- Added interactive /review command with mode selection for base
branch, uncommitted changes, specific commit, or custom
instructions.
- Added subprocess tool registry for extracting and rendering tool
data from subprocess agents.
- Added combined review result rendering with verdict and findings
tree structure.
- Changed bundled commands to be overridable by user/project
commands with same name.
- Removed findings_count parameter from submit_review tool in
favor of automatic counting.
- Added fix npm script for auto-fixable lint issues with unsafe
fixes.
The --model CLI flag now accepts provider/model format directly,
eliminating the need for a separate --provider flag. The Task tool's
model-resolver returns provider/model format, which the subprocess
CLI correctly parses.
Previously, subprocesses were passed --model gpt-5.2 without --provider,
causing the CLI to ignore the model and fall back to settings default.
Files referenced with @path/to/file.ext syntax are automatically read
and injected as synthetic Read tool calls before the agent responds,
eliminating the need for manual file reads.
- Write input file (.in.md) before spawning subprocess
- Use subtask-specific session file so pi writes jsonl in real-time
- Output file (.out.md) still written at completion (summary)
Enables observing/debugging running subagents via their session files.
message_update events contain full accumulated text, not deltas.
Previously appending them caused repeated content in artifacts and
progress display. Now message_update only updates progress display
(replacing, not accumulating) and message_end handles final output.
- Added hidden property for custom tools to exclude them from default tool list unless explicitly requested.
- Added explicitTools option to createAgentSession for enabling hidden tools by name.
- Added example review tools (report_finding, submit_review) with structured findings accumulation and verdict rendering.
- Added /review command for interactive code review with branch comparison, uncommitted changes, and commit review modes.
- Updated bundled reviewer agent to use structured review tools with priority-based findings (P0-P3) and formal verdict submission.
commands.enableClaudeUser and commands.enableClaudeProject settings
were not being used - fixed by passing settingsManager.getCommandsSettings()
to discoverSlashCommands in createAgentSession.
- Added custom TypeScript slash commands loaded from ~/.pi/commands/ and .pi/commands/ directories.
- Added support for loading markdown slash commands from Claude's ~/.claude/commands/ and .claude/commands/ directories.
- Added commands.enableClaudeUser and commands.enableClaudeProject settings to control Claude command directory loading.
- Implemented CustomCommandAPI with exec, typebox, and pi module injection for command execution context.
- Integrated custom commands into autocomplete provider for interactive mode discovery.
- Added interruptMode option to control when queued messages interrupt tool execution.
- Implemented "immediate" mode (default) to check queue after each tool and interrupt remaining tools.
- Implemented "wait" mode to defer queue processing until the entire turn completes.
- Exposed interruptMode configuration in interactive settings UI with immediate/wait options.
- Wired interrupt mode through AgentSession and SDK to enable runtime configuration and persistence.
- Renamed SMALL_MODEL_PRIORITY to SMOL_MODEL_PRIORITY
- Renamed findSmallModel to findSmolModel
- Updated all role strings from 'small' to 'smol'
- Updated UI labels and comments
- Added --smol and --slow CLI args for ephemeral model role overrides
- Added PI_SMOL_MODEL and PI_SLOW_MODEL env var support
- Overrides are applied at runtime without persisting to settings.json
- Added pi/smol alias for explore agent (fast/cheap model)
- Added pi/slow alias for reviewer agent (reasoning/codex model)
- Added L key binding in model selector to set slow model
- Added 🧠 marker for slow model in selector
- Slow model priority chain: gpt-5.2-codex > codex > gpt > opus > pro
- Settings now use modelRoles map (e.g., { default: 'anthropic/claude-sonnet-4-5', small: 'anthropic/claude-haiku-4-5' })
- Session persistence uses 'provider/modelId' string format with optional role field
- Model selector: Enter sets default, S sets small, Escape closes
- Added role markers in selector: checkmark for default, lightning bolt for small
- Task tool supports pi/<role> aliases (e.g., model: pi/small, haiku, flash, mini)
- Title generation uses configured small model with auto-discovery fallback
- Explore agent now uses pi/small as primary model preference
- Added automatic session title generation using small models (Haiku/GPT-4o-mini/Flash) to improve session identification.
