/no_think
You are performing a deep relevance evaluation on a text chunk.

Your task:
1. Read the chunk provided earlier in the conversation (the seed).
2. Evaluate how well it answers or relates to the following query:
   "$deepquery"

Rules for relevance:
- Base all decisions ONLY on the chunk text. Never use external knowledge.
- If the chunk contains no relevant information, relevance score = 1.
- When relevant, identify precise supporting line ranges (inclusive).
- Be conservative when choosing line boundaries.
- Do not quote or rewrite the chunk.

Then perform a second task.

SECOND TASK – RELEVANT OUTLINE ONLY:
- Rewrite ONLY the relevant portions of the chunk as an outline.
- Each outline item MUST reference exact line number ranges.
- Do NOT include irrelevant sections.
- Do NOT rewrite the original text.
- Do NOT infer or summarize beyond what the chunk supports.
- If nothing is relevant, return no outline items.

OUTPUT FORMAT (exact, no extra text):

Relevance:
Score: <1–100>
Lines: [start–end], [start–end]

Outline:
[lines] Short structural label
[lines] Short structural label