New Models/MutableModels/createModels collection: provider map, async
model listing (best-effort aggregation), getAuth decision tree with
double-checked locked OAuth refresh, stream/complete with per-field
auth merge over lazyStream.
Auth substrate: ProviderAuth { apiKey?, oauth? }, one type-tagged
credential per provider, CredentialStore (read/modify/delete; modify
is the only write path, serialized RMW), OAuthAuth login/refresh/toAuth
split, prompt()/notify() login callbacks, browser-safe default
AuthContext.
types.ts: Provider alias renamed to ProviderId; ApiOptionsMap and
ApiStreamOptions<TApi> for typed per-API stream options; hasApi()
runtime narrowing guard.
moveCursor() repositioned the cursor but never called updateAutocomplete()
or cancelAutocomplete(), unlike insertCharacter()/handleBackspace(). So an
open autocomplete picker went stale when the cursor moved: e.g. typing
`/cmd ` (argument menu open) then arrowing Left back into the command name
left the argument list showing against a `/cmd` prefix, and a Tab there
concatenated the stale suggestion onto the partial command name
(e.g. `/swarepo`).
Sync the picker at the end of moveCursor(): when a picker is open, re-query
via updateAutocomplete(), which refreshes the list for the new cursor
position or closes it when nothing matches.
Adds a regression test in test/editor.test.ts (verified to fail without the
fix). Fixesearendil-works/pi#5496.
- Wrap package.json read in try/catch to handle missing file gracefully
- Use defaults (APP_NAME=pi, CONFIG_DIR_NAME=.pi, etc.) when package.json not found
- Preserves rebranding via piConfig when package.json exists
- Only ignores ENOENT; other errors still throw
- Update GitHub Copilot tests to use available models (claude-haiku-4.5, gemini-2.5-pro)
- Update OpenRouter tests to use available models (google/gemini-2.5-flash, deepseek/deepseek-chat)
- Fix over-broad sed that incorrectly renamed OpenAI Responses gpt-4o test
- Remove nvidia.md from repo root
- Update packages/ai/CHANGELOG.md
The HTTP timeout setting (httpIdleTimeoutMs) was only used as a fallback for the
openai-codex-responses API. For other providers like openai-completions
(llama.cpp), the SDK default timeout (10 min) was used instead, ignoring the
user's disabled timeout setting.
Now httpIdleTimeoutMs applies universally as the default SDK request timeout
for all providers that support timeoutMs. Setting HTTP timeout = false (0)
correctly disables SDK timeouts across the board by sending max int32
(effectively infinite) instead of 0, since SDKs treat timeout=0 as immediate
timeout.
closes#5294
Mario Zechner