ThunderAI/claude-spec/05-options.md

Options & Settings System

Overview

All extension preferences are stored in browser.storage.local. Defaults and the full list of valid keys are defined in options/mzta-options-default.js.

Key Exports from mzta-options-default.js

Export	Description
prefs_default	All preference keys with their default values
integration_options_config	Per-provider API settings structure
getDynamicSettingsDefaults(keysFilter)	Returns per-special-prompt integration defaults
getDynamicSettingValue(prefs, prefix, settingName)	Reads a prefixed setting for a special prompt

Settings Structure

Global Integration Settings

Stored flat in prefs_default with {provider}_{key} naming:

chatgpt_api_key, chatgpt_model, chatgpt_developer_messages, chatgpt_temperature, chatgpt_store
ollama_host, ollama_model, ollama_num_ctx, ollama_temperature, ollama_think
openai_comp_host, openai_comp_model, openai_comp_api_key, openai_comp_use_v1, openai_comp_chat_name, openai_comp_temperature
google_gemini_api_key, google_gemini_model, google_gemini_system_instruction, google_gemini_thinking_budget, google_gemini_temperature
anthropic_api_key, anthropic_model, anthropic_version, anthropic_max_tokens, anthropic_system_prompt, anthropic_temperature

Plus the global connection selector:

connection_type   (default: 'chatgpt_web')
use_specific_integration   (default: false)

Special Prompt Integration Overrides

The 6 special prompts (add_tags, spamfilter, summarize, get_calendar_event, get_task, translate) each get their own use_specific_integration and connection_type keys:

{prefix}_use_specific_integration   (default: false)
{prefix}_connection_type            (default: 'chatgpt_api')

These are generated programmatically at the bottom of mzta-options-default.js using special_prompts_with_integration array.

UI & Feature Preferences

Key	Default	Description
do_debug	false	Enable debug logging
chatgpt_win_height	800	ChatGPT window height
chatgpt_win_width	700	ChatGPT window width
chatgpt_win_top	''	Window top position
chatgpt_win_left	''	Window left position
chatgpt_win_save_position	false	Remember window position
default_chatgpt_lang	''	Force response language
default_sign_name	''	Default signature name
reply_type	'reply_all'	Default reply type
composing_plain_text	false	Use plain text in compose
chatgpt_web_model	''	ChatGPT Web model override
chatgpt_web_tempchat	false	Use temporary chat
chatgpt_web_project	''	ChatGPT Web project
chatgpt_web_custom_gpt	''	Custom GPT URL
chatgpt_web_load_wait_time	1000	Wait time (ms) for ChatGPT page
dynamic_menu_force_enter	false	Force Enter to submit in popup
dynamic_menu_order_alphabet	true	Internal migration flag only; no UI. Set to false by migrateMenuOrderAlphabetic() on first boot after upgrade to bootstrap position-based ordering. See claude-spec/02-prompts.md for details.
placeholders_use_default_value	false	Use placeholder defaults when empty
max_prompt_length	30000	Max prompt string length

Feature Flags

Key	Default	Description
add_tags	false	Enable auto-tagging feature
add_tags_maxnum	3	Max tags to apply
add_tags_hide_exclusions	false	Hide excluded tags from menu
add_tags_exclusions_exact_match	false	Exact match for exclusions
add_tags_first_uppercase	true	Capitalize first letter of tags
add_tags_force_lang	true	Force language for tags
add_tags_auto	false	Auto-tag on message open
add_tags_auto_force_existing	false	Only use existing tags
add_tags_auto_only_inbox	true	Auto-tag only inbox messages
add_tags_auto_uselist	false	Use tag allow-list
add_tags_auto_uselist_list	''	Tag allow-list content
add_tags_enabled_accounts	[]	Accounts where auto-tag is active
get_calendar_event	true	Enable calendar event extraction
get_calendar_event_from_clipboard	false	Enable calendar from clipboard
get_task	true	Enable task creation
calendar_enforce_timezone	false	Force specific timezone
calendar_timezone	''	Timezone to enforce
calendar_no_selection	false	Skip selection prompt
spamfilter	false	Enable spam filter
spamfilter_threshold	70	Spam confidence threshold (%)
spamfilter_enabled_accounts	[]	Accounts where spam filter is active
spamfilter_show_msg_panel	true	Show info panel on spam detection
summarize	false	Enable email summarization
summarize_auto	1	Auto-summarize mode: 0 = disabled, 1 = manual (show "click to generate" button), 2 = automatic (generate on message open), 3 = generate on email receive (background pre-cache via onNewMailReceived, no UI during generation)
summarize_display_mode	'inline'	Where to display summaries: 'inline' = message pane banner, 'webchat' = AI chat window. Note: summarize_auto = 2 and summarize_auto = 3 always use inline regardless of this setting.
summarize_max_display_length	0	Maximum characters shown in inline summary before truncation. 0 = no limit (show full text). When set, text is truncated at a word boundary and a "See more"/"See less" toggle link is shown.
summarize_strip_formatting	false	Strip HTML and Markdown formatting from AI-generated summaries, showing plain text only.
translate	true	Enable email translation
translate_auto	0	Auto-translate mode: 0 = disabled, 1 = manual (show button), 2 = automatic (translate on message open), 3 = generate on email receive (background pre-cache via onNewMailReceived, no UI during generation)
translate_max_display_length	0	Maximum characters shown in inline translation before truncation. 0 = no limit (show full text). When set, text is truncated at a word boundary and a "See more"/"See less" toggle link is shown.
translate_lang	''	Target language for translation. Falls back to default_chatgpt_lang if empty.