- Implemented terminal window title updates to display current session context.
- Updated session selector to show title with dimmed first message preview for better session browsing.
- Extended SessionInfo and SessionHeader interfaces to persist and display generated titles.
- Added title model preference hierarchy with fallback to any available model for reliability.
- Added LSP format on write to automatically format code files using language server formatters after writing.
- Implemented LSP server warmup at startup to avoid cold-start delays when first writing files.
- Added LSP server status display in welcome banner to show which language servers are active and ready.
- Fixed recent sessions display in welcome banner which was never wired up.
- Improved grep tool description to document pagination options and clarify recursive search behavior.
- Added edit fuzzy match setting to allow users to control whitespace/indentation tolerance in the edit tool.
- Added comprehensive LSP server configurations for 40+ languages including Rust, Go, Python, Java, Kotlin, Scala, Haskell, OCaml, Elixir, Ruby, PHP, C#, Lua, Nix, and more.
- Extended LSP config file search paths to include lsp.json, .lsp.json in project root, .pi/ subdirectory, and user-level configs in ~/.pi/ and home directory.
- Increased LSP diagnostics timeout from 2000ms to 5000ms to accommodate slower language servers.
- Added LSP diagnostics integration to write tool for immediate feedback on syntax errors and type issues after writing code files.
- Implemented multi-server LSP diagnostics to query all applicable language servers (e.g., TypeScript + Biome) and merge results for comprehensive error reporting.
- Added Biome language server configuration to enable combined type-checking and linting diagnostics for TypeScript/JavaScript projects.
- Introduced lsp.diagnosticsOnWrite setting to allow users to control automatic diagnostics behavior.
- Enhanced TUI tool execution component to display LSP diagnostics with color-coded severity indicators.
- Replaced vendored highlight.js with CDN-hosted version to reduce bundle size by ~1200 lines.
- Added CSS, JS, and HTML minification functions to reduce exported file size.
- Implemented caching for minified template and JS assets to avoid redundant processing.
- Updated share preview URL to use gistpreview.github.io for session sharing.
- Introduced declarative settings definitions to eliminate duplicated setting logic across UI and handlers.
- Extracted settings configuration into settings-defs.ts for single-source-of-truth maintenance.
- Simplified SettingsSelectorComponent to generate UI from definitions rather than hardcoded items.
- Added bash interceptor toggle to allow users to control shell command interception behavior.
- Updated CLI help text to reflect all available tools with accurate default states.
- Consolidated settings change handling into unified handler dispatching by setting ID.
- Replaced hardcoded shell path detection with user's $SHELL preference for better environment compatibility.
- Simplified commandExists function to use Bun.which instead of spawning processes for faster tool detection.
- Updated getToolPath to return full binary path from Bun.which for more reliable tool resolution.
- Updated CI workflows to include explicit coding-agent build step for proper binary compilation.
- Changed TypeBox validation from ts-expect-error to explicit any cast to resolve CJS/ESM type mismatch.
- Replaced build-binaries workflow with unified publish workflow to consolidate release automation.
- Added matrix-based parallel binary builds across platforms for faster release cycles.
- Simplified CI workflow by removing concurrency settings and adding CLI smoke tests.
- Added npm publish step with provenance support for secure package distribution.
- Introduced skip_npm input to allow binary-only releases when needed.
- Replaced process group kill with descendant-aware tree traversal to handle non-leader processes.
- Added pgrep-based child discovery with /proc fallback for broader Linux compatibility.
- Introduced optional grace period parameter to allow SIGTERM before SIGKILL for graceful shutdown.
- Simplified focus return value in interactive mode for consistency.
- Added TabBar component to enable horizontal tab navigation with keyboard support.
- Refactored settings selector to use tabbed interface for better organization of Config, Exa, and Plugins sections.
- Added Shift+Space handling via Kitty protocol to fix space insertion in editor.
- Removed redundant LSP explanation text from system prompt for conciseness.
