ollama_chat 0.0.109 → 0.0.111

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: e6c85b95f4c5051b6b0f257dd8f63ee2a3997146c527fd8af251458fee641039
-  data.tar.gz: 45a66e25a1a92d8cbd86253a991ad28eeb48d1a8ca68c1f8bbb8df90bbbba37f
+  metadata.gz: 54b2bf70b7f843ad09b5de5f0881bf810ec78a08c79c5e1cd93d86aa75b68696
+  data.tar.gz: 33e475ea5e115609e02ba6b92cdf6a9b1b1dc6eccbfde6978fc80623daa3040b
 SHA512:
-  metadata.gz: a401173a7deb626af01ad8c9adf358259a6ffd7ea292644bb1a5a4553df2f70e9c74bb4ff0807d04914a7aa47e8140e81e0bd46745e33900962d8559950ea914
-  data.tar.gz: 6b04fcce611684e73bbe54cdadb01859fc4219f57960604a64c3037b414e56a02c7389a5bfef2341d0a2aa2a44d8a723761469f9c677ec8784c2c9d9e8b6fae0
+  metadata.gz: e7dad8f668a54b9c1e89937a445894eeb9872e1541d54dfc36fcf136d5c759bde7e0c21c9b818ce2e8ffa3bd7274163ada6b18553590dc812f3e89bdae839a27
+  data.tar.gz: 8bbd695f1eacc33c6928f8a382acdb8ed68b57348a0ce5199a9203bd7070df59f71a7a37e2862297c6c7dbc12e0d00aa9c46ef90f6392fc7579b87c8de438b7d