Summarize Settings Page (pages/summarize/)

The summarize settings page provides:

1. Specific integration checkbox — enables per-feature API override (like other special prompts)
2. Auto-summarize dropdown (summarize_auto) — three modes:
   • 0 (Disabled) — no inline summaries
   • 1 (Manual) — shows a "Click to generate summary" button in message display
   • 2 (Automatic) — generates summary immediately when message is opened
3. Display mode dropdown (summarize_display_mode) — controls where summaries are shown:
   • 'inline' — summary banner in the message pane (default)
   • 'webchat' — opens the AI chat window
   • Note: summarize_auto = 2 always generates inline regardless of this setting. Context menu summarize with multiple messages always falls back to webchat.
4. Max display length (summarize_max_display_length) — number input, limits inline summary text to N characters. 0 = no limit. When truncated, a "See more"/"See less" toggle link is appended.
5. Strip formatting (summarize_strip_formatting) — checkbox, removes HTML/Markdown formatting from AI summary responses, displaying plain text only. Default: off.
6. Three editable prompts (used by context menu summarize and webchat mode):
   • Summarize instruction prompt (prompt_summarize)
   • Email template prompt (prompt_summarize_email_template)
   • Email separator prompt (prompt_summarize_email_separator)
   • Each has Save/Reset buttons and placeholder autocomplete
   • Default text comes from i18n strings (prompt_summarize_full_text, etc.)

Menu Order Page (pages/menu_order/)

Entry point from the options page via the "Menu Order" button (next to "Manage your prompts"). Provides drag-and-drop reordering and toggle-based visibility control for both the popup and the context menu. See claude-spec/02-prompts.md ("Menu Order Page") for the full behaviour, data flow, and exclusion rules.

Translate Settings Page (pages/translate/)

The translate settings page provides:

1. Specific integration checkbox — enables per-feature API override (like other special prompts)
2. Auto-translate dropdown (translate_auto) — three modes:
   • 0 (Disabled) — no inline translations
   • 1 (Manual) — shows a "Get AI Translation" button in message display
   • 2 (Automatic) — generates translation immediately when message is opened
3. Max display length (translate_max_display_length) — number input, limits inline translation text to N characters. 0 = no limit. When truncated, a "See more"/"See less" toggle link is appended.
4. Target language (translate_lang) — text input for the destination language. If empty, falls back to default_chatgpt_lang.
5. One editable prompt — the translation instruction prompt (prompt_translate_this) with Save/Reset buttons and placeholder autocomplete. Default text comes from i18n string prompt_translate_this_full_text.

Adding a New Preference

1. Add the key and default value to prefs_default in options/mzta-options-default.js
2. Add UI control to options/mzta-options.html
3. Add load/save logic to options/mzta-options.js
4. Add i18n label to _locales/en/messages.json
5. Read the pref in the relevant module via browser.storage.local.get()

Reading Preferences at Runtime

const prefs = await browser.storage.local.get(prefs_default);
// prefs now contains all keys with defaults for any unset values
const myPref = prefs.my_new_pref;