Commit Graph
1818 Commits
Author SHA1 Message Date
can1357 38b61ae342 fix(session): honored explicit retry-after over reason backoff
- A provider-supplied retry-after now bypasses the transient rate/concurrency
  heuristic window instead of being overridden by it (regression from the
  subscription-cap retry change).
- Updated event-controller/ui-helpers test doubles for provenance-gated
  renderer selection (hasBuiltInTool), aggregated retryErrors on
  auto_retry_end, and Bedrock override compat gaining streamIdleTimeoutMs.
2026-08-07 23:38:25 +02:00
roboompandcan1357 7d4b2e7998 fix(agent): re-check context on cooldown-expiry revert in auto-continue path
A cooldown-expiry model revert runs at a turn boundary. The user-prompt
path reverts then re-checks accumulated context against the restored
model via runPrePromptCompactionIfNeeded, but the automatic
agent.continue() path (#scheduleAgentContinue) reverted and issued the
next request with no such check. When a transient failure had fallen
back to a larger-window model and the conversation then grew past the
original model's window, restoring the primary once its cooldown expired
sent a predictably oversized request to the smaller model.

maybeRestoreRetryFallbackPrimary now reports whether it actually
switched, and the auto-continue path runs the same post-revert
context-fit maintenance (compaction/promotion) the prompt path already
runs, but only when a revert occurred.

Fixes #7952
2026-08-07 23:38:24 +02:00
can1357 0e8142ad0e Merge PR #7904: fix(session): surface real handoff errors instead of false cancel (@roboomp) 2026-08-07 14:52:57 +02:00
roboomp 1e6638d8cd fix(session): surfaced real handoff errors instead of false cancel
The handoff catch in session-handoff.ts and the /handoff handler in
command-controller.ts mapped any error named AbortError to "Handoff
cancelled" regardless of whether the handoff signal was actually
aborted. Providers throw name-AbortError errors on non-user conditions
(stalls, idle timeouts, nested resolution failures), so a genuine
generation failure surfaced as a user cancellation and hid the cause.

Only report "Handoff cancelled" when handoffSignal.aborted is set;
re-throw the real error otherwise. The controller now trusts the
normalized "Handoff cancelled" message and drops its own AbortError
check so re-thrown provider failures render as "Handoff failed: ...".

Fixes #7903
2026-08-07 12:00:06 +00:00
can1357 a8a8188e4b Merge PR #7756: fix(coding-agent): preserve before_agent_start prompt override across base rebuilds (@roboomp) 2026-08-07 13:39:53 +02:00
can1357 fe7774a1aa Merge PR #7806: fix(coding-agent): detect non-English questions in todo reminder guard (@roboomp) 2026-08-07 13:39:52 +02:00
can1357 e8e47a3b34 fix(coding-agent): report only matching prewalk arms 2026-08-07 13:39:51 +02:00
can1357 5f758778b8 Merge PR #7785: fix(coding-agent): make prewalk lifecycle one-shot (@eggpeat) 2026-08-07 13:39:51 +02:00
can1357 68dece477f Merge PR #7772: fix(session): handle subscription-cap retry exhaustion (@roboomp) 2026-08-07 13:37:54 +02:00
Brent 2efe8896ea fix(coding-agent): make prewalk lifecycle one-shot 2026-08-06 19:57:10 +00:00
roboomp 181f63ebc0 fix(coding-agent): detect non-English questions in todo reminder guard
The isAwaitingUserAnswer guard that suppresses the todo completion reminder while the agent waits on a user question only recognized English question words and pronouns. A '?'/'?'-terminated Chinese, Japanese, Korean, or Spanish prompt went undetected, so the <system-reminder> interrupted the pause and the model misread it as the user's answer.

isQuestionPromptLine now also treats a question-mark-terminated line containing any non-ASCII character as a pending user question.

