OpenCodeFormFields reset its fetched model list and invalidated in-flight fetches whenever the headers prop changed identity. MiniMax Code / DeepSeek Harness forms pass a fresh {} on every render, so the list vanished within ~2s while the local route was running. Compare headers by serialized content instead, matching KimiProviderForm's headersKey.
OpenCode, MiniMax Code and DeepSeek Harness share one model editor that
still showed a "Token Limits" header and dumped reasoning and modalities as
raw JSON properties. Lay it out like the Pi and Kimi model rows: capability
switches for extended thinking and image input, then context length and max
output fields with the same labels.
DeepSeek Harness also gets structured thinking levels (offered levels and
their upstream values) and per-model interface compatibility options. The
thinking switch follows each app's semantics: OpenCode drops `reasoning`
instead of writing false so models.dev defaults still apply, while DSH
writes `reasoningEfforts: false`, which is the only field it reads.
Fields with an unrecognised shape stay under Model Properties so they can
still be edited by hand.
Each model in the picker dialog now shows its context window, reasoning
and image input, resolved with the same lookup as the parameter fill
(same-endpoint preset, then models.dev by endpoint, then the vendor's own
entry), so what the dialog shows is what Apply fills in. Prices are left
out: models.dev only has them when the endpoint is recognized. The vendor
column has a fixed width so the parameters line up. The dialog body mounts
only while open, which is also when it subscribes to the models.dev query.
Since the dialog already showed the values, models added from it are
filled without the "parameters filled, please check" toast. The fill
function of useModelMetadataFill now takes { baseUrl, silent } instead of a
bare base URL (the Pi form passes its per-model address that way). Picking
a model from a row's dropdown still shows the toast.
ClaudeStackModelsField takes the provider's base URL so its dialog can
show parameters too; its rows have no such fields to fill.
The fetched-model picker was an inline box under the model list that could
only add models: configured ones showed as checked and disabled. It is now
a dialog opened from the list header (Claude Code Aggregation, both Codex
model list layouts, OpenCode and MiniMax Code). A successful fetch opens it;
after that the button reads "Select models" and opens it without fetching
again, with a refresh button inside.
A checkbox now means "in the list": configured models start checked and
unchecking removes their rows on Apply, while Cancel changes nothing. Rows
that stay keep their order and fields, new picks are appended and get their
parameters filled. Configured models the provider did not return are listed
first in their own group so they can be removed too. In Codex, removing the
default model's row moves the default to the new first row, as deleting
the row does. The dialog has search, select all / none for the visible
models, a selected-only filter and a selected count.
Replaces fetchedModelsTitle, modelAlreadyAdded and addSelectedModels with
providerForm.modelPicker.*.
The extra-options row editor keeps every value as display text, and any
row change re-parsed all rows with JSON.parse. Untouched string values
such as "123", "false" or "null" were silently turned into numbers,
booleans or null on save, in both the OpenCode and MiniMax Code forms.
Merge row edits into the stored options instead: a row whose text still
matches the stored value's display form keeps the original value and
type, and a renamed key inherits its old value. Renaming an option onto
a key that already exists is rejected, matching the model field editor.
Fixes#7833
Pages (phase 4):
- Usage: v7 layout (filter row, four metrics with more on demand,
switchable trend, request log / provider / model / pricing tabs, data
sources sheet). Latency becomes output speed: output / (duration -
first token), shown only for requests with at least 100 output tokens;
provider totals sum output and generation time instead of averaging.
- Sessions: single-column list grouped by project or time, drilling into
a reader with resume, find, question jump and copy actions.
- MCP and Skills share one app matrix (neutral checks, retry on failed
writes, column bulk actions with undo); MCP add/edit moves into a
drawer with form/JSON modes; Skills merges installed and discover and
always shows a Pi column.
- Prompts: one page for eight apps including Hermes SOUL.md, enable and
disable as buttons, delete with undo instead of a confirm, copy to
other apps, import existing files; app pages show the active prompt.
- Add provider is two steps: pick a preset (always-on search including
Chinese names and domains, categories inferred for display, sorted by
name with no promoted entries), then fill the form. Edit pages name
the provider and warn when it is the one in use.
- App page menu links to that app's usage.
Visual pass (phase 6):
- Secondary pages cover only the content area so the sidebar stays.
- Cmd/Ctrl+K jumps to apps, pages and settings sections.
- Legacy button variants, primary color, dialogs, toasts and sheets
follow v7 tokens; hard-coded palette classes and 10px text across
the remaining components are replaced with tokens.
A provider's /models endpoint only returns model IDs, so a model picked
from the fetched list left its context window, output limit, reasoning
levels, and modalities empty, while the same model in a preset had them.
When a model is picked in the Codex, OpenClaw, Hermes, OpenCode, or
MiniMax Code form, look up its metadata field by field:
1. the preset model with the same endpoint (hand-verified values);
2. the models.dev provider whose API endpoint matches;
3. the vendor's own models.dev entry, without its price.
Endpoints match only when host and path are equal after dropping a
trailing version segment or request path, so plans that share a host
(z.ai standard vs Coding Plan, OpenCode Zen vs Go) never borrow each
other's values. Models are never matched by name across providers: the
same model ID often has very different context windows per deployment.
Only empty fields are filled, and reasoning or image input is only ever
turned on. models.dev is prefetched once the model list is fetched and
shared with the usage pricing import; when it arrives late the row is
filled again, located by its stable row key. A single toast asks the
user to review what was filled.
Follow up on the maintainer review for #6333 by protecting shared provider metadata fields and hardening interrupted-composition blur handling.
Commit OpenCode key edits from the blur event value so force-committed IME text is not lost to stale local state.