Commit Graph

10 Commits

Author SHA1 Message Date
Tianyi Cui
226dc7a249 docs: translate remaining READMEs 2026-07-26 05:06:39 +08:00
Tianyi Cui
6b44d01bc8 Merge master into compact-tool-result-prune 2026-07-20 16:41:59 +08:00
Tianyi Cui
e8eddc7ef8 Rename RFCs to Agent Notes 2026-07-19 22:52:03 +08:00
Hypatia May
ba4a57c7d1 refactor(compact): align pruning contracts (round 3) 2026-07-16 18:51:26 +08:00
Hypatia May
ce96104a77 feat(compact): prune tool results before summarization (round 1) 2026-07-16 18:02:15 +08:00
Hypatia May
f038780ff6 feat(llm): add replay token metering (PR2 round 1) 2026-07-15 14:47:29 +08:00
Tianyi Cui
cc15ef54ef Sync remaining chars-per-token prose in overview READMEs
The convergence pass found four summary-level sites still describing
the estimator as fixed char/4: packages/README.md (twice), the compact
group and interface READMEs, and compact-basic's package.json
description. All now say chars-per-token with the charsPerToken
default, matching the authoritative package README/module doc/RFC.
2026-07-04 18:15:54 +08:00
Hypatia May
cec32faa4e refactor(compact): turn-agnostic retention + dedicated agent/pre-request seam
Reform the compaction blueprint so a runaway turn survives and the design
stops drifting across review rounds:

- Drop in-flight-turn protection ("layer 2"). Retention is a uniform tail→head
  whole-unit walk; the only structural guard is step-alignment. A single turn
  that alone exceeds the window now compacts its own early closed steps instead
  of being retained verbatim (the failure mode that motivated this).
- Move auto-compaction off the agent/request waterfall onto a new awaited
  agent/pre-request loop seam, fired before history derivation. Compaction
  mutates the surface; the loop derives once from the result — no double-derive,
  and a listener structurally cannot act on not-yet-derived messages.
- Tighten compactIfNeeded to required (session, system, model, signal).
- Enforce a single-pass convergence invariant in resolveConfig: reject configs
  where summarizationMaxTokens + retainTokens exceeds the threshold, so a
  compaction can never immediately re-trigger.
- Document the crash vs recoverable failure taxonomy; core session repair stays
  compaction-agnostic (a log-only orphaned compact/start is inert).
- Wire dsh-compact-basic into examples/coding-agent and add a with-key
  compaction e2e (compaction's first real-world exercise + runaway net).
- Rewrite the RFC to encode the blueprint and move it to implemented/.

The runaway-turn snapshot is a named deferred follow-up: dsh-llm-replay cannot
yet serve the interleaved summarization model call.
2026-06-26 08:59:33 +08:00
Hypatia May
aa9afcefc7 feat(compact-basic): baseline compaction backend (squashed from compact-basic)
Collapses the per-round review churn of the prior compact-basic branch into a
single clean baseline on top of compact-interface, so the upcoming retention
refactor lands as fresh, well-scoped commits rather than stacking on a history
of fixes that are being superseded.
2026-06-25 17:30:42 +08:00
Hypatia May
894653763e docs(compact): add the compact group/service to package and architecture docs 2026-06-23 16:40:24 +08:00