Fixes #7803
2026-08-06 06:47:27 +00:00
roboomp f6c5a43a1f fix(session): handled subscription-cap retry exhaustion
- Classified subscription and plan rate caps as credential-rotatable usage limits while preserving transient per-minute throttles.

- Applied reason-specific backoff to transient rate limits and collapsed exhausted retry attempts behind one budget-labeled terminal error.

- Covered classification, delay selection, persisted transcript aggregation, and retry event propagation.

Fixes #7767
2026-08-06 01:31:22 +00:00
roboomp 3df56506c7 fix(coding-agent): preserve before_agent_start prompt override across base rebuilds
The per-turn systemPrompt returned by before_agent_start was applied only to the agent state, so any base-prompt rebuild firing in the prompt window (context-overflow compaction/promotion, memory promotion, MCP/RPC tool refresh, hindsight MM-TTL refresh) re-pushed the rebuilt base via setSystemPrompt and silently dropped the override before the request.

SessionTools now tracks the active per-turn override and re-applies it on every base rebuild during the turn, clearing it when the turn ends.

Fixes #7755
2026-08-05 21:28:37 +00:00
can1357 0474e797da Merge PR #7678: fix(session): migrate hashed session dirs back to legacy names (@roboomp) 2026-08-05 22:16:28 +02:00
can1357 677b6f10f5 Merge PR #7735: fix(coding-agent): return ToolInfo[] from getAllTools extension API (@roboomp) 2026-08-05 22:15:46 +02:00
can1357 3de6dabf98 Merge PR #7740: fix(session): isolate rewind reports by checkpoint cycle (@roboomp) 2026-08-05 21:50:23 +02:00
can1357 a11eacc9d6 fix(advisor): drop unreachable status guard 2026-08-05 21:50:23 +02:00
can1357 c19665ef90 Merge PR #7745: fix(advisor): fail a refusal over to the model fallback chain (@mvid) 2026-08-05 21:50:23 +02:00
Mantas Vidutis a650938d3f Merge upstream/main into fix/advisor-refusal-fallback 2026-08-05 11:43:16 -07:00
roboomp 2cf500dcc0 fix(session): isolated rewind reports by checkpoint cycle
- Bounded rewind report recovery to messages created after the active checkpoint.

- Added resumed-context regression coverage for late stale rewind results.

Fixes #7739
2026-08-05 16:26:56 +00:00
roboomp 83496b8211 fix(coding-agent): returned ToolInfo[] from getAllTools extension API
The ExtensionAPI getAllTools() wired to session.getAllToolNames(),
returning bare tool-name strings. Upstream @earendil-works/pi-coding-agent
promises ToolInfo[] with sourceInfo, so extensions loaded through the
legacy-pi shim (e.g. gentle-pi) crashed on t.sourceInfo.source at every
session start.

Added SourceInfo/ToolInfo types plus SessionTools.getAllToolInfos(), which
returns { name, description, parameters, sourceInfo } and classifies each
tool as builtin/mcp/sdk/extension. Rewired every getAllTools action site
(interactive, acp, print/rpc, subagent executor) and the example extension.

Fixes #7732
2026-08-05 15:42:06 +00:00
can1357 e9888367d1 refactor: migrated packages to internal utility modules and removed external dependencies
- Implemented in-house, zero-dependency utility modules in `pi-utils` covering DOM manipulation, markdown parsing, templating, browser automation helpers, and terminal buffers.
- Migrated packages across the repository to consume the new internal utilities and `omptype` schema validators instead of external dependencies.
- Removed multiple external runtime and development dependencies including Zod, Marked, LRU cache, Turndown, and Puppeteer browser packages.
2026-08-05 13:39:09 +02:00
Can BölükandGitHub 1e492d6ff9 Merge pull request #7354 from brymko/fix/browser-timeout-crash-recovery
Fix/browser timeout crash recovery
2026-08-05 12:39:14 +02:00
brymko 56931915f1 feat(coding-agent): added fatal session recovery hints
Registered live session resume commands with postmortem handling so a
fatal rejection or exception identifies every recoverable agent before
cleanup. Escaped terminal control characters in recovery output.
2026-08-05 14:31:03 +08:00
roboomp 338451ffdb fix(session): migrate hashed session dirs back to legacy names
The 17.2.9 revert (172ce9b) restored the legacy path-based session
directory names but removed all migration, including the reverse path.
Sessions written under the short-lived hashed scheme (17.2.5-17.2.8,
`<scope>-<readable>-<sha256>`) were left orphaned, so `omp -r`
current-folder scope reported no sessions.