data/CHANGES.md

@@ -1,5 +1,118 @@
 # Changes
 
+## 2026-08-25 v0.0.111
+
+### Added
+
+*   `OllamaChat::Tools::ExecuteJiraTwg` tool
+    (`lib/ollama_chat/tools/execute_jira_twg.rb`) as a `twg` CLI wrapper,
+    featuring `Shellwords.shellsplit` argument parsing,
+    `OllamaChat::ExecuteError` handling for non-zero exits, and
+    `Kramdown::ANSI::Width.truncate` for feedback messages.
+*   `OC::OLLAMA::CHAT::TOOLS::JIRA::TWG` configuration constant, defaulting to
+    `which 2>/dev/null twg`.
+*   `execute_jira_twg` entry in `default_config.yml` with `default: false` and
+    `require_confirmation: true`.
+*   Two-step tool approval workflow in `FollowChat#handle_tool_calls`,
+    replacing single `chat.ask?` with `chat.confirm?` followed by `chat.ask?`
+    to allow single-keypress confirmation or free-text instructions.
+*   `session_sync` method in `session_management.rb` to persist messages,
+    links, history, update `working_directory`, and re-lock the session.
+*   `context_usage` method in `lib/ollama_chat/information.rb` to calculate and
+    format context usage, including `current_context_length` and
+    `context_filled` helpers.
+*   Context usage gauge in `info_session` displaying tokens used, total, and
+    percentage.
+*   `api_ps.json` asset containing a **27.3B** `qwen3.8:27b` model entry with a
+    **262144** context length.
+*   `spec/ollama_chat/tools/execute_jira_twg_spec.rb` with 8 specs covering
+    identity, execution, quoted arguments, nil guards, exception rescue, and
+    non-zero exits.
+*   `#session_sync` spec verifying ordered calls to
+    `store_messages_in_session`, `links.sync`, `save_history`, and
+    `session.lock`.
+
+### Changed
+
+*   `session_close` now delegates to `session_sync` before calling
+    `session.unlock`.
+*   `change_session` `ensure` block now uses `session_sync` instead of inline
+    `session.update(working_directory:)`.
+*   `set_new_session` now includes `ensure session.lock` to guarantee locking
+    on early return.
+*   `context_usage` logic extracted from `dynamic_runtime_information_values`
+    for improved maintainability.
+*   `session_management.rb` now logs the formatted `context_usage` string
+    instead of raw token counts and byte sizes.
+*   `self.register_name` in `lib/ollama_chat/tools/gem_path_lookup.rb`
+    simplified to use endless method definition syntax.
+*   Context ingestion logic now guards against empty file sets, emitting a
+    `STDERR` message and canceling if no files match the given patterns.
+*   `parsing_spec.rb` and `directory_structure_spec.rb` expected file counts
+    updated from **23** to **24**.
+
+### Fixed
+
+*   `current_context_length` now uses safe navigation operators to guard
+    against missing `ollama.ps` or `models`.
+*   `context_usage` body wrapped in a guard to return `nil` when context length
+    is unavailable.
+*   `which` stderr silenced in all OC config defaults using `2>/dev/null`.
+
+## 2026-08-20 v0.0.110
+
+### Added
+
+*   **App State Support**: Introduced `009_add_app_states_table.rb` migration
+    and `app_state.rb` model.
+*   **Prompt Synchronization**: Added `/prompt sync` subcommand with unified
+    diffs and boot check capabilities.
+    *   Implemented `AppState` model with `get`/`set` helpers and a `seed` boot
+        check using per-prompt `Digest::SHA256` hashes XOR-folded into a
+        fingerprint.
+    *   Boot check prints `~`/`+`/`-` detail report on drift and prompts user
+        to acknowledge; declining retains old fingerprint for repeated notice
+        on next boot.
+    *   Added `OC::DIFF_COMMAND` setting (`diff -u --color=always`) to `oc.rb`
+        for unified diff display, decoded via `Shellwords.split`.
+    *   Implemented `prompt_sync` and `show_prompt_diff` methods in
+        `prompt_management.rb` to compare DB prompts against shipped defaults,
+        show unified diff, offer `OC::DIFF_TOOL` resolution, and clean up
+        orphaned default prompts.
+    *   Wired `/prompt sync` subcommand into `commands.rb` (regexp, completion,
+        dispatch).
+*   **Test Mode**: Added `OllamaChat.test_mode` flag in `ollama_chat.rb` to
+    suppress interactive prompts during specs; set in `spec_helper.rb`.
+*   **Documentation**: Added YARD `@!attribute` docs for `test_mode` and
+    `test_mode?`.
+
+### Changed
+
+*   **Web Searching**: Updated `web_searching.rb` to store `search_engine` in
+    local `engine` to avoid double method call.
+*   **Session Management**: Added call to `session_close` after
+    `clean_messages!` succeeds in `lib/ollama_chat/commands.rb`.
+*   **Error Messages**: Simplified stale context error message in `PatchFile`
+    when a checksum mismatch occurs.
+
+### Fixed
+
+*   **Spec Expectations**: Updated `web_searching_spec.rb` to replace `allow`
+    with `expect` for message expectations.
+    *   Changed `log` expectations to `.at_least(:once)` since `links.add` also
+        triggers `log` internally.
+    *   Removed unused `before { allow(chat).to receive(:log) }` from the
+        `#web` describe block.
+    *   Added `.at_least(:once)` to `document_policy.selected` as it is called
+        in both `if` and `elsif` branches.
+    *   Changed `fetch_source`, `summarize`, and `import` expectations to
+        `.twice` since they are called once per URL (2 URLs in test data).
+
+### Tests
+
+*   Added 7 new spec examples for `prompt_sync` and `show_prompt_diff` in
+    `prompt_management_spec.rb`.
+
 ## 2026-08-18 v0.0.109
 
 ### Added

data/lib/ollama_chat/commands.rb

@@ -484,14 +484,15 @@ module OllamaChat::Commands
 
   command(
     name: :prompt,
-    regexp: %r(^/prompt(?:\s+(edit|info|add|delete|list|duplicate|import|export|reset|rename|-e))?(\s+(?:-[ef]|-c\s+(?:\w+|\?)))?(?:\s+([^-].*))?$),
-    complete: [ 'prompt', %w[ edit info add delete list duplicate import export reset rename ] ],
+    regexp: %r(^/prompt(?:\s+(edit|info|add|delete|list|duplicate|import|export|reset|rename|sync|-e))?(\s+(?:-[ef]|-c\s+(?:\w+|\?)))?(?:\s+([^-].*))?$),
+    complete: [ 'prompt', %w[ edit info add delete list duplicate import export reset rename sync ] ],
     optional: true,
     options: '[-c CONTEXT|-e|-f]',
     help: <<~EOT,
       📝 Manage prompt templates:
-         Subcommands: edit, info, add, delete, list,
-         duplicate, import, export, reset, rename.
+          Subcommands: edit, info, add, delete, list,
+          duplicate, import, export, reset, rename,
+          sync.
          Options: -c [context]
                      (? for interactive in /prompt),
                   -e (edit next)
@@ -544,6 +545,8 @@ module OllamaChat::Commands
           STDOUT.puts "No default value found for prompt #{bold{prompt.name}}."
         end
       end
+    when 'sync'
+      prompt_sync(context:)
     when nil, '-e'
       if prompt = choose_prompt(
           prompt: 'Which template shall guide the next response? %s',
@@ -591,14 +594,15 @@ module OllamaChat::Commands
 