- Added specialized Exa web search tools for deep search, code context, URL crawling, LinkedIn, and company research.
- Implemented settings UI for granular control over Exa tool availability with per-feature toggles.
- Changed Exa addon defaults to disabled to prevent unexpected API usage for LinkedIn, company, researcher, and websets features.
- Added settings manager methods to persist individual Exa feature preferences.
- Removed ast and replace tools to simplify toolset and reduce maintenance overhead.
- Added LSP call hierarchy support (incoming_calls, outgoing_calls) for better code navigation.
- Added Exa as web search provider alongside Anthropic and Perplexity for broader search options.
- Enhanced tool descriptions with detailed usage guidelines to improve agent tool selection.
- Simplified task tool by removing single-mode execution and agentScope parameter.
- Added GitHub tree rendering and improved web-fetch documentation for better URL handling.
- Added fuzzy matching with Levenshtein distance to handle whitespace and indentation differences in edit operations.
- Implemented configurable similarity threshold (default 95%) to balance precision with flexibility.
- Added detailed error messages showing closest match location and similarity percentage to aid debugging.
- Updated edit tool schema to expose fuzzy parameter for user control over matching behavior.
- Renamed npm package scope from @mariozechner to @oh-my-pi across all packages for consistent branding.
- Removed packages/pods directory and all related pod management functionality.
- Updated import paths and module declarations throughout codebase to reflect new package names.
- Reformatted code with consistent trailing comma removal and ternary operator alignment.
- Added MCP (Model Context Protocol) support with stdio and HTTP transports for external tool integration.
- Implemented plugin system with install, enable, configure, and doctor commands for extensibility.
- Added 22 Exa MCP tools for web search, LinkedIn, company research, and websets operations.
- Introduced LSP tool with rust-analyzer support for code intelligence (diagnostics, hover, references, code actions).
- Added task tool for delegating work to specialized subagents with parallel execution support.
- Implemented web_search and web_fetch tools with Anthropic and Perplexity provider support.
- Added git context loading to system prompt to provide AI with repository state awareness.
- Implemented automatic detection of main/master branch to support varied repository conventions.
- Included current branch, git status, and recent commits snapshot for contextual understanding.
- Added emergency terminal restore function to reset terminal state on crashes without requiring instance access.
- Implemented crash handlers for SIGTERM, SIGHUP, uncaught exceptions, and unhandled rejections to ensure clean terminal exit.
- Added active terminal tracking to enable blind restore when no instance is available.
- Implemented a new welcome screen component for startup.
- Introduced a powerline-style footer with integrated git status, model info, and token usage.
- Consolidated extension loading messages into a new `/status` command.
- Added `?` hotkey to display shortcuts when the editor is empty.
Merged 53 commits from pi/main including:
- Export HTML rewrite with tree sidebar and /share command
- Hooks API enhancements (custom status, editor, session management)
- Edit diff preview before tool execution
- TUI fixes (Unicode format chars, OSC 8 hyperlinks, emoji optimization)
- Gemini CLI rate limit retry with server-provided delay
- Various bug fixes and documentation updates
Resolved conflicts by preserving Bun migration (spawn→Bun.spawn,
fs→Bun.file) while accepting upstream's new features and APIs.
- Migrated from Node.js to Bun runtime to leverage native TypeScript execution and faster startup times.
- Replaced npm with bun package manager and updated lockfile format.
- Converted Node.js APIs (child_process, fs, readline) to Bun equivalents across all packages.
- Updated shebang lines from node/tsx to bun for CLI entry points.
- Removed TypeScript build step by using source files directly as package entry points.
- Updated CI workflows to use Bun instead of Node.js for installation and testing.
BorderedLoader receives theme as a parameter but was creating
DynamicBorder() without passing it, causing DynamicBorder to
fall back to the global theme variable which is undefined when
loaded via jiti (separate module cache).
Added doc comment to DynamicBorder explaining the jiti issue.
Move line count from header to footer to avoid changing the first line
during streaming, which was triggering full screen re-renders in the
TUI's differential rendering logic.
can1357