openclacky 0.9.0 → 0.9.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 732978cd6072f33b558e6260af317ffa188529f922913193d9b7505745f0bb2b
-  data.tar.gz: 3a03afdbb194bbf3e52cd89d173a05186e01ce4cb8564410e294e0e5bb7db501
+  metadata.gz: 634937d9d7a20aa0a76b046ba079ef2d82c398ac81d59014968f7dbeb325726c
+  data.tar.gz: 7cd5c7eab980c54b5da38b6744dbc21dd3ac48af00c541a609d5537d6867ee70
 SHA512:
-  metadata.gz: 0ad076fe8aa80fe40b8f9908cdf6bab87c08110c8940c76aa3d8fe5944886f565735b7dd06008e9f54fc0a13a3b1d5966800d3bb02bf1d7268cf99faab181dd4
-  data.tar.gz: e29e2688839d08a271b105830df847517cfaab57ecae36d50787687180b1ba9a41e15a68f6e3afc21a800dd78b6eea6c7556c97c503aabdc46c2763bdf1c0bb8
+  metadata.gz: b874781caf58c502536666c33b8aff0c459076689909dd36bd82151f7092953d453d51b2e90259bbfdb6b2eb0ba389e5e56c083a1364118ca0a15edbef471e02
+  data.tar.gz: 1639031ccf11848ab3b8c2a18d3eb7d87c487dc1e13b856e4c0cdc8cb35e3ebf4c8001660ea607f013e64a40fa1ef09255915652af9d0bc44ad8ddf9c0b1dff2

data/CHANGELOG.md

@@ -7,6 +7,21 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.9.2] - 2026-03-15
+
+### Fixed
+- **Version upgrade button now appears reliably**: the new version check now queries RubyGems directly instead of relying on local gem mirror sources (which often lag behind by hours or days), so the upgrade badge shows up promptly when a new version is available. Falls back to the local mirror if RubyGems is unreachable.
+- **Edit confirmation diff output restored**: the file diff was not displaying correctly when the input area paused during an edit confirmation prompt; this is now fixed.
+
+## [0.9.1] - 2026-03-15
+
+### Added
+- **Session context auto-injection**: the agent now automatically injects the current date and active model name into each conversation turn, so it always knows what day it is and which model it's running — helpful for time-sensitive tasks and multi-model setups
+- **Kimi/Moonshot extended thinking support**: reasoning content is now preserved and echoed back correctly in message history, fixing HTTP 400 errors when using Kimi's extended thinking API
+
+### Improved
+- **Browser tool install UX**: the `agent-browser` setup flow has been redesigned with a dedicated install script and clearer guidance, making first-time setup smoother
+
 ## [0.9.0] - 2026-03-14
 
 ### Added

data/lib/clacky/agent.rb

@@ -182,6 +182,9 @@ module Clacky
         @messages << system_message
       end
 
+      # Inject session context (date + model) if not yet present or date has changed
+      inject_session_context_if_needed
+
       # Format user message with images and files if provided
       user_content = format_user_content(user_input, images, files)
       @messages << { role: "user", content: user_content, task_id: task_id, created_at: Time.now.to_f }
@@ -423,6 +426,9 @@ module Clacky
       end
       # Store token_usage in the message so replay_history can re-emit it
       msg[:token_usage] = response[:token_usage] if response[:token_usage]
+      # Preserve reasoning_content so it is echoed back to APIs that require it
+      # (e.g. Kimi/Moonshot extended thinking — omitting it causes HTTP 400)
+      msg[:reasoning_content] = response[:reasoning_content] if response[:reasoning_content]
       @messages << msg
 
       response
@@ -871,6 +877,33 @@ module Clacky
       content
     end
 
+    # Inject a session context message (date + model) into the conversation.
+    # Only injects when:
+    #   1. No context message exists yet in this session, OR
+    #   2. The existing context is from a previous day (cross-day session)
+    # Marked with system_injected: true so existing filters (replay_history,
+    # get_recent_user_messages, etc.) automatically skip it.
+    # Cache-safe: always inserted just before the current user message,
+    # so no historical cache entries are ever invalidated.
+    private def inject_session_context_if_needed
+      today = Time.now.strftime("%Y-%m-%d")
+
+      # Find the last injected context message
+      last_ctx = @messages.reverse.find { |m| m[:session_context] }
+
+      # Skip if we already have a context for today
+      return if last_ctx && last_ctx[:context_date] == today
+
+      content = "[Session context: Today is #{Time.now.strftime('%Y-%m-%d, %A')}. Current model: #{current_model}]"
+      @messages << {
+        role: "user",
+        content: content,
+        system_injected: true,
+        session_context: true,
+        context_date: today
+      }
+    end
+
     # Track modified files for Time Machine snapshots
     # @param tool_name [String] Name of the tool that was executed
     # @param args [Hash] Arguments passed to the tool

