Repository navigation
[pull] main from can1357:main - #112
Merged
Merged
Conversation
…nking levels, document dialect
…oad docs, no formatter churn
…ounding - Queued background completions (finished jobs, exited supervised processes) no longer raise the cooperative steering signal, so ordinary foreground bash/eval calls keep running instead of detaching milliseconds in; the freshly detached job's own completion used to detach the next command, chaining short commands into separate background deliveries. - Completions still hard-abort interruptible waits (hub wait) and inject at the batch boundary, unchanged. - The event-driven steering watch now stops only once the cooperative signal is up, so a genuine user steer arriving after a completion-only trigger still reaches foreground tools. Fixes #12869
- Parsed the console's monthly bucket into a `credits:monthly` limit so a monthly-only plan still produces a usage report. - Promoted the monthly window class for every single-bucket subscription provider, adding alibaba-token-plan next to cursor and opencode-go. - Covered both layers with regression tests and documented the new window. Fixes #12873
- Recorded server startup time for diagnostic freshness decisions. - Kept empty unversioned cold-server publishes pending for late delivery. - Covered the first-edit placeholder publish followed by completed analysis. Fixes #12889
The minimizer defs moved to crates/pi-shell, and pi-shell's build.rs generates builtin_filters.toml there. pi-natives kept a copy that watched crates/pi-natives/src/shell/minimizer/defs. That path no longer exists, and cargo treats a missing rerun-if-changed path as changed, so pi-natives recompiled on every build (about 60s on an M-series Mac). Nothing in pi-natives reads the file it generated. Measured: two consecutive build:native runs took 64s then 64s before, 200s (cold after the edit) then 3s after.
…version - `isJTDSchema` treated any root with `properties` and no `type` as JTD, so a native JSON Schema root was rebuilt by `convertSchema`, which models JTD keywords only and dropped `items`/`required`; OpenAI strict mode then rejected the yield tool with `array schema missing items`. - Added `isJTDDocument`, an RFC 8927 grammar check (known keywords only, JTD primitive `type` values, declared scalar value types, recursion through `elements`/`values`/`properties`/`optionalProperties`/`definitions`/`mapping`), and gated the ambiguous properties-form rule on it. - Genuine JTD properties documents still convert with `required` and `additionalProperties: false`; documents carrying JSON-Schema-only keywords now stay on the `normalizeJsonSchemaNode` passthrough path. Fixes #12893
The OTLP exporters are constructed without explicit config, so @opentelemetry/otlp-exporter-base already resolves request headers from OTEL_EXPORTER_OTLP_HEADERS and the per-signal *_HEADERS variants; only the env reference omitted them. Extends otel-export-probe.ts to assert the env-configured headers reach the collector (common list merged, signal-specific key winning, values percent-decoded). Fixes #12894
- `tab.fill`, `frame.click` and `frame.fill` waited on `Locator` preconditions (viewport intersection, stable bounding box, enabled probe) that only settle while the tab produces animation frames. Managed headless tabs are backgrounded targets that produce none, so every call burned the full 8s action ceiling instead of acting. - Routed all three through the handle paths the element API already uses (`#resolveActionHandle` + `fillViaHandle`, `clickElement`), so no browser action depends on frame production. - Moved `fillViaHandle` next to `clickElement` in `interactions.ts` so frame-scoped actions can share it without a cycle. - Added a regression test that halts frame production via paused virtual time and then fills a page selector and an iframe selector. Fixes #12892
- Matched the Copilot Responses roster by `grok-4.` prefix in `api-routes` so every 4.x revision and its synthesized `-1m` sibling routes to /responses instead of one exact id per release. - Added grok-4.7/grok-4.7-1m to the Copilot cache-invalidation list so a cache written before the pin is refetched. - Scoped static compat overrides to the transport they were authored for: a bundled chat-completions row no longer strips the effort dial once discovery re-routes the id to Responses. Fixes #12901
…ustom providers (@LingLambda)
…k for extensions (@aloyzzz)
…l discovery (@ArpitMishra17) # Conflicts: # packages/catalog/src/compat/rules.json
… get_tree, get_available_thinking_levels to RPC protocol (@44madfire)