   command(
     name: :conversation,
-    regexp: %r(^/conversation\s+(clean|save|load)(\s+-c)?(?:\s+([^-].*\.jsonl?))?$),
-    complete: [ 'conversation', %w[ save load clean ] ],
+    regexp: %r(^/conversation\s+(clean|compact|save|load)(\s+-c)?(?:\s+([^-].*\.jsonl?))?$),
+    complete: [ 'conversation', %w[ save load clean compact ] ],
     options: '[-c] [FILENAME]',
     help: <<~EOT
       💾 Save/Load conversation state
          (-c to clean before saving)
          FILENAME ends with .json or .jsonl
          Or clean inplace (removes tool content, images, and thinking)
+         Or compact (summarize old messages, keep recent)
     EOT
   ) do |subcommand,opts,path|
     if %w[ save load ].include?(subcommand) && path.blank?
@@ -619,10 +623,23 @@ module OllamaChat::Commands
         )
       then
         messages.clean_messages!
+        session_sync
         STDOUT.puts "Conversation cleaned."
       else
         STDOUT.puts 'Cancelled.'
       end
+    when 'compact'
+      if confirm?(
+          prompt: '🔔 Compact conversation? Old messages will be summarized. (y/n) ',
+          yes: /\Ay/i
+        )
+      then
+        messages.compact!
+        session_sync
+        STDOUT.puts "Conversation compacted."
+      else
+        STDOUT.puts 'Cancelled.'
+      end
     end
     :next
   end

data/lib/ollama_chat/database/migrations/009_add_app_states_table.rb

@@ -0,0 +1,11 @@
+Sequel.migration do
+  change do
+    create_table :app_states do
+      primary_key :id
+      String :key, null: false, unique: true
+      Text :value # JSON-encoded payload
+      DateTime :created_at, default: Sequel::CURRENT_TIMESTAMP, null: false
+      DateTime :updated_at, default: Sequel::CURRENT_TIMESTAMP, null: false
+    end
+  end
+end