data/lib/clacky/client.rb

@@ -659,13 +659,20 @@ module Clacky
           end
         end
 
-        {
+        result = {
           content: message["content"],
           tool_calls: parse_tool_calls(message["tool_calls"]),
           finish_reason: data["choices"].first["finish_reason"],
           usage: usage_data,
           raw_api_usage: raw_api_usage
         }
+
+        # Preserve reasoning_content if present (e.g. Kimi/Moonshot extended thinking).
+        # The API requires this field to be echoed back in the message history on
+        # subsequent requests, otherwise it returns HTTP 400.
+        result[:reasoning_content] = message["reasoning_content"] if message["reasoning_content"]
+
+        result
       else
         raise_error(response)
       end

data/lib/clacky/server/http_server.rb

@@ -687,10 +687,38 @@ module Clacky
         latest
       end
 
-      # Query the latest openclacky version via `gem list -r openclacky`.
-      # Runs through login shell so gem source mirrors configured via rbenv/mise work correctly.
-      # Output format: "openclacky (0.9.0)"
+      # Query the latest openclacky version.
+      # Strategy: try RubyGems official REST API first (most accurate, not affected by mirror lag),
+      # then fall back to `gem list -r` (respects user's configured gem source).
       private def fetch_latest_version_from_gem
+        fetch_latest_version_from_rubygems_api || fetch_latest_version_from_gem_command
+      end
+
+      # Try RubyGems official REST API — fast and always up-to-date.
+      # Returns nil if the request fails or times out.
+      private def fetch_latest_version_from_rubygems_api
+        require "net/http"
+        require "json"
+
+        uri      = URI("https://rubygems.org/api/v1/gems/openclacky.json")
+        http     = Net::HTTP.new(uri.host, uri.port)
+        http.use_ssl     = true
+        http.open_timeout = 5
+        http.read_timeout = 8
+
+        res = http.get(uri.request_uri)
+        return nil unless res.is_a?(Net::HTTPSuccess)
+
+        data = JSON.parse(res.body)
+        data["version"].to_s.strip.then { |v| v.empty? ? nil : v }
+      rescue StandardError
+        nil
+      end
+
+      # Fall back to `gem list -r openclacky` via login shell.
+      # Respects the user's configured gem source (rbenv/mise mirrors, etc.).
+      # Output format: "openclacky (0.9.0)"
+      private def fetch_latest_version_from_gem_command
         shell  = Clacky::Tools::Shell.new
         result = shell.execute(command: "gem list -r openclacky", soft_timeout: 15, hard_timeout: 30)
         return nil unless result[:exit_code] == 0

data/lib/clacky/tools/browser.rb

@@ -13,14 +13,24 @@ module Clacky
       self.tool_description = <<~DESC
         Browser automation for login-related operations (sign-in, OAuth, form submission requiring session). For simple page fetch or search, prefer web_fetch or web_search instead.
 
-        isolated param: true = built-in browser (default, works immediately, login persists). false = user's Chrome (keeps cookies/login, needs one-time debug setup; opens URLs in new tab).
+        isolated: true = built-in browser (default, no setup, login persists). false = user's Chrome (keeps cookies/login, needs one-time debug setup; opens URLs in new tab).
 