computeDefaultSessionDir now reconstructs the hashed dir name for the
cwd and performs a one-way best-effort migration into the legacy name,
alongside the existing legacy-absolute migration.

Fixes #7677
2026-08-05 06:22:22 +00:00
Kyle McCleary 5cc4f5c93a Merge main into refactor/agent-hub-fullscreen 2026-08-04 18:15:42 -07:00
can1357 eb56330c9c chore(changelog): normalized unreleased entries after merges 2026-08-05 02:32:48 +02:00
roboompandcan1357 99748bbe61 fix(ai): omitted codex optional response defaults
Stopped OpenAI Codex requests from sending reasoning.summary, reasoning.context, and text.verbosity unless explicitly configured, matching the safer native Codex request shape for GPT-5.x models.

Preserved explicit overrides and added regression coverage for transformer and settings-aware stream behavior.

Fixes #4949

(cherry picked from commit 15d86667d6572faf453e74b922eed33dd2d3f3c9)
2026-08-05 02:32:36 +02:00
can1357 a6eafcbf6b fix(ai): keep concurrency caps out of auth rotation
(cherry picked from commit bd6285ad9b16ae6f0a75e336a1c4bf961d3e43c7)
2026-08-05 02:32:36 +02:00
metaphoricsandcan1357 5726fac646 fix(ai): tighten concurrency-cap classification and account-cap gating
Reset-window rotation requires account-specific wording; concurrency caps require an actual cap signal; credential removal gated on AuthFailed without UsageLimit so a valid-but-blocked 403 credential is retained.

(cherry picked from commit 2f72752c2586352a4f7e9e814af1cdb0cf192af4)
2026-08-05 02:32:35 +02:00
metaphoricsandcan1357 7262d9762e fix(ai): honor account reset windows and statusless concurrency caps
Account-reset hint evaluated before short retry hints; account-scoped caps rotate on status 403 or undefined (Devin statusless trailer); statusless concurrency caps marked transient; transient same-model retries use the concurrency backoff.

Refuted: quota-worded concurrency caps were already excluded from rotation before the usage-limit text match.
(cherry picked from commit f2b9a18d715ddbcb6ae703670f2212da36bb2826)
2026-08-05 02:32:35 +02:00
Kyle McCleary 5b68e99796 Merge main into refactor/agent-hub-fullscreen 2026-08-04 17:23:50 -07:00
Kyle McCleary 8e5f619502 fix(coding-agent): harden Agent Hub lifecycle and persistence 2026-08-04 16:29:15 -07:00
can1357 89931e109a fix(session): preserve colliding legacy sessions 2026-08-05 01:12:54 +02:00
can1357 94c838faea Merge PR #7539: fix(coding-agent): complete usage-aware fallback integration (@eggpeat) 2026-08-05 01:12:02 +02:00
can1357 975d0a2e9d Merge PR #7501: fix(agent): classify thinking-only unexpected stops (@roboomp) 2026-08-05 01:12:00 +02:00
can1357 0cf5d55ac0 Merge PR #7629: fix(coding-agent): retry buffered print streams (@roboomp) 2026-08-05 01:12:00 +02:00
roboomp 172ce9bf6c revert(session): restored legacy project directory names
Reverted PR #7397 session directory hashing and automatic migration.

Restored existing script compatibility with the legacy cwd-derived naming scheme.

Fixes #7646
2026-08-04 21:10:54 +00:00
Mantas Vidutis 6e32705e71 fix(advisor): fail a refusal over to the model fallback chain
A classifier refusal returned before the `onTurnError` hook that owns
model fallback, so `AdvisorRuntime` treated one provider's policy verdict
as terminal: `Refusal (cyber)` on the advisor model disabled the advisor
outright even with a fallback chain configured. Its only recovery was
stripping echoed primary reasoning and resending once, which does nothing
for a refusal about the content itself.

Route a refusal that outlives the strip through the same hook the generic
failure path uses, mirroring its epoch guard, session-transition requeue,
and requeue-on-recovery. The cascade walks the chain to exhaustion and
only reports the advisor unavailable once the host runs out of candidates,
matching what turn-recovery already allows for the primary.

Each cascade visits a model at most once. A switch re-arms
`#includeThinking` through `#syncModelIdentity`, so chain keys that point
back at each other (A to B, B to A) would otherwise strip-and-resend
against the same pair forever. A successful turn or a reset starts a
fresh walk.

Also stop `/advisor status` throwing when a live advisor has no roster
entry: `formatAdvisorStatus` guarded only the inactive case before
dereferencing `stats.advisors[0]`, and `#ensureAdvisors` clears
`#advisorStatuses` before repopulating it, so a status call landing in
that window hit `undefined.contextWindow`.
2026-08-04 11:47:54 -07:00
roboomp 929ab1a20f fix(coding-agent): allowed retries for buffered print output
Made text replay safety depend on whether the active output sink has committed streamed text.

Kept tool calls, images, and server tools replay-unsafe while covering text and JSON print policies plus a transient socket-close recovery.

Fixes #7625
2026-08-04 13:50:09 +00:00
Kyle McCleary 91467c2f27 fix(coding-agent): restore Agent Hub lineage metrics 2026-08-03 19:42:11 -07:00
Brent 2da96d77a7 fix(coding-agent): close fallback lifecycle races 2026-08-03 17:51:32 +00:00
Brent 353bbc034b fix(coding-agent): harden usage fallback review races 2026-08-03 17:15:57 +00:00
can1357 1a8caad23e chore: update docs + rename reset to clear 2026-08-03 18:37:23 +02:00
Brent db97103c32 fix(coding-agent): complete usage-aware fallback integration 2026-08-03 16:34:35 +00:00
can1357 fc04aa6fa7 refactor(coding-agent): deferred startup parsing and schema compilation
- Wrapped security contract schemas in a lazy initializer with jitless scopes to eliminate startup JIT compilation tax.
- Enabled jitless scope configuration in auth-broker wire schemas to skip definition-time codegen.
- Deferred startup changelog parsing to overlap with interactive session creation.
2026-08-03 16:31:27 +02:00
can1357 a418920ec1 feat: made /reset semantically different
Closes #4447
2026-08-03 15:46:46 +02:00
roboomp 50228c3656 fix(agent): gate thinking-only stops on signature
Align isUnexpectedStopCandidate with #isEmptyAssistantStop: only a signed (non-whitespace thinkingSignature) thinking-only stop is a candidate. Unsigned thinking-only stops stay with the empty-stop retry path and its cap, so unexpected-stop classification no longer re-handles a turn the empty-stop cap already gave up on.

Fixes #7499
2026-08-03 13:18:54 +00:00
can1357 c53a47f97d Merge PR #7287: fix(coding-agent): prune archived session stats (@alphastorm)
# Conflicts:
#	packages/coding-agent/src/session/session-manager.ts
2026-08-03 15:15:36 +02:00
can1357 267856deb8 Merge PR #7475: fix(session): prevent stale /btw branch promotion (@roboomp) 2026-08-03 15:12:03 +02:00