openclacky 1.2.8 → 1.2.10

data/lib/clacky/session_manager.rb

@@ -179,16 +179,17 @@ module Clacky
       deleted
     end
 
-    # Keep only the most recent N sessions by created_at; delete the rest.
-    # Returns count of deleted sessions.
+    # Keep only the most recent N non-pinned sessions by created_at; the rest
+    # are soft-deleted (moved to the session trash, recoverable). Pinned
+    # sessions are never deleted and do not count toward the cap.
+    # Returns count of soft-deleted sessions.
     def cleanup_by_count(keep:)
-      sessions = all_sessions # already sorted newest-first
-      return 0 if sessions.size <= keep
+      non_pinned = all_sessions.reject { |s| s[:pinned] } # already sorted newest-first
+      return 0 if non_pinned.size <= keep
 
-      sessions[keep..].each do |session|
-        filepath = File.join(@sessions_dir, generate_filename(session[:session_id], session[:created_at]))
-        _hard_delete_session_with_chunks(filepath) if File.exist?(filepath)
-      end.size
+      victims = non_pinned[keep..]
+      victims.each { |session| soft_delete(session[:session_id]) }
+      victims.size
     end
 
     # ── Session trash (delegates to Tools::TrashManager) ──────────────

data/lib/clacky/telemetry.rb

@@ -23,6 +23,11 @@ module Clacky
   #   POST /api/v1/telemetry/startup
   #   POST /api/v1/telemetry/task
   module Telemetry
+    LAUNCH_SOURCES = {
+      "installer" => "installer",
+      nil         => "cli"
+    }.freeze
+
     class << self
       # Called on every CLI startup (agent and server mode).
       # No local dedup — the server deduplicates by device_hash for unique
@@ -32,11 +37,13 @@ module Clacky
 
         brand = Clacky::BrandConfig.load
         payload = {
-          device_id:    resolve_device_id(brand),
-          version:      Clacky::VERSION,
-          os:           RbConfig::CONFIG["host_os"],
-          ruby_version: RUBY_VERSION,
-          brand:        brand.branded? ? brand.package_name : nil
+          device_id:     resolve_device_id(brand),
+          version:       Clacky::VERSION,
+          os:            RbConfig::CONFIG["host_os"],
+          ruby_version:  RUBY_VERSION,
+          brand:         brand.branded? ? brand.package_name : nil,
+          launch_source: LAUNCH_SOURCES.fetch(ENV["CLACKY_LAUNCHED_BY"], "cli"),
+          container:     detect_container
         }.compact
 
         fire_and_forget("/api/v1/telemetry/startup", payload)
@@ -56,7 +63,8 @@ module Clacky
         payload = {
           device_id: resolve_device_id(brand),
           version:   Clacky::VERSION,
-          brand:     brand.branded? ? brand.package_name : nil
+          brand:     brand.branded? ? brand.package_name : nil,
+          container: detect_container
         }
         payload.merge!(extract_task_metrics(result)) if result.is_a?(Hash)
 
@@ -74,6 +82,18 @@ module Clacky
         brand.device_id
       end
 
+      private def detect_container
+        return "docker" if File.exist?("/.dockerenv")
+
+        begin
+          cgroup = File.read("/proc/1/cgroup")
+          return "docker" if cgroup.include?("docker") || cgroup.include?("containerd")
+        rescue Errno::ENOENT, Errno::EACCES
+          nil
+        end
+        nil
+      end
+
       private def extract_task_metrics(result)
         cache = result[:cache_stats] || {}
         duration = result[:duration_seconds]

data/lib/clacky/ui2/layout_manager.rb

@@ -381,15 +381,19 @@ module Clacky
         render_all
       end
 
-      def cleanup_screen
+      def cleanup_screen(clear_screen: false)
         @render_mutex.synchronize do
-          fixed_start = fixed_area_start_row
-          (fixed_start...screen.height).each do |row|
-            screen.move_cursor(row, 0)
-            screen.clear_line
+          if clear_screen
+            screen.clear_screen(mode: :reset)
+          else
+            fixed_start = fixed_area_start_row
+            (fixed_start...screen.height).each do |row|
+              screen.move_cursor(row, 0)
+              screen.clear_line
+            end
+            screen.move_cursor([@output_row, 0].max, 0)
+            print "\r"
           end
-          screen.move_cursor([@output_row, 0].max, 0)
-          print "\r"
           screen.show_cursor
           screen.flush
         end

data/lib/clacky/ui2/ui_controller.rb

@@ -150,9 +150,9 @@ module Clacky
       end
 
       # Stop the UI controller
-      def stop
+      def stop(clear_screen: false)
         @running = false
-        @layout.cleanup_screen
+        @layout.cleanup_screen(clear_screen: clear_screen)
       end
 
       # Clear the input area

data/lib/clacky/ui_interface.rb

@@ -135,6 +135,6 @@ module Clacky
 
     # === Path redaction (for encrypted brand skill tmpdirs) ===
     # === Lifecycle ===
-    def stop; end
+    def stop(clear_screen: false); end
   end
 end

data/lib/clacky/utils/model_pricing.rb

@@ -350,80 +350,104 @@ module Clacky
         cache:  { write: 0.30, read: 0.03 }
       },
 
