find-text

Find Text

Use this skill when the user wants to read a specific passage they can name or roughly describe. The goal is to return the text itself (Hebrew source, English translation, or both), not to discover new sources — for that, use find-reference.

When to invoke

Trigger on requests like:

• "Show me Genesis 1:1"
• "What does Berakhot 2a say?"
• "Give me the Hebrew of Shema Yisrael" → Deuteronomy 6:4
• "Read me Pirkei Avot 1:1 in English"
• "What's the text of Rashi on Genesis 1:1?"
• "Show me Mishneh Torah, Hilchot Shabbat 1:1"

Skip for open-ended topical questions ("what does Judaism say about X") — that's find-reference.

Procedure

1. Resolve the reference. If the user's wording is ambiguous, partial, or transliterated unusually, call clarify_name_argument (with type_filter: "ref" when looking for a textual citation) to get the canonical Sefaria reference string before fetching. Examples that benefit from clarification: "the part of Avot about Hillel", "Sanhedrin daf 90", "Rambam on repentance".

2. Fetch the text. Call get_text with the canonical reference. Pass version_language:

   • "source" — Hebrew/Aramaic only
   • "english" — English translation only
   • "both" — side-by-side (default for most reading requests)
   • omit — every available version

3. Translation choice. If the user wants a particular English translation or wants to see all available translations side-by-side, call get_english_translations instead of (or in addition to) get_text.

4. Commentary on the passage. If the user wants to read what commentators say on this verse/sugya (Rashi, Ramban, Tosafot, Steinsaltz, etc.), call get_links_between_texts with with_text: "1" to pull the linked commentary text in one go. Filter or summarise the returned links by commentator if the response is large.

5. Manuscripts. If the user asks about historical manuscript witnesses or wants to see a scanned page, call get_available_manuscripts and offer get_manuscript_image for a specific image URL.

6. Bibliographic / structural questions. If the user asks "how is this book organised" or "who wrote this and when", use get_text_catalogue_info (index data) or get_text_or_category_shape (hierarchy) — not get_text.

Presentation

• For Hebrew + English, render Hebrew right-aligned in a quote block followed by the English. Preserve verse numbers / daf-amud markers from Sefaria's response.
• Hebrew text from Sefaria includes nikkud (vowel points) by default — keep them. Strip only if the user explicitly asks for unvocalized text.
• Always cite the canonical reference Sefaria returned, plus a Sefaria URL of the form https://www.sefaria.org/<reference-with-underscores> so the user can open it in a browser.
• For long passages (full chapters, full daf), summarise structure first and ask whether to dump the full text, unless the user explicitly asked for the complete passage.

Edge cases

• Talmud daf references use side notation: Berakhot 2a, Berakhot 2b. If a user says "Berakhot 2", clarify which side, or fetch both and present them.
• Mishneh Torah / Shulchan Aruch require the full path: Mishneh Torah, Teshuvah 3:4, Shulchan Aruch, Orach Chayim 1:1. Use clarify_name_argument if unsure.
• Zohar references use the section + page format: Zohar 1:15a.
• If get_text returns empty for a valid-looking reference, the version may not exist in that language — fall back to version_language: null to see what's available, or call get_english_translations to enumerate.