-        WORKFLOW:
-        - Default: use isolated=true (built-in browser). No setup, works immediately.
-        - If the user explicitly wants their Chrome (e.g. "use my Chrome"), set isolated=false. We auto-open chrome://inspect/#remote-debugging when needed — guide user to enable the toggle. When opening URLs, we use a new tab to avoid replacing the user's current page.
+        SNAPSHOT — always run before interacting with a page. Refs (@e1, @e2...) expire after page changes, always re-snapshot before acting on a changed page:
+        - 'snapshot -i -C' — interactive + cursor-clickable elements (recommended default)
+        - 'snapshot -i' — interactive elements only (faster, for simple forms)
+        - 'snapshot' — full accessibility tree (when above miss elements)
 
-        Commands: 'open <url>', 'snapshot -i', 'click @e1', 'fill @e2 "text"', etc.
-        IMPORTANT: Always use 'snapshot -i' to inspect page state. NEVER use 'screenshot' without explicit user permission — it costs significantly more tokens. If snapshot is insufficient, tell the user and ask if they want a screenshot.
+        ELEMENT SELECTION — prefer in this order:
+        1. Refs: 'click @e1', 'fill @e2 "text"'
+        2. Semantic find: 'find text "Submit" click', 'find role button "Login" click', 'find label "Email" fill "user@example.com"'
+        3. CSS: 'click "#submit-btn"'
+
+        OTHER COMMANDS:
+        - 'open <url>', 'back', 'reload', 'press Enter', 'key Control+a'
+        - 'scroll down/up', 'scrollintoview @e1', 'wait @e1', 'wait --text "..."', 'wait --load networkidle'
+        - 'dialog accept/dismiss', 'tab new <url>', 'tab <n>'
+
+        SCREENSHOT: NEVER call on your own — costs far more tokens than snapshot. Last resort only. Ask user first: "Screenshots cost more tokens. Approve?" When approved: 'screenshot --screenshot-format jpeg --screenshot-quality 50'.
       DESC
       self.tool_category = "web"
       self.tool_parameters = {
@@ -47,11 +57,29 @@ module Clacky
       CHROME_DEBUG_PORT = 9222
       BROWSER_COMMAND_TIMEOUT = 30
       CHROME_DEBUG_PAGE = "chrome://inspect/#remote-debugging"
+      MIN_AGENT_BROWSER_VERSION = "0.20.0"
 
       def execute(command:, session: nil, isolated: nil, working_dir: nil)
-        unless agent_browser_installed?
-          install_result = auto_install_agent_browser
-          return install_result if install_result[:error]
+        # Handle explicit install command
+        if command.strip == "install"
+          return do_install_agent_browser
+        end
+
+        if !agent_browser_installed?
+          return {
+            error: "agent-browser not installed",
+            message: "agent-browser is required for browser automation but is not installed.",
+            instructions: "Tell the user: 'agent-browser is not installed. It's required for browser automation. Run `browser(command: \"install\")` to install it — this may take a minute. Would you like me to install it now?' Wait for user confirmation before calling install."
+          }
+        end
+
+        if agent_browser_outdated?
+          current = `agent-browser --version 2>/dev/null`.strip.split.last
+          return {
+            error: "agent-browser version too old",
+            message: "agent-browser #{current} is installed but version >= #{MIN_AGENT_BROWSER_VERSION} is required.",
+            instructions: "Tell the user: 'agent-browser needs to be upgraded from #{current} to #{MIN_AGENT_BROWSER_VERSION}+. Run `browser(command: \"install\")` to upgrade — this may take a minute. Would you like me to upgrade it now?' Wait for user confirmation before calling install."
+          }
         end
 
         # Default to built-in browser (isolated=true). Only use user's Chrome when explicitly isolated=false.
@@ -94,12 +122,6 @@ module Clacky
           result = Shell.new.execute(command: full_command, hard_timeout: BROWSER_COMMAND_TIMEOUT, working_dir: working_dir)
         end
 
-        if playwright_missing?(result)
-          pw_result = install_playwright_chromium
-          return pw_result if pw_result[:error]
-          result = Shell.new.execute(command: full_command, hard_timeout: BROWSER_COMMAND_TIMEOUT, working_dir: working_dir)
-        end
-
         if use_auto_connect && !result[:success] && connection_error?(result)
           if we_launched_chrome
             result = Shell.new.execute(command: full_command, hard_timeout: BROWSER_COMMAND_TIMEOUT, working_dir: working_dir)
@@ -363,52 +385,23 @@ module Clacky
         !!find_in_path(AGENT_BROWSER_BIN)
       end
 