agent-identity: dev3.direct.omp.hppn5dm9 agent-persona: generalist agent-supervisor: unavailable agent-tool: OMP agent-tool-version: 18.1.14 agent-runtime: OMP 18.1.14 agent-model: openai-codex/gpt-5.6-sol tooling-profile: dotfiles@183d780
agent-identity: dev3.direct.omp.hppn5dm9 agent-persona: generalist agent-supervisor: unavailable agent-tool: OMP agent-tool-version: 18.1.14 agent-runtime: OMP 18.1.14 agent-model: openai-codex/gpt-5.6-sol tooling-profile: dotfiles@183d780
agent-identity: dev3.direct.omp.hppn5dm9 agent-persona: generalist agent-supervisor: unavailable agent-tool: OMP agent-tool-version: 18.1.14 agent-runtime: OMP 18.1.14 agent-model: openai-codex/gpt-5.6-sol tooling-profile: dotfiles@183d780
agent-identity: dev3.direct.omp.hppn5dm9 agent-persona: generalist agent-supervisor: unavailable agent-tool: OMP agent-tool-version: 18.1.14 agent-runtime: OMP 18.1.14 agent-model: openai-codex/gpt-5.6-sol tooling-profile: dotfiles@183d780
agent-identity: dev3.direct.omp.qa2w9jj3 agent-persona: generalist agent-supervisor: unavailable agent-tool: OMP agent-tool-version: 18.2.0 agent-runtime: OMP 18.2.0 agent-model: openai-codex/gpt-5.6-sol tooling-profile: dotfiles@3c1a4a0-dirty
- SDK discoverAuthStorage now derives auth.accountPolicies and retry.usageReservePct from effective Settings (explicit instance, matching global instance, or a read-only load), so omp commit, compress, cleanse, models, read, setup, bench and other standalone entrypoints honor PI_CONFIG_FILES, --config and project overlays. - Callers holding Settings pass it and sequence Settings.init before discovery instead of racing them. - main.ts, usage-cli and auth-gateway share the same loader instead of per-callsite plumbing. - Replaced the option-forwarding test with a PI_CONFIG_FILES routing regression.
- Print mode now blocks the first prompt until every configured MCP server loads tools or fails, bounded by OMP_MCP_TIMEOUT_MS, then refreshes the session tool list; interactive, RPC and ACP startup stay non-blocking. - Added MCPManager.waitForStartup() reporting connected, pending and failed servers, including timeout-triggered reconnects. - Made the initial discovery window configurable via mcp.startupTimeoutMs and OMP_MCP_STARTUP_TIMEOUT_MS (default 250ms). - Unavailable servers are named on stderr; OMP_MCP_REQUIRE_READY=1 exits 1 before the turn. Fixes #12188
- Strict config parsing threw "must contain a YAML object" for an empty or comment-only config.yml, which main treated as empty; auth discovery then failed at startup for users with nothing configured.
- Session resume restored account affinity through the explicit user-pin API, so an Anthropic session resumed past the warm window stuck to its old account instead of re-ranking, even without account policies. - pinSessionOAuthAccount now takes restoredAtMs for automatic restores; user pins stay explicit.
…for failed judgments - Filtered judge role chain to keep only native candidates after the first native, preventing prompted models from serving as fallback. - Added logger.warn on judgment candidate failure, exposing candidate identity, status, error, and rejection cooldown. - Extended usageReportingTypeSafeJudge to emit usage events for failed native judgments with stopReason "error" and errorMessage. - Inserted "openrouter/~typesafe/jev-latest" into priority.json as a native judge candidate. - Adjusted ChainJudge candidate building to skip session model injection when any native candidate exists.
…ffinity - Modularized `AuthStorage` into distinct sub-namespaces (`credentials`, `keys`, `oauth`, `limits`, `health`, `blocks`, `resets`, `usage`) to improve API clarity and organization. - Implemented `SessionAffinity` and `CredentialPool` to support persistent, sticky session-to-credential pins and improved load-balancing. - Added persistent rate-limit block tracking and automated block-healing logic with support for account-specific routing policies. - Introduced `KeyCascade` to unify hierarchical authentication precedence across runtime overrides, config, and stored credentials. - Replaced `node:crypto` hashing utilities with optimized `Bun.SHA256` and `Bun.SHA1` implementations throughout the codebase.
- Introduced `omp login [provider]` CLI command for terminal-based OAuth authentication. - Unified terminal OAuth flow handling into a shared module for `login` and `auth-broker login`. - Integrated automatic model discovery refresh upon successful terminal login. - Added support for auto-opening the authentication URL in the default browser during terminal login. - Updated successful login feedback to include identity details (email/organization).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )