Commit graph

9 commits

Author SHA1 Message Date
nesquena-hermes
f7d727d061 fix(compress): make manual /compress durable across reload + restart (#4986)
Manual /compress shrinks the model-facing context_messages but keeps the
visible messages[] transcript. The old context returned via two paths (#4836):
(1) append-only state.db reconciliation re-appending pre-compression rows, and
(2) startup .bak recovery treating the intentional shrink as data loss.

Half A (routes.py _handle_session_compress): persist truncation_watermark +
truncation_boundary (= watermark of the compressed context), set
compression_anchor_mode="manual", refresh last_prompt_tokens, stamp missing
timestamps on the compressed context, and delete the now-stale .bak. The
boundary==watermark stamp drives the #4772 reconciliation logic onto its
conservative path (block replay of pre-compression rows) while post-compression
turns still merge once sidecar timestamps advance past the watermark.

Half B (session_recovery.py): the .bak recovery guard. MAINTAINER FIX over the
original PR — the contributor suppressed recovery whenever
compression_anchor_mode=="manual", a flag set once at compress and never
cleared, which PERMANENTLY disabled #1558 crash-recovery for any compressed
session (real data loss). Recovery is now suppressed only when the session was
intentionally compressed AND the .bak is genuinely the pre-compression backup,
discriminated by the compaction marker (the same _context_messages_include_
compression_marker signal reconciliation uses): a marked .bak post-dates the
compression -> recover; an unmarked .bak with a larger context is the
shrink-undoing pre-compression one -> suppress. Fail-open on any error.

Resolves an Opus-gate edge in the first cut (length-only heuristic wrongly
suppressed a loss that shrank BOTH messages and context). Two new non-vacuous
regression tests: post-compression real loss recovers, and the both-shrunk
marked-backup case recovers.

Co-authored-by: hyl-ailab <hyl-ailab@users.noreply.github.com>
2026-06-27 21:43:27 +00:00
nesquena-hermes
4bf9ff7518 stage #4594 onto master (v0.51.550) — preserve visible transcript after manual /compress (#3133)
Manual /compress wrote the compressed payload into BOTH s.messages (visible) and s.context_messages
(model-facing) + wiped s.tool_calls -> messages vanished from the window. Fix keeps s.messages +
tool history durable, deep-copies compressed only into s.context_messages, anchors off visible msgs.
Co-authored-by: franksong2702 <franksong2702@users.noreply.github.com>
2026-06-21 06:46:47 +00:00
Michael Lam
5bd1f1441d fix: route background worker credentials through profile env 2026-05-16 01:02:19 -07:00
Michael Lam
9894157120 fix: hybridize background profile env routing 2026-05-15 23:19:37 -07:00
starship-s
aa1c7c24f4 fix(profiles): route background aux workers via session profile 2026-05-15 03:02:42 -06:00
Frank Song
8fa92c680f Fix manual compression proxy timeouts 2026-05-12 17:33:59 +08:00
Frank Song
1bec8070f2 fix(1833): persist compression anchor summary for reload UI 2026-05-10 16:45:16 +08:00
nesquena-hermes
d3a686a266
fix(compress): prefer persisted reference handoff after completion — v0.50.90 (PR #699 by @franksong2702)
Fixes the /compress reference card showing only a short 3-line summary immediately after compression. Now prefers the persisted compaction message (full handoff) over the raw API summary, matching what is shown after page reload. Closes #695.
2026-04-19 04:29:07 +00:00
nesquena-hermes
b49de92893
feat(/compress): manual session compression with focus topic — closes #469 (PR #619 by @franksong2702)
POST /api/session/compress with optional focus_topic. Transcript-inline cards: command, running, complete (collapsible green), reference. /compact alias kept. Fixes: var(--green) undefined color, focus_topic 500-char cap. Independent review by @nesquena (4 passes).
2026-04-18 06:55:04 +00:00