A new /routeragent clear command has been added to reset the agent's conversational memory. This is useful for starting a new line of inquiry without prior context influencing the agent.
• Improved Prompting: The agent's core instructions have been updated to better understand user intent, prefer reading files for context over asking clarifying questions, and pass user requests to the /code command verbatim for more predictable results.

🛠️ A Closer Look at the Changes

Code Refinements & Improvements

• UI Styling System: The terminal UI was refactored to use a new TerminalTextStyles system. Instead of relying on generic markdown, output is now tagged with a PrintType (e.g., ERROR, PROCESSING, SUCCESS). This allows for precise, centralized, and user-configurable styling for all terminal text.
• Agent Pipeline Readability: All phases of the code generation pipeline (Analyze, Plan, Execute, etc.) now use the new PrintType system to tag their status updates. This makes the step-by-step process much easier to follow in the terminal.
• Robust File Requests: A new "salvage" mechanism has been implemented. If the LLM generates a malformed file request, the system now attempts to parse and correct it, reducing failures in the context-gathering phase.
• Error Handling: The message streaming service now includes more robust error handling. If an issue occurs while streaming a chat response, the error is caught and reported in the UI, preventing silent failures.

New Functionality

• TerminalTextStyles Class: A new class responsible for loading, managing, and saving user-defined terminal color and text styles.
• TerminalStylesWidget: The new UI component within the settings screen that allows users to edit and save their style preferences.
• Router Agent Context Reset: The application state now supports clearing the router's message history, enabling the /routeragent clear command.
• reasoning in ToolCall: The ToolCall data structure was extended to include a reasoning field, capturing the agent's justification for a chosen action.
  Generated by JrDev AI using google/gemini-2.5-pro

v0.1.3-alpha

JrDev v0.1.3-alpha: Natural Language Commands & Workflow Enhancements

This release introduces a major step forward in user interaction with the new Router Agent, allowing you to issue commands in natural language. The /code command is now powered by a more robust and reliable agent pipeline, and the application's documentation and installation process have been significantly improved.

Natural Language Command Router

The most significant new feature is the introduction of a Router Agent that interprets natural language. Instead of needing to know the exact command syntax, you can now type your requests directly into the terminal.

• Conversational Commands: Simply type what you want to do, such as "add error handling to the main function in main.py". The agent will analyze your request, determine the necessary steps, and execute the appropriate tools and commands.
• Agentic Workflow: The router can use tools to gather information (e.g., reading files or listing the directory tree) before executing a final command like /code, making it more autonomous and effective.
• Configuration: A new /routeragent command has been added to manage the agent's settings, such as setting the maximum number of iterations it can perform for a single request (/routeragent set-max-iter 5).

Simplified Installation and Documentation

Getting started with JrDev is now easier than ever, and the documentation has been completely overhauled for clarity.

• PyPI Installation: JrDev is now available on the Python Package Index (PyPI). You can install it with a simple pip install jrdev.
• Revamped README: The project's README.md has been rewritten to provide clearer instructions, simplified workflow explanations, and better overall guidance for both new and existing users.

Enhanced /code Agent and Task Execution

The agent responsible for handling the /code command has been significantly refactored for better reliability and transparency.

• Robust Pipeline: The code generation process is now broken into distinct, more robust phases: Analyze, Plan, Execute, Review, and Validate. This structured approach improves reliability and provides better feedback during execution.
• Real-Time Cost Tracking: The Task Monitor in the TUI now includes a cost column, displaying a real-time cost estimate for each running task. This gives you immediate visibility into token usage and expenses.

Expanded Provider and Model Management

Managing API providers and models is now more flexible.

• Custom API Providers: The new /provider command allows you to add and manage custom OpenAI-compatible API providers. This is ideal for users who want to connect to local models via services like Ollama or use other custom endpoints.
• Updated Model List: The built-in model list has been updated with newer models and more accurate pricing information.

Additional Improvements

• Command Help: The help text and docstrings for most commands have been improved to provide more detailed descriptions and clear usage examples.
• Git Integration: The underlying prompts for generating PR summaries and reviews (/git pr summary and /git pr review) have been rewritten to produce higher-quality, more structured, and professional output.
• Code Quality: The project now uses black and pylint for consistent code formatting and quality, configured via a new pyproject.toml file.
• Cost Reporting: The /cost command has been refined to provide a clearer and more accurate breakdown of session expenses.
  Generated by JrDev AI using google/gemini-2.5-pro

What's Changed

• docs: update installation and usage instructions by @presstab in #84
• Lint and clean agents/ and commands/ by @presstab in #86
• fix: invoke profile_manager() when retrieving advanced reasoning model by @presstab in #87
• Feat: AI-Powered Intent Router for a More Natural UX by @presstab in #88
• Model Profile Updates & New Routing Capabilities by @presstab in #89
• Enhanced PR Prompt Structure for Clarity and User Focus by @presstab in #90
• fix: accurately calculate tokens per second in streaming responses by @presstab in #92
• System Process Refinements for Command Pipeline Optimization by @presstab in #93
• Add cost streaming to task monitor widget by @presstab in #94
• bump version 0.1.3-alpha by @presstab in #95

Full Changelog: v0.1.2-alpha...v0.1.3-alpha

v0.1.2-alpha

What's Changed

• Add Interactive Git Overview Panel with File Staging, Unstaging, and Diff Viewing by @presstab in #80
• Refactor: Decompose CodeProcessor into a Multi-Stage Agent Pipeline by @presstab in #81
• Enhance Git Commit Workflow with AI Assistance by @presstab in #82
• Polish Packaging & Asset Inclusion. Bump to 0.1.2-alpha by @presstab in #83

Full Changelog: v0.1.1-alpha...v0.1.2-alpha