+      # M3 (released 2026-06-01) is MiniMax's multimodal flagship. Official
+      # pricing is tiered by context length (≤512K vs 512K–1M); per the
+      # project's "displayed ≤ actual" convention we record only the lowest
+      # (≤512K) tier as a flat rate — the global TIERED_PRICING_THRESHOLD is
+      # 200K, so applying the 512K–1M rate to the 200K–512K band would over-
+      # charge. Listed at original (non-promotional) prices: input $0.60,
+      # output $2.40, cache read $0.12 per 1M tokens.
+      "minimax-m3" => {
+        input:  { default: 0.60, over_200k: 0.60 },
+        output: { default: 2.40, over_200k: 2.40 },
+        cache:  { write: 0.60, read: 0.12 }
+      },
+
       "minimax-m2.7" => {
         input:  { default: 0.30, over_200k: 0.30 },
         output: { default: 1.20, over_200k: 1.20 },
         cache:  { write: 0.30, read: 0.06 }
       },
 
-      # Qwen (Alibaba DashScope) — USD per 1M tokens, Singapore region list price.
-      # Source: Alibaba Cloud Model Studio international pricing.
-      # Cache convention (mirrors DeepSeek/Kimi/GLM "displayed ≤ actual"):
-      #   - DashScope has two cache modes; implicit is auto-on, explicit is opt-in.
-      #     Implicit: write @ 100% input, read @ 20% input (no setup, no guarantee)
-      #     Explicit: write @ 125% input, read @ 10% input (cache_control marker)
-      #   - We bill writes at the regular input rate (matches implicit, and avoids
-      #     surprising users with the explicit 25% surcharge).
-      #   - We bill reads at 20% (implicit rate) — the conservative side; users on
-      #     explicit caching will see real bills slightly *lower* than displayed.
+      # Qwen (Alibaba DashScope) - USD per 1M tokens, international (Singapore) list price.
+      # Source: Alibaba Cloud Model Studio international console per-model pages.
+      #
+      # Pricing convention:
+      #   - These rates are used for user-facing cost ESTIMATION, so we always use
+      #     the standard LIST price and intentionally ignore any limited-time promo
+      #     discounts. A promo lowers the user's actual bill, never raises it, so
+      #     estimating at list price keeps the estimate a safe upper bound and avoids
+      #     churn whenever a promo starts or ends.
+      #   - We record the model's LOWEST context tier (e.g. input<=256k / <=128k) as a
+      #     flat rate, since the global TIERED_PRICING_THRESHOLD is 200K and does not
+      #     match Qwen's per-model breakpoints.
+      #   - cache.write = official explicit-cache-create price.
+      #   - cache.read  = official explicit-cache-hit price.
+      #   - When a model has NO published explicit-cache price (e.g. qwen3.6-27b,
+      #     qwen-plus-latest), cache.write/read fall back to the input rate.
+      # qwen3.7-max: NOT tiered (single flat tier per Alibaba's definition).
+      #   List price: input 2.5, output 7.5, explicit write 3.125, explicit read 0.25.
       "qwen3.7-max" => {
-        input:  { default: 1.20, over_200k: 1.20 },
-        output: { default: 6.00, over_200k: 6.00 },
-        cache:  { write: 1.20, read: 0.24 }
+        input:  { default: 2.5, over_200k: 2.5 },
+        output: { default: 7.5, over_200k: 7.5 },
+        cache:  { write: 3.125, read: 0.25 }
       },
 
+      # qwen3.7-plus: list price (<=256k tier):
+      #   input 0.4, output 1.6, explicit write 0.5, explicit read 0.04.
       "qwen3.7-plus" => {
-        input:  { default: 0.40, over_200k: 0.40 },
-        output: { default: 2.40, over_200k: 2.40 },
-        cache:  { write: 0.40, read: 0.08 }
-      },
-
-      "qwen3.7-flash" => {
-        input:  { default: 0.15, over_200k: 0.15 },
-        output: { default: 0.90, over_200k: 0.90 },
-        cache:  { write: 0.15, read: 0.03 }
+        input:  { default: 0.4, over_200k: 0.4 },
+        output: { default: 1.6, over_200k: 1.6 },
+        cache:  { write: 0.5, read: 0.04 }
       },
 
+      # qwen3.6-plus: list price (<=256k tier). Official explicit-cache prices.
+      #   input 0.50, output 3.00, explicit write 0.625, explicit read 0.05
       "qwen3.6-plus" => {
-        input:  { default: 0.40, over_200k: 0.40 },
-        output: { default: 2.40, over_200k: 2.40 },
-        cache:  { write: 0.40, read: 0.08 }
+        input:  { default: 0.50, over_200k: 0.50 },
+        output: { default: 3.00, over_200k: 3.00 },
+        cache:  { write: 0.625, read: 0.05 }
       },
 
