v0.62.0

0.62.0 (2026-10-28)

Features

• briefing: an imported latest gets a head note and the active topic says imported (e9e37ca)
• cli: daimon import serializes the given transcripts of one project in one bounded run (16508e6)
• import: a window of captured sessions names the way to older history, in the plan and in both CLI reference rows (45b1167), closes #1236
• import: discover a project's transcripts on disk and plan the N newest, skips applied inside that window (df85e93), closes #1236
• ledger: an import marker opens an attempt that heal never retries (9cefcb2), closes #1236
• llm: capture_cost stamps how many calls the backends answered (de92774)
• provenance: a session key per host, and the hosts whose transcripts record no project (712b257)
• provenance: public HOSTS and roots, one listing pattern per root, and the import table iterates the host set (80b2910), closes #1236
• provenance: SourceResolver.transcripts lists each host's files from its own roots (3e68cbf), closes #1236
• serializer: count the model calls a transcript's serialize will make before any spend (4c88aea)
• serialize: the import door stamps capture_door and marks its success line (381f577), closes #1236
• transcript: read the working directory a transcript recorded (5c2c900)
• view: open reports the import door count and when the newest imported session ended (9a00602)

Bug Fixes

• briefing: a planted age key on disk is dropped before the stamp (c280410)
• briefing: the import note is set only once the body has something to show (82a6f7d)
• briefing: the import note travels with the body and its facts are read on demand (313c85a)
• carry: a capture behind the stream's latest folds nothing backward (aabb447), closes #1236
• import: a transcript root the user cannot list is one named plan line, and the other roots still list (3c84abe), closes #1236
• import: a transcript the capture reader cannot decode is the named unreadable skip, before any spend (9bb0f3e), closes #1236
• import: bound on spent calls, key sessions by host, accept no-project hosts on a typed --project (5229987)
• import: one stat decides a missing transcript, and an unreadable one is a named skip (5c581fc)
• import: the reason a rootless host cannot be imported is provenance's, Gemini reads no registered root, and a Gemini path given by hand is refused by name (ab8b5c8), closes #1236
• ledger: any success line marks a session captured for the import's done-set (0879f41)
• ledger: every spawn line owns its attempt's reason, and a held import names its own cure (9c94ab2), closes #1236
• provenance: a Codex rollout present in both sessions and archived_sessions lists once, from sessions (dcaaace), closes #1236
• store: an older checkpoint never inherits a birth stamp newer than its own session (603ad91), closes #1236
• transcript: recorded_cwd stops at the first row that records one, and both readers answer None on undecodable bytes (df7d4c5), closes #1236
• transcript: the end stamp is the newest stamp in the tail, and the first row is read no further than the tail size (a3bad07), closes #1236

Performance Improvements

• transcript: last_timestamp reads a bounded tail and falls back to the whole file only when the tail has no stamp (dbde9b8), closes #1236

Documentation

• changelog: list each 0.61.0 change once (6357cb6), closes #1176
• claims: count the quote downgrade rate over fresh items only, 11.3% (385a3b0)
• cli: document daimon import in the reference, with the Spanish mirror (fc49c80)
• cli: the import row quotes the not-discovered line and says Gemini and Hermes cannot be imported yet, EN and ES (9babbea), closes #1236
• cli: the import row says about, spent versus planned, every refusal and exit code (e447900)
• cli: the import row says how discovery picks the N newest and which hosts stay path-given (e45b5fd), closes #1236
• hosts: the Claude Code page says import discovers sessions by the recorded working directory (7a6eac9)
• hosts: the Codex import section points at the quickstart in its own words (9ac0862)
• hosts: the Codex page says import discovers rollouts in both roots by the session_meta cwd (9c58087)
• hosts: the Gemini import cell reads no registered root (75e52a7)
• hosts: the Gemini page quotes its plan reason and describes the exit 2 refusal of a hand-given file (e01fea1)
• hosts: the Gemini page says a path given with --transcript is refused (80e4137)
• hosts: the Gemini page says import does not reach this host (f772ca8)
• hosts: the host table gains an Import column from the importer capability rows (d478046)
• hosts: the Kimi import section points at the quickstart in its own words (07b61ca)
• hosts: the Kimi page says import takes the wire log by path with the project typed (cd86c8c)
• hosts: the unsupported row says no registered root, and the Spanish mode labels read as table cells (09229d2)
• hosts: the Windsurf import section says transcript like the rest of its Spanish page and points at the quickstart in its own words (41f1990)
• hosts: the Windsurf page says import takes a transcript by path with the project typed (1d97d32)
• index: the landing page counts six quickstart steps (d496244)
• quickstart: a step fills the briefing from past sessions with daimon import, its plan, bound, resume and the imported-history note (501acbc)
• quickstart: Gemini is unreached by root, not by file, the compaction note is its own fact, the bound reads planned calls, and the plain path names NO_COLOR and DAIMON_PLAIN (7428dfa)
• quickstart: the plan's not-discovered line names gemini by its missing root and hermes by its missing file (163e2d4)
• readme: say what heal, the Windsurf adapter and team attribution do (6f2c381)
• schema: capture_door names its reader in the field table and the published schema (6c113b9)
• schema: describe capture_door as stamped with no reader yet (e3ada11), closes #1236
• schema: one constant names the capture_door key and the table says the count can shrink (8e28a6f)
• transcript: recorded_cwd names the one row shape where its line split could change the answer (a28130e), closes #1236