v0.7.1

tesserae ingest <file|url> — incremental single-source ingestion

Adds a first-class tesserae ingest command that merges one document or URL into the knowledge base without re-extracting the whole corpus, while remaining byte-identical to a full compile (parity-gated, with an automatic full-recompile fallback for correctness).

• New top-level tesserae ingest <input>... for local file paths and http(s) URLs (distinct from tesserae code ingest).
• URLs are fetched, converted to markdown, and persisted as tracked sources under data/ingested/<slug>.md with provenance front-matter (source_url, fetched_at, content_sha256, arxiv_id). URL ingest is behind a new optional extra: pip install tesserae[ingest-url].
• Default is an incremental fast path; --exact forces a full recompile. Flags: --exact, --dry-run, --title, --source-kind. Out-of-corpus files are copied into the tracked corpus.
• Internals: a per-invocation incremental_override seam (inert by default; engine parity preserved) + a golden-parity test gating the feature; YAML-safe provenance front-matter.

Full notes: docs/release-notes/v0.7.1.md (ko/zh/ja/ru/es/fr/de translations included).