data/lib/ollama_chat/database/models/app_state.rb

@@ -0,0 +1,122 @@
+require 'digest'
+
+# Represents a key/value pair of persistent application state stored in
+# the database.
+#
+# This model provides a generic, JSON-backed store for small bits of
+# runtime state that need to survive across sessions—such as the
+# XOR-SHA256 fingerprint of the shipped default prompts used for boot
+# drift detection.
+class OllamaChat::Database::Models::AppState < Sequel::Model(OllamaChat::DB)
+  plugin :timestamps, update_on_create: true
+  plugin :serialization, :json, :value
+  plugin :validation_helpers
+
+  # Validates the application state entry.
+  #
+  # Ensures that the `key` is present and unique.
+  def validate
+    super
+    validates_presence :key
+    validates_unique :key
+  end
+
+  # Retrieves the JSON payload stored under the given key.
+  #
+  # @param key [String, Symbol] the state key to look up
+  # @return [Hash, nil] the deserialized JSON payload, or nil if no
+  #   record exists for the key
+  def self.get(key)
+    where(key: key.to_s).first&.value
+  end
+
+  # Creates or updates the JSON payload stored under the given key.
+  #
+  # @param key [String, Symbol] the state key to write
+  # @param value [Hash, nil] the JSON-serializable payload to store
+  # @return [OllamaChat::Database::Models::AppState] the saved record
+  def self.set(key, value)
+    if found = where(key: key.to_s).first
+      found.update(value:)
+      found
+    else
+      create(key: key.to_s, value:)
+    end
+  end
+
+  # Computes per-prompt SHA256 hashes and an XOR fingerprint of all shipped
+  # default prompts.
+  #
+  # At boot, if the stored fingerprint differs from the current one, a detail
+  # report (modified / added / removed) is printed to STDERR and the user is
+  # prompted to accept the new fingerprint. If declined, the old fingerprint
+  # is retained and the notice will reappear on the next boot.
+  #
+  # This method is invoked automatically by the model seeding loop in
+  # `OllamaChat::Database.setup_models`.
+  #
+  # @param chat [OllamaChat::Chat] the chat instance providing the config
+  #   and `confirm?`
+  def self.seed(chat)
+    stored = get(:prompt_defaults_fingerprint)
+
+    stored_fingerprint = stored&.dig('fingerprint')
+
+    hashes = {}
+    chat.config.prompts.to_h.each do |context, prompts|
+      prompts.each do |name, content|
+        hashes["#{context}/#{name}"] =
+          Digest::SHA256.hexdigest(content.to_s)
+      end
+    end
+
+    fingerprint = hashes.values.reduce(0) do |sum, hex|
+      sum ^ hex.to_i(16)
+    end.to_s(16)
+
+    return true if fingerprint == stored_fingerprint
+
+    if stored&.dig('hashes')
+      old     = stored['hashes']
+      added   = hashes.keys - old.keys
+      removed = old.keys - hashes.keys
+      changed = hashes.select { |k, v| old[k] && old[k] != v }.keys
+
+      STDERR.puts "⚠️  Shipped prompts changed since last boot:"
+      changed.each { |k| STDERR.puts "  ~ #{k} (modified)" }
+      added.each   { |k| STDERR.puts "  + #{k} (new)" }
+      removed.each { |k| STDERR.puts "  - #{k} (removed)" }
+    elsif !OllamaChat.test_mode?
+      STDERR.puts '⚠️  First run — storing prompt fingerprints.'
+      set(:prompt_defaults_fingerprint,
+          { fingerprint:, hashes:, computed_at: Time.now.iso8601 })
+      return true
+    end
+
+    prompt = <<~EOT.chomp << ' '
+      Keep local prompts (stop nagging)?
+      Consider adopting via /prompt sync or /prompt reset. (y/n) %s
+    EOT
+    if OllamaChat.test_mode? || chat.confirm?(prompt:, yes: /\Ay/i, timeout: 5)
+      set(:prompt_defaults_fingerprint,
+          { fingerprint:, hashes:, computed_at: Time.now.iso8601 })
+    end
+
+    true
+  end
+
+  # @!attribute [v] id
+  #   @return [Integer] The primary key for the state entry.
+  #
+  # @!attribute [v] key
+  #   @return [String] The unique state key.
+  #
+  # @!attribute [v] value
+  #   @return [Hash, nil] The JSON-serialized state payload.
+  #
+  # @!attribute [v] created_at
+  #   @return [Time, nil] The timestamp when the state was created.
+  #
+  # @!attribute [v] updated_at
+  #   @return [Time, nil] The timestamp of the last update to the state.
+end

data/lib/ollama_chat/follow_chat.rb

@@ -137,18 +137,25 @@ class OllamaChat::FollowChat
       end
       STDOUT.puts
       confirmed = :implicit
-      function = JSON.pretty_generate(tool_call.function)
+      resolve   = nil
+      function  = JSON.pretty_generate(tool_call.function)
       chat.log(:info, "Tool call received", data: { tool: name, function: })
       if chat.tool_function(name).require_confirmation?
-        prompt = "🔔 I want to execute tool %s\n%s\nConfirm? (y/n) " % [
-          bold { name },
-          italic { function },
+        STDOUT.puts "🔔 I want to execute tool %s\n%s\n" % [
+          bold { name }, italic { function },
         ]
-        prompt.gsub!('%', '%%')
-        if chat.confirm?(prompt:, yes: /\Ay/i)
+        prompt  = '❓ Allow ✅[y]es / ⛔️[n]o / 📝[i]nstruct? '
+        answer  = chat.confirm?(prompt:)&.to_s&.downcase
+        resolve = 'You **MUST** ask the user for instructions on how to proceed!!!'
+        case answer
+        when 'y'
           confirmed = :explicit
+        when 'n'
+          confirmed = :denied
         else
           confirmed = :denied
+          instr     = chat.ask?(prompt: '📝 Instructions: ')&.strip
+          resolve   = instr.full? ||  resolve
         end
       else
         STDOUT.puts "Executing tool %s\n%s" % [
@@ -160,10 +167,7 @@ class OllamaChat::FollowChat
       result = nil
       case confirmed
       when :denied
-        result = JSON(
-          message: 'User denied confirmation!',
-          resolve: 'You **MUST** ask the user for instructions on how to proceed!!!',
-        )
+        result = JSON(message: 'User denied confirmation!', resolve:)
         STDOUT.printf(
           "\n%s Execution of tool %s denied by user.\n\n", ?🚫, bold { name }
         )
@@ -174,11 +178,7 @@ class OllamaChat::FollowChat
           "\n%s Execution of tool %s confirmed.\n\n", symbol, bold { name }
         )
         result = OllamaChat::Tools.registered[name].execute(tool_call, chat:)
-        if confirmed == :explicit
-          chat.log(:info, "Tool execution confirmed", data: { tool: name, confirmed: })
-        else
-          chat.log(:info, "Tool execution confirmed", data: { tool: name, confirmed: })
-        end
+        chat.log(:info, "Tool execution confirmed", data: { tool: name, confirmed: })
       end
 