-      def auto_install_agent_browser
-        npm = find_or_install_npm
-        unless npm
-          return {
-            error: "agent-browser not installed",
-            message: "agent-browser is required for browser automation but is not installed. " \
-              "Node.js not found; tried to install via mise but failed.\n\n" \
-              "Please run: mise install node@22 && mise use -g node@22"
-          }
-        end
-
-        result = Shell.new.execute(command: "#{npm} install -g agent-browser", hard_timeout: 120)
-        unless result[:success]
-          return {
-            error: "Failed to auto-install agent-browser",
-            message: "npm install -g agent-browser failed: #{result[:stderr]}\n\nPlease run it manually."
-          }
-        end
-
-        {}
-      end
-
-      def find_or_install_npm
-        npm = find_in_path("npm")
-        return npm if npm
-
-        mise = find_mise_bin
-        return nil unless mise
-
-        path = `#{Shellwords.escape(mise)} which npm 2>/dev/null`.strip
-        return path if path && !path.empty? && File.executable?(path)
-        system(mise, "install", "node@22", out: File::NULL, err: File::NULL)
-        system(mise, "use", "-g", "node@22", out: File::NULL, err: File::NULL)
-
-        path = `#{Shellwords.escape(mise)} which npm 2>/dev/null`.strip
-        return path if path && !path.empty? && File.executable?(path)
-
-        nil
+      def agent_browser_outdated?
+        version = `agent-browser --version 2>/dev/null`.strip.split.last
+        return false if version.nil? || version.empty?
+        Gem::Version.new(version) < Gem::Version.new(MIN_AGENT_BROWSER_VERSION)
+      rescue StandardError
+        false
       end
 
-      def find_mise_bin
-        mise = find_in_path("mise")
-        return mise if mise
-
-        candidate = "#{Dir.home}/.local/bin/mise"
-        File.executable?(candidate) ? candidate : nil
+      def do_install_agent_browser
+        script = File.expand_path("../../../../scripts/install_agent_browser.sh", __FILE__)
+        result = Shell.new.execute(command: "bash #{Shellwords.escape(script)}", hard_timeout: 180)
+        if result[:success]
+          version = `agent-browser --version 2>/dev/null`.strip.split.last
+          { success: true, message: "agent-browser #{version} installed successfully. You can now use browser commands." }
+        else
+          { error: "Failed to install agent-browser", message: result[:stdout].to_s.strip }
+        end
       end
 
       def command_name_for_temp(command)
@@ -445,25 +438,6 @@ module Clacky
         { content: first_part.join + notice, temp_file: temp_file }
       end
 
-      def playwright_missing?(result)
-        output = "#{result[:stdout]}#{result[:stderr]}"
-        output.include?("Executable doesn't exist") ||
-          output.include?("Please run the following command to download new browsers")
-      end
-
-      def install_playwright_chromium
-        playwright = find_in_path("playwright")
-        cmd = playwright ? "#{playwright} install chromium" : "npx playwright install chromium"
-
-        result = Shell.new.execute(command: cmd, hard_timeout: 300)
-        unless result[:success]
-          return {
-            error: "Failed to install Playwright Chromium",
-            message: "Automatic browser installation failed. Please run manually:\n  npx playwright install chromium"
-          }
-        end
-        {}
-      end
     end
   end
 end

data/lib/clacky/ui2/layout_manager.rb

@@ -59,8 +59,15 @@ module Clacky
               screen.clear_line
             end
 
-            # Re-render fixed areas at new position
-            render_fixed_areas
+            # When input is paused (InlineInput active), fixed_area_start_row has grown
+            # (input_area.required_height returns 0 while paused), so the cleared rows
+            # now belong to the output area. Re-render output from buffer to fill them in.
+            if input_area.paused?
+              render_output_from_buffer
+            else
+              # Re-render fixed areas at new position
+              render_fixed_areas
+            end
             screen.flush
           end
         end

data/lib/clacky/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Clacky
-  VERSION = "0.9.0"
+  VERSION = "0.9.2"
 end

data/scripts/install.sh

@@ -590,40 +590,9 @@ main() {
 # Install agent-browser (browser automation tool)
 # This step is optional — failures are silently skipped with a hint.
 install_agent_browser() {
-    print_step "Installing agent-browser (optional)..."
-
-    # Try to find npm; if missing, attempt to install Node.js via mise
-    if ! command_exists npm; then
-        if command_exists mise || [ -x "$HOME/.local/bin/mise" ]; then
-            print_info "Installing Node.js via mise..."
-            "$HOME/.local/bin/mise" install node@22 > /dev/null 2>&1 || true
-            "$HOME/.local/bin/mise" use -g node@22 > /dev/null 2>&1 || true
-            # Reload mise so node/npm are on PATH
-            eval "$("$HOME/.local/bin/mise" activate bash 2>/dev/null)" 2>/dev/null || true
-        fi
-    fi
-
-    if ! command_exists npm; then
-        print_warning "agent-browser skipped (Node.js/npm not found)."
-        print_info "To enable browser automation later:"
-        print_info "  mise install node@22 && mise use -g node@22 && npm install -g agent-browser"
-        return 0
-    fi
-
-    if command_exists agent-browser; then
-        print_success "agent-browser already installed"
-        return 0
-    fi
-
-    print_info "Running: npm install -g agent-browser"
-    if npm install -g agent-browser > /dev/null 2>&1; then
-        print_success "agent-browser installed"
-    else
-        print_warning "agent-browser installation failed — skipping."
-        print_info "To install manually later: npm install -g agent-browser"
-    fi
-
-    return 0
+    local script_dir
+    script_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+    bash "$script_dir/install_agent_browser.sh" || true
 }
 
 # Post-installation information

data/scripts/install_agent_browser.sh

@@ -0,0 +1,67 @@
+#!/bin/bash
+# Install or upgrade agent-browser (browser automation tool)
+# Can be sourced by install.sh or run standalone (e.g. from Ruby via shell)
+#
+# Exit codes:
+#   0 — success (installed/upgraded)
+#   1 — failure (npm not found or install failed)
+
+set -e
+
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+BLUE='\033[0;34m'
+NC='\033[0m'
+
+print_info()    { echo -e "${BLUE}ℹ${NC} $1"; }
+print_success() { echo -e "${GREEN}✓${NC} $1"; }
+print_warning() { echo -e "${YELLOW}⚠${NC} $1"; }
+print_error()   { echo -e "${RED}✗${NC} $1"; }
+print_step()    { echo -e "\n${BLUE}==>${NC} $1"; }
+
+command_exists() { command -v "$1" >/dev/null 2>&1; }
+
+print_step "Installing agent-browser..."
+
+# Try to find npm; if missing, attempt to install Node.js via mise
+if ! command_exists npm; then
+    mise_bin=""
+    if command_exists mise; then
+        mise_bin="mise"
+    elif [ -x "$HOME/.local/bin/mise" ]; then
+        mise_bin="$HOME/.local/bin/mise"
+    fi
+
+    if [ -n "$mise_bin" ]; then
+        print_info "Installing Node.js via mise..."
+        "$mise_bin" install node@22 > /dev/null 2>&1 || true
+        "$mise_bin" use -g node@22 > /dev/null 2>&1 || true
+        eval "$("$mise_bin" activate bash 2>/dev/null)" 2>/dev/null || true
+    fi
+fi
+
+if ! command_exists npm; then
+    print_error "agent-browser installation failed: Node.js/npm not found."
+    print_info "Please run: mise install node@22 && mise use -g node@22 && npm install -g agent-browser"
+    exit 1
+fi
+
+print_info "Running: npm install -g agent-browser"
+if npm install -g agent-browser > /dev/null 2>&1; then
+    version=$(agent-browser --version 2>/dev/null | awk '{print $NF}')
+    print_success "agent-browser ${version} installed/updated"
+else
+    print_error "agent-browser installation failed."
+    print_info "To install manually: npm install -g agent-browser"
+    exit 1
+fi
+
+print_info "Installing Playwright Chromium..."
+if npx playwright install chromium > /dev/null 2>&1; then
+    print_success "Playwright Chromium installed"
+else
+    print_error "Playwright Chromium installation failed."
+    print_info "To install manually: npx playwright install chromium"
+    exit 1
+fi

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: openclacky
 version: !ruby/object:Gem::Version
-  version: 0.9.0
+  version: 0.9.2
 platform: ruby
 authors:
 - windy
@@ -397,6 +397,7 @@ files:
 - lib/clacky/web/version.js
 - lib/clacky/web/ws.js
 - scripts/install.sh
+- scripts/install_agent_browser.sh
 - scripts/uninstall.sh
 - sig/clacky.rbs
 homepage: https://github.com/yafeilee/clacky