v0.7.0

• Added: body_chars cap on read_topic to spend less context
• Fixed: elide_quotes missed Outlook/Exchange/Apple (no->) quoting
• Changed: route disposition/resolution questions to the chair messages
• Added: read_topic thread map as a saturation signal
• Fixed: stop presenting [mlkem] as a subject prefix every corpus has
• Added: completeness signal and relevance scores on read_topic
• Fixed: SKILL message-numbering note was stale for read_topic
• Added: summarise the effective gather config at start
• Changed: document the new query filters and server knobs
• Added: most-active-threads heat signal in overview
• Added: read_digest sort=activity and exclude_mechanical filters
• Changed: fold mechanical events out of overview recent activity
• Changed: explain the overview citation counts
• Added: overview flags drafts blocked on an IESG DISCUSS
• Fixed: harden MCP tools against invalid input
• Added: server-side per-call timeout so stuck tools fail fast
• Changed: number read_topic messages globally across the narrative
• Fixed: fetch_by_url advertised mail URLs the corpus never stores
• Changed: tally_positions recognises WGLC review phrasings
• Fixed: find_replies rendered each message header twice
• Changed: collapse older draft revisions in search_corpus by default
• Fixed: list_labels example prefix and read_digest missing-kind message
• Fixed: drop bogus authors from "Additional contact information:" sub-headings
• Added: separate active Internet-Drafts from concluded work in overview
• Changed: help clients load the ietf-llm tool set in one search
• Changed: show the gather date on every top-level response
• Changed: surface the last ~10 timeline events in overview
• Changed: split Internet-Drafts from published RFCs in overview
• Added: writer->reader round-trip test for the threads digest
• Fixed: overview charter excerpt returning the file header
• Fixed: read_digest duplicating single-table digests and ignoring filters
• Changed: reflect the corpus subject line in README and ARCHITECTURE
• Added: per-corpus subject line to list_corpora and --list
• Changed: trim schema-duplicating prose from SKILL.md
• Changed: rename MCP list tool and wg param for the corpus model
• Changed: corpus-not-WG naming across README, CLI help, and ARCHITECTURE
• Changed: document --add-mentioned-drafts in README and ARCHITECTURE
• Added: --add-mentioned-drafts (pull drafts a corpus cites but lacks)
• Changed: document --new-drafts and --author in README and ARCHITECTURE
• Added: --author corpus (every draft a person has written)
• Added: --new-drafts subscription corpus (rolling --months window)
• Changed: document the corpus-kind model in ARCHITECTURE and README
• Added: MCP list_working_groups reports corpus kind and status
• Changed: --list shows corpus kind and status
• Added: infer non-WG corpora (list / custom) instead of a --list-only flag
• Changed: gather non-WG/RG groups (edwg, BoF) cleanly without charter noise