Daily English Speaking Practice · публичный навык Open Skill
Daily English Speaking Practice
Daily English Speaking Practice — публичный Open Skill на Digen. Откройте, чтобы увидеть, что создаёт этот процесс.
Other
2 2 2
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4224455/2026-10-08/r449wby4.jpg
# Daily English Speaking Practice ## When to use - The user wants to improve their English speaking using daily-life phrases or sentences. - The user wants a set of natural, everyday English phrases/sentences for a real-life situation (ordering coffee, small talk, at the doctor, job interview, travel, shopping, etc.). - The user wants each phrase translated into a language they choose, and/or pronunciation audio. - The user wants a speaking-practice lesson card generated. ## Steps 1. **Recall context first.** Call `memory_query` to check whether the user's native language and preferred practice situations are already known. If found, use them as defaults; if not, ask the user: - Which language they want the English phrases translated into — **offer all languages as options** and let them pick (e.g. Chinese, Spanish, Hindi, French, Arabic, etc.). - Which situation/topic they want (e.g. ordering coffee, small talk, at the doctor, job interview, travel, shopping). If they don't specify, pick one common daily situation. 2. **Write the phrases.** Generate a set of ~10–15 short, natural, everyday English phrases for that situation, the kind a native speaker actually uses. Group them logically (e.g. greetings, key questions, common responses, useful expressions). Keep each phrase short and conversational. 3. **Translate.** Provide a natural translation of every phrase into the user's chosen language. 4. **Generate pronunciation audio.** For each phrase, call `generate_tts` with `model='tts.design'`, `text` = the English phrase, and a `voice_instruction` in English describing a clear, friendly, natural English voice (e.g. "a clear, warm, natural American English female voice, moderate pace, friendly and encouraging"). Pass `name` = the phrase text and `entity_type='voiceovers'`. Generate one audio per phrase so the user can hear each one. 5. **Create the lesson card.** Call `create_document` with `content` = a Markdown lesson card that includes: the situation, the phrases grouped by category, each phrase with its translation, a short pronunciation tip, and a brief example dialogue using the phrases. Pass `name` = "Daily English — <situation>" and `entity_type='lessons'`. 6. **Reply.** Summarize the lesson in natural language: list the phrases with their translations, point to the pronunciation audio and the saved lesson card. Keep it conversational; don't paste large blocks of URLs. 7. **Remember for next time.** Call `memory_store` to save the user's native language and preferred situations (kind `preference`), so next time you can skip asking. ## Notes - `generate_tts` `voice_instruction` must be written in English. - Use `model='tts.design'` for pronunciation audio (no reference audio needed). - Asset conventions: pronunciation audio uses `entity_type='voiceovers'`; lesson cards use `entity_type='lessons'`. - If the user wants to tweak an existing lesson (different situation, more phrases, different language), reuse the saved document via `get_document`/`update_document` on its `asset_id`. - Report failures honestly; never fabricate URLs or audio that wasn't generated.