       chat.tool_call_results[name] << result

data/lib/ollama_chat/information.rb

@@ -132,6 +132,12 @@ module OllamaChat::Information
       output.puts "  No persona selected."
     end
     output.puts "🧠 Current chat model is #{bold{@model}}."
+    context_usage = '%s of %s (%s)' % [
+      session.estimate_tokens.tokens_formatted,
+      format_tokens(current_context_length),
+      bold { '%.1f%%' % (100 * context_filled)},
+    ]
+    output.puts  "  Context Length: #{context_usage}"
     output.print '  '; think_mode.show(output:)
     output.print '  '; think_loud.show(output:)
     output.print '  '; think_strip.show(output:)
@@ -360,6 +366,7 @@ module OllamaChat::Information
       tools_support:        tools_support.on? ? 'enabled' : 'disabled',
       voice:                voice.on? ? 'enabled' : 'disabled',
       weekday:              now.strftime('%A'),
+      context_usage:        ,
     }
   end
 
@@ -373,4 +380,47 @@ module OllamaChat::Information
   def dynamic_runtime_information
     prompt(:dynamic_runtime_info).to_s % dynamic_runtime_information_values
   end
+
+  # Resolves the effective context window size for the current model.
+  #
+  # Prefers the session-level `num_ctx` override (set via `/model_options`
+  # or the stored model options profile). Falls back to the model's native
+  # `context_length` in the default profile or as reported by the Ollama server
+  # if possible.
+  #
+  # @return [Integer, NilClass] the context length in tokens or nil
+  def current_context_length
+    get_session_model_options[:num_ctx] ||
+    get_stored_model_options(@model)[:num_ctx] ||
+      ollama.ps&.models&.find { _1.name == @model }&.context_length
+  end
+
+  # Computes the fraction of the context window currently in use.
+  #
+  # Divides the estimated token count of the session's messages by the
+  # effective context length, clamped to the range 0.0–1.0.
+  #
+  # @return [Float] the ratio of used context (e.g. `0.73` for 73%)
+  def context_filled
+    es = session.estimate_tokens
+    (es.tokens.to_f / current_context_length).clamp(0..1).to_f
+  end
+
+  # Formats the current context usage as a human-readable string.
+  #
+  # Returns a string like `"167.7 KT of 262.1 KT (64.0%)"` combining
+  # the estimated tokens in use, the effective context length, and
+  # the percentage (via +context_filled+).
+  #
+  # @return [String, nil] the formatted usage string, or +nil+ if
+  #   the context length cannot be determined.
+  def context_usage
+    if cl = current_context_length
+      '%s of %s (%s)' % [
+        session.estimate_tokens.tokens_formatted,
+        format_tokens(current_context_length),
+        '%.1f%%' % (100 * context_filled),
+      ]
+    end
+  end
 end

data/lib/ollama_chat/input_content.rb

@@ -95,14 +95,22 @@ module OllamaChat::InputContent
     format = @context_format.selected
     myself = self
     if patterns
-      ContextSpook::generate_context(verbose: true, format:) do |context|
+      count = 0
+      ctx = ContextSpook::generate_context(verbose: true, format:) do |context|
         context do
           myself.file_set_each(patterns, all:) do |filename|
             filename.file? or next
             file filename.to_path
+            count += 1
           end
         end
-      end.send("to_#{format.downcase}")
+      end
+      if count > 0
+        ctx.send("to_#{format.downcase}")
+        STDOUT.puts "✅ Ingesting context now."
+      else
+        STDERR.puts "❌ No files in context. ⇨ Cancelled."
+      end
     else
       if context_filename = choose_filename('.contexts/*.rb')
         ContextSpook.generate_context(context_filename, verbose: true, format:).

data/lib/ollama_chat/oc.rb

@@ -1,5 +1,6 @@
 require 'const_conf'
 require 'pathname'
+require 'shellwords'
 
 # Environment configuration module for OllamaChat
 #
@@ -38,7 +39,7 @@ module OC
     description 'Pager command to use in case terminal lines are exceeded by output'
 
     default do
-      if fallback_pager = `which less`.full?(:chomp) || `which more`.full?(:chomp)
+      if fallback_pager = `which 2>/dev/null less`.full?(:chomp) || `which 2>/dev/null more`.full?(:chomp)
         fallback_pager << ' -r'
       end
     end
@@ -48,7 +49,7 @@ module OC
     description 'Editor to use'
 
     default do
-      if  editor = %w[ vim vi ].find { `which #{_1}`.full?(:chomp) }
+      if  editor = %w[ vim vi ].find { `which 2>/dev/null #{_1}`.full?(:chomp) }
         editor
       else
         warn 'Need an editor command configured via env var "EDITOR"'
@@ -60,15 +61,23 @@ module OC
     description 'Browser to use'
 
     default do
-      %w[ open xdg-open ].find { `which #{_1}` }.full?(:chomp)
+      %w[ open xdg-open ].find { `which 2>/dev/null #{_1}` }.full?(:chomp)
     end
   end
 
+  DIFF_COMMAND = set do
+    description 'Command (string) to generate a unified diff with'
+
+    required true
+    default { 'diff -u --color=always' }
+    decode { Shellwords.split(_1) if _1.present? }
+  end
+
   DIFF_TOOL = set do
     description 'Diff tool to apply changes with'
 
     default do
-      if  diff = `which vimdiff`.full?(:chomp)
+      if  diff = `which 2>/dev/null vimdiff`.full?(:chomp)
         diff
       else
         warn 'Need a diff tool configured via env var "DIFF_TOOL"'
@@ -184,7 +193,7 @@ module OC
 
         CTAGS_TOOL = set do
           description 'Tools ctags path'
-          default { `which ctags`.full?(:chomp) }
+          default { `which 2>/dev/null ctags`.full?(:chomp) }
           check   { value.blank? || File.exist?(value) }
         end
 
@@ -231,6 +240,11 @@ module OC
             sensitive   true
             required { OC::OLLAMA::CHAT::TOOLS::JIRA::URL? }
           end
+
+          TWG = set do
+            description 'Path to the twg CLI binary'
+            default { `which 2>/dev/null twg`.full?(:chomp) }
+          end
         end
 
         module IMAGE_GENERATOR

data/lib/ollama_chat/ollama_chat_config/default_config.yml

@@ -73,6 +73,7 @@ prompts:
           80%% of this width in your output of responses, especially when using
           markdown tables.
       - **Session**: %{session_name}
+        - Context usage is %{context_usage}.
         - Markdown output is %{markdown}. **Never** output markdown as your
           responses if it is disabled.
         - Voice output is %{voice}. Speak naturally and short if voice
@@ -240,7 +241,6 @@ prompts:
       How does this implementation handle malformed JSON or missing keys in the
       middle of a 2GB file without crashing the process?
 
-
       What is the time complexity of the current insert logic, and can we implement
       bulk inserts to reduce database round trips?
 
@@ -508,6 +508,10 @@ tools:
       result_display_timeout: 3
       default: false
       require_confirmation: true
+    execute_jira_twg:
+      result_display_timeout: 3
+      default: false
+      require_confirmation: true
     get_rfc:
       url: "https://www.rfc-editor.org/rfc/rfc%{rfc_id}.txt"
       result_display_timeout: 3