Repository navigation
[pull] main from can1357:main - #132
Merged
Merged
Conversation
- run arg prompts headless without a stdin TTY; fail bare non-TTY launches with exit 2 - clean --tools usage error listing all built-in tools - reject invalid --thinking/--approval-mode/--mode values - keep explicit non-zero exit codes in the unsettled-command report - suggest closest models for a mistyped --model in print mode
- defer --mode/--thinking/--approval-mode validation until the extension-aware reparse so extension flags that shadow built-ins keep working; report alongside unknown flags (exit 2) and in the ACP session factory - --tools errors separate unknown names from built-ins unavailable in this session instead of listing the catalog as available - process-level test for non-TTY launches (bare exits 2, prompt runs print mode) - give interactive-startup tests a TTY stdin now that non-TTY interactive launches exit 2 - changelog attribution for #13623
…esolution With pending unknown flags or invalid enum values, the pre-extension TTY guard hid the real flag error and exited before an extension owning the flag received it. Skip the early guard then and recheck after the extension-aware flag report.
…parse The bootstrap parse cannot tell an extension string flag's value from a prompt, so 'omp -e ext --spawn-peer reviewer </dev/null' auto-printed with no prompt. When print mode came only from argv and the extension-aware reparse leaves no prompt, fail as a bare non-TTY launch.
An invalid --mode left mode unset, so '--mode bogus --no-ui' exited 1 with '--no-ui requires --mode rpc' before the extension-aware reparse reported the invalid value. Skip the early guard while flag-value errors are pending and recheck it after they are reported.
ACP only validated recorded invalid --mode/--thinking/--approval-mode values in the per-session factory, so 'omp --mode acp --thinking bogus' served and failed every session/new. When a value is pending, load the launch cwd's extensions (only then, keeping ACP startup lazy), reparse, and exit 2 before serving if it is still invalid.
- Integrated web search into model roles, removing the dedicated search provider picker in favor of automatic, free-only chain resolution. - Introduced model-scoped capabilities (`webSearchModel`, `hostedImage`, `imageModel`) to enable native, host-agnostic search and generation via session models. - Renamed legacy image model `openai-codex/gpt-image-1` to `openai-codex/gpt-image-2`. - Restricted keyless search to free providers (Parallel, Exa, Firecrawl) by default, while paid engines remain accessible via explicit model role selection. - Removed setup wizard search configuration, unifying provider and search selection under model role definitions. - Implemented auto-failover for hosted search, allowing chains to proceed if a model fails to return grounded sources.
…sion scope
- Without a terminal on stdin, always start headless and decide after the extension-aware reparse whether any prompt remains; a boolean extension flag shadowing --mode ('--mode compact') now runs 'compact' as the prompt instead of failing on the terminal requirement.
- ACP: reject pending invalid enum values before serving only when extension discovery is off, where every session loads the same explicit extensions; with discovery on, a session cwd's extension may own the flag, so the per-session factory keeps the verdict.
…ime extension binding - --export returns before any extension loads, so a value no extension can own was silently ignored; report it (exit 2) instead of exporting. - ACP no longer binds explicit extensions at launch just to learn their flags (their factories have side effects and ran again per session). Reject pending invalid enum values at launch only when discovery is off and no -e/--hook/trusted extension exists; otherwise the per-session factory decides.
…ocale
Number formatters passed `undefined` as the locale, so figures followed the
browser locale while labels, percents, durations and dates are hard-coded
English. On a tr-TR browser the Overview read `$38.003,33`, `1,4 Mr` and
`546 B` (B = "bin", thousand) beside `97.0%` and `10.7s`, and the existing
`formatEstimatedCost` test ("$1.50") fails on such a machine.
Pin formatInteger/formatCompact/formatCost to en-US and route the three
direct `number.toLocaleString()` calls (BarList, Table, TimelineCanvas)
through formatInteger so every figure uses one convention.
cli ux gap fixes
…less of browser locale (@NaC-L) fix(stats): format dashboard numbers in en-US regardless of browser locale
Hoist the template compile out of composeAppendPrompt and group the test heading constant with the other fixtures.
…-ups (@andrebrait) fix(prompt): address append provenance review follow-ups
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 : )