omp · v18.14.2

@oh-my-pi/pi-agent-core

Added

The soft steering signal a tool sees as toolCall.steeringSignal now aborts with the source of the queued message ("user", "irc", …) as its reason, so a tool can yield only to the sources it should (#17537)

Fixed

Reject incomplete local compaction, turn-prefix, handoff, and branch summaries ending with stopReason: "length" instead of using truncated text as complete context. Omit incomplete display-only short summaries without discarding a completed full summary (#17499 by @Xytronix).

@oh-my-pi/pi-ai Breaking Changes

Replaced isLoopGuardedModel with loopGuardLevel, which returns "off", "exact", or "semantic" from the model's thinking-loop-guard catalog rule

Added

Added disableServerSideFallback to FactoryDroidOptions and SimpleStreamOptions: a Factory Droid request leaves out the server-side refusal fallback chain and fails, before any of its content is streamed, if another model answers it (#12229 by @Xytronix)

Changed

DeepSeek and Grok models on Amazon Bedrock now get the same near-duplicate reasoning-loop checks as on other hosts

Fixed

Fixed Claude turns failing with "Thinking loop detected" when summarized thinking repeated a progress line such as "Still writing the parser..."; Claude models no longer run the loop guard Fixed the loop guard discarding turns from other models whose reasoning summary repeats one status line a few times; whole-line repetition now has to cover about 1 KB before it counts as a loop Fixed per-request OpenAI-compatible headers being combined with differently cased model headers instead of overriding them (#17513 by @jchanghong023) Fixed a stalled Anthropic stream that keeps sending empty content deltas never timing out, because those deltas were treated as stream progress (#17484 by @jchanghong023) Fixed Anthropic-compatible endpoints that put a thinking signature on the streamed content block losing it, so replayed reasoning keeps its signed block (#17483 by @jchanghong023) Fixed AuthStorage.close() and replaceStore() losing a rotated OAuth refresh token when a refresh was still in flight, with or without durable refresh leases; the store now stays open until that token is saved, and new refreshes are refused (#17518) Fixed /usage showing Cursor's "Other Models" meter as an inflated dollar amount; it now shows the percentage Cursor reports, like Cursor Models (#17462 by @eggpeat) Fixed the omp auth-gateway Responses endpoint rejecting documented null values for temperature, top_p, output limits, reasoning, and reasoning summaries (#17466 by @Xytronix). Fixed provider HTTP errors printing the response body twice when the thrown error already carried it, including responses with a retry header (#17503 by @metaphoric-dev) Fixed Gemini tool calls through the omp auth-gateway chat-completions endpoint failing on the next turn: the gateway now returns each call's thought signature in extra_content and replays the one a client sends back (#17509 by @justdoGIT)

@oh-my-pi/pi-catalog Breaking Changes

Removed thinkingLoopGuard from compat records; thinking-loop-guard is now a catalog axis that applies on every host, with "semantic" for Gemini, DeepSeek, and Grok and "off" for Claude

Fixed

Fixed GitHub Copilot Claude 5.x base models (claude-opus-5, claude-opus-5.5, claude-sonnet-5, claude-sonnet-5.5) reporting the 1M window of their -1m variants once the model list was read from cache; discovered limits now survive cached reads instead of reverting to bundled values (#17481)

@oh-my-pi/pi-coding-agent

Added

Added task.perCallModel (off by default, eval, on) so a task item or an eval agent()/workpool() call can run on one exact model the operator already names: a chat role from your own (non-project) settings, the selected agent's own model, or the parent's current model. Glob values (anthropic/*) never approve a model. task.perCallModelAllow narrows that list, task.maxEffort caps the worker's effort, and the worker never switches to another model. Project settings can only narrow this: they never turn it on or approve a model (#12229 by @Xytronix) Added Settings.reloadFromDisk({ includeParents: true }), which reloads the persisted settings behind an overlay while keeping the overlay's own overrides (#12229 by @Xytronix)

Changed

model is now declared in every task schema: a task call, eval agent() or workpool() that passes a model the operator's configuration does not allow is refused with an error (#12229 by @Xytronix) Viewing a subagent or returning to the main session no longer posts a "Viewing agent …" / "Returned to main session" status (a popup in Tern); the status line and Tern's composer already show which agent is in view In a Tern pane, omp stats shows the dashboard in place of the block it runs in, with a Close button (or Esc, or Ctrl+C) that stops it and brings the terminal back, and /stats opens it in a browser block beside omp, instead of the system browser bash tool commands now see PI_TOOL_CALL_ID, the id of the call running them, so a script the agent runs can point at that call in the session transcript (#17489 by @will-bogusz) Native judgments with a configured fallback now move to it after a 3-second candidate deadline (configurable with judgment.attemptTimeoutMs) and briefly skip the timed-out route; without a configured fallback a slow judge is never cut off or replaced (#17465 by @Xytronix).

Fixed

Fixed a ! or $ command run while the agent works floating over the composer in Tern, where expanding its output covered the conversation; it now sits at the end of the conversation Fixed out:// replies sent from an eval cell (tool.write to out://, or tool.bash with > out://) never appearing even though the write reported "Shown to the user."; they now draw as replies after the eval card, and are refused where nothing would show them (a backgrounded cell, a browser tab script) Fixed a write to out:// (such as out://baselines.md) drawing as a Write card with line numbers instead of a reply streaming as prose Fixed table charts drawing a ratio or speedup column as a series beside the timings it was computed from, which read as "719× less" against the baseline Fixed read of a directory hiding existing dist, build, node_modules, .git, and other dependency/build/cache subdirectories, which made them look absent (#17539) Fixed artifact writes reporting the staging-file cleanup error instead of the original short-write or publication failure (#17485 by @jchanghong023) Fixed MCP tool calls failing without reconnecting after DNS or socket-open errors (#17461 by @Xytronix). Fixed MCP connection timeouts and cancellation allowing a stdio server to start after the connection attempt had already failed. (#17512 by @jchanghong023) Fixed ACP updates and extension dialogs switching to an unadvertised session handle after a session reset, and a client reopening a continued conversation after a restart landing on the original session (#17517 by @jchanghong023) Fixed IRC coordination cards remaining visible when Hide Tool Activity is enabled, including live messages and replay (#17498 by @dannyboy-ai). Fixed a message you queue while the agent waits on ask sitting behind the unanswered dialog; the dialog now closes without cancelling the turn, while IRC messages and background completions leave it open (#17537). Fixed JS eval output() ignoring its documented offset, limit, query and format options and returning the raw read tool result; it now returns the selected text (or the format: "json" metadata) like Python's output() (#17455) Fixed session lookup losing readable project sessions when an unrelated project folder is unreadable (#17467 by @Xytronix). Fixed the advisor resending a byte-identical refused review before trying its model fallback when the review contained no primary reasoning to strip (#17473) Fixed read of a whole URL returning no content when the page's first line is over the 50 KB output budget, as on minified pages read with :raw; it now shows that line's first 50 KB (#17500 by @will-bogusz) Fixed an advisor whose usage-limited account kept being reported as switchable retrying the same capped account thousands of times; it now tries each sibling account at most once per provider per failing review, then falls back, waits, or pauses on the quota (#17472) Fixed LSP definition, type definition, implementation and reference results carrying a stray carriage return on every source context line from files with Windows (CRLF) line endings (#17549 by @JustasMasiulis)

@oh-my-pi/pi-mnemopi

Fixed

Fixed heuristic episodic facts storing word fragments (Drafting is active. → ctive), truncated multiword objects, or nothing for has an … sentences (#17511)

@oh-my-pi/pi-natives

Added

Added listWorkspace({ pruneExcludedDirs: false }) to stop pruning standard dependency/build directory names; hidden directories such as .git also require hidden: true (#17539)

@oh-my-pi/omp-stats

Added

Added setStatsCloseHandler: while a host sets one, the dashboard shows a Close button in its top bar and closes on Esc or Ctrl+C (with no text selected), for hosts that show it in place of their terminal

Fixed

Fixed tool error rates counting skipped tool calls that were explicitly marked as not executed (#17463 by @Xytronix).

@oh-my-pi/pi-tui

Added

Added assertModelSupportsEffortCeiling(), which throws when a model has no supported thinking effort at or below a task.maxEffort ceiling (#12229 by @Xytronix) Added iconButton(), quietButton() and segmentedTabs() to native/overlay, and disabled, busy and icon options to actionButton() Added nativeTail to Composer.setRuntimeChildren options: containers whose children continue the native surface's main after the transcript instead of riding in the dock Added isHighlightLanguage and createCodeFrameStyler for renderers that highlight source lines in their file's language (#17550 by @JustasMasiulis)

Changed

Grep and read results with gaps between line runs now render in Tern as one code block that scrolls sideways as a whole, with … gap rows aligned in the line-number gutter Tern builds that draw the button kind get omp's buttons as real buttons (pointer cursor, accessible pressed/disabled state, a disabled button's tooltip says why) and segmented controls as tabs variant: "segmented"; older Tern builds keep the role-styled rows Source lines in Grep, AST Grep and Find results and LSP symbol names are syntax-highlighted in their file's language instead of drawn in one color; consecutive lines highlight together, so multi-line strings and comments keep their colors (#17550 by @JustasMasiulis)

Fixed

Fixed table charts plotting a ratio column against the timings it divides (ratio 0.997 beside C++ (ms) drew as "719× less"): analyzeTable reads a column equal to the quotient of two others as restating them, written bare or as × Fixed IRC coordination cards remaining visible when Hide Tool Activity is enabled, including incoming messages, automatic replies, relays, and workpool messages in live and replayed transcripts (#17498 by @dannyboy-ai). Fixed text after a nested color losing the outer color, such as the rest of the /mcp and /ssh hints after the highlighted command (Run /mcp list to see all configured servers.) falling back to the default text color instead of staying muted (#17521 by @JustasMasiulis) Fixed LSP definition, type definition and implementation results showing the raw Found 1 definition(s): text; the response now counts them (1 definition) and shows each location over its source lines, its own line marked like a grep match (#17549 by @JustasMasiulis) Fixed an LSP definition, type definition or implementation location whose path starts with whitespace being dropped, with its source lines shown under the previous location Fixed session tree user and assistant turns looking identical in themes whose accent and success colors match or nearly match (dark-terminal, dark-poimandres, light-poimandres): user rows are now bold, the assistant: label switches to the path color, and custom and advisor messages are muted (#17520 by @wolfiesch)

@oh-my-pi/pi-utils

Added

Added oklabDistance, the perceptual distance between two hex colors (#17520)

@oh-my-pi/pi-wire

Added

Added nums to code props (TspCodeProps): each line's number, so one block can show non-contiguous lines with … gap rows where the numbers jump. Added the TSP button kind (TspButtonProps: label, icon, keycaps, default/primary/quiet/danger variants, sm/md sizes, pressed, disabled with a reason, busy) and tabs variant: "segmented".

What's Changed

• fix(tui): tell session tree user turns from assistant turns beyond hue by @wolfiesch in #17520
• chore(coding-agent): remove SQLite files committed with the RPC test fixture by @will-bogusz in #17501
• fix(ai): accept nullable Responses request controls by @Xytronix in #17466
• fix(advisor): bounded credential switches on repeated quota failures by @roboomp in #17476
• fix(catalog): kept discovered rows on cached additive-catalog reads by @roboomp in #17482
• fix(ai): stop empty Anthropic deltas from renewing stream timeouts by @jchanghong023 in #17484
• fix(coding-agent): preserve artifact write errors when cleanup fails by @jchanghong023 in #17485
• feat(coding-agent): expose PI_TOOL_CALL_ID to bash commands by @will-bogusz in #17489
• fix(compaction): reject length-truncated summaries and handoffs by @Xytronix in #17499
• fix(coding-agent): show the head of an over-budget first line on whole-URL reads by @will-bogusz in #17500
• fix(ai): print provider error bodies once by @metaphoric-dev in #17503
• fix(mcp): cancel pending stdio creation when connection ends by @jchanghong023 in #17512
• fix(ai): honor case-insensitive per-request header overrides by @jchanghong023 in #17513
• fix(mnemopi): kept whole words and phrases in episodic fact objects by @roboomp in #17516
• fix(acp): keep protocol session handles stable across resets by @jchanghong023 in #17517
• fix(ai): drained in-flight oauth refresh before closing auth store by @roboomp in #17519
• fix(tui): keep a color applied over styled text after the inner styling ends by @JustasMasiulis in #17521
• test(coding-agent): make three tests independent of HOME and the inline image budget's value by @sjawhar in #17527
• fix(ask): interrupted pending dialogs for queued steering by @roboomp in #17541
• fix(coding-agent): listed build and dependency dirs in read directory output by @roboomp in #17543
• fix(lsp): render definition results as locations over their source by @JustasMasiulis in #17549
• feat(tui): highlight search results and LSP symbols in their file's language by @JustasMasiulis in #17550
• fix(coding-agent): honoured js eval output() offset/limit/query/format by @roboomp in #17460
• fix(mcp): retry DNS and socket-open failures by @Xytronix in #17461
• fix(ai): show Cursor's Other Models meter as a percentage by @eggpeat in #17462
• fix(stats): exclude synthetic non-executions from tool errors by @Xytronix in #17463
• fix(judgment): bound stalled native attempts and fail over by @Xytronix in #17465
• fix(coding-agent): isolate unavailable session buckets by @Xytronix in #17467
• fix(advisor): skipped the no-op reasoning strip on refusals by @roboomp in #17475
• fix(ai): keep a thinking signature delivered on content_block_start by @jchanghong023 in #17483
• fix(tui): make IRC cards respect Hide Tool Activity by @dannyboy-ai in #17498
• fix(ai): keep Google thought signature in gateway by @justdoGIT in #17509
• feat(task,eval): per-call model selector limited to operator-configured models by @Xytronix in #12229
• docs: add Simplified Chinese README (README.zh-CN.md) by @bigHippoZz in #17480

New Contributors

@metaphoric-dev made their first contribution in #17503

Full Changelog: v18.14.1...v18.14.2