+      # qwen3.6-max (qwen3.6-max-preview): list price (<=128k tier).
+      #   input 1.30, output 7.80, explicit write 1.625, explicit read 0.13
       "qwen3.6-max" => {
-        input:  { default: 1.20, over_200k: 1.20 },
-        output: { default: 6.00, over_200k: 6.00 },
-        cache:  { write: 1.20, read: 0.24 }
+        input:  { default: 1.30, over_200k: 1.30 },
+        output: { default: 7.80, over_200k: 7.80 },
+        cache:  { write: 1.625, read: 0.13 }
       },
 
+      # qwen3.6-27b: list price, no explicit-cache pricing published.
+      #   Cache write/read fall back to the input rate (no cache discount).
       "qwen3.6-27b" => {
-        input:  { default: 0.20, over_200k: 0.20 },
-        output: { default: 0.80, over_200k: 0.80 },
-        cache:  { write: 0.20, read: 0.04 }
+        input:  { default: 0.60, over_200k: 0.60 },
+        output: { default: 3.60, over_200k: 3.60 },
+        cache:  { write: 0.60, read: 0.60 }
       },
 
+      # qwen3.6-flash: list price (<=256k tier).
+      #   input 0.25, output 1.50, explicit write 0.3125, explicit read 0.025
       "qwen3.6-flash" => {
-        input:  { default: 0.15, over_200k: 0.15 },
-        output: { default: 0.90, over_200k: 0.90 },
-        cache:  { write: 0.15, read: 0.03 }
+        input:  { default: 0.25, over_200k: 0.25 },
+        output: { default: 1.50, over_200k: 1.50 },
+        cache:  { write: 0.3125, read: 0.025 }
       },
 
+      # qwen-plus-latest: list price (<=256k tier), no explicit-cache pricing.
+      #   Cache write/read fall back to the input rate (no cache discount).
       "qwen-plus-latest" => {
         input:  { default: 0.40, over_200k: 0.40 },
         output: { default: 1.20, over_200k: 1.20 },
-        cache:  { write: 0.40, read: 0.08 }
-      },
-
-      "qwen-vl-plus" => {
-        input:  { default: 0.14, over_200k: 0.14 },
-        output: { default: 0.41, over_200k: 0.41 },
-        cache:  { write: 0.14, read: 0.028 }
+        cache:  { write: 0.40, read: 0.40 }
       },
 
-      "qwen-vl-max" => {
-        input:  { default: 0.52, over_200k: 0.52 },
-        output: { default: 2.08, over_200k: 2.08 },
-        cache:  { write: 0.52, read: 0.104 }
+      # qwen3-vl-plus: replaces the retiring qwen-vl-plus. List price
+      #   (128k<input<=256k tier). input 0.60, output 4.80,
+      #   explicit write 0.75, explicit read 0.06.
+      "qwen3-vl-plus" => {
+        input:  { default: 0.60, over_200k: 0.60 },
+        output: { default: 4.80, over_200k: 4.80 },
+        cache:  { write: 0.75, read: 0.06 }
       },
 
     }.freeze
@@ -583,6 +607,8 @@ module Clacky
           "glm-4.7"
         # MiniMax — model ids in providers.rb use capitalised "MiniMax-M2.x"
         # but we match case-insensitively and map to the lowercased table key.
+        when /^minimax-m3$/i
+          "minimax-m3"
         when /^minimax-m2\.5$/i
           "minimax-m2.5"
         when /^minimax-m2\.7$/i
@@ -591,14 +617,12 @@ module Clacky
         # Qwen (Alibaba DashScope) — strict anchored match per registered
         # model id in providers.rb. qwen3.7-* is the latest flagship line;
         # qwen3.6-* are the previous generation; qwen-plus-latest is the
-        # rolling alias for the latest Qwen-Plus release; qwen-vl-* are
-        # the multimodal SKUs.
+        # rolling alias for the latest Qwen-Plus release; qwen3-vl-plus is
+        # the multimodal SKU (replaces the retired qwen-vl-plus/max).
         when /^qwen3\.7-max$/i
           "qwen3.7-max"
         when /^qwen3\.7-plus$/i
           "qwen3.7-plus"
-        when /^qwen3\.7-flash$/i
-          "qwen3.7-flash"
         when /^qwen3\.6-plus$/i
           "qwen3.6-plus"
         when /^qwen3\.6-max$/i
@@ -609,10 +633,8 @@ module Clacky
           "qwen3.6-flash"
         when /^qwen-plus-latest$/i
           "qwen-plus-latest"
-        when /^qwen-vl-plus$/i
-          "qwen-vl-plus"
-        when /^qwen-vl-max$/i
-          "qwen-vl-max"
+        when /^qwen3-vl-plus$/i
+          "qwen3-vl-plus"
 
         # OpenAI GPT-5.x models — match various dashed/dotted/compact forms
         # (e.g. "gpt-5.5", "gpt-5-5", "gpt5.5", "gpt55")

data/lib/clacky/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Clacky
-  VERSION = "1.2.8"
+  VERSION = "1.2.10"
 end