mirror of
https://github.com/deepseek-ai/deepseek-harness
synced 2026-08-15 21:04:50 +00:00
fix(llm): honor model metadata cancellation and defaults
This commit is contained in:
Yichen Jiang
@@ -2,5 +2,5 @@
|
||||
# side as of the last confirmed-consistent state. Both languages carry equal authority;
|
||||
# after editing either side, bring the other along and re-record with:
|
||||
# pnpm run verify-translation-pairing --write
|
||||
2026-07-17-dedicated-full-screen-tui-front-door.md: 3b018de9d2360e68ad00cdb7e4d5178c9e3bc6ea
|
||||
2026-07-17-dedicated-full-screen-tui-front-door.zh.md: 46165ce41bb0704b64d4447e683b70839e188562
|
||||
2026-07-17-dedicated-full-screen-tui-front-door.md: 8d7c7b00c8d9b15ea3f2419ed44ca88209e60dad
|
||||
2026-07-17-dedicated-full-screen-tui-front-door.zh.md: 49e9541c9f98c9f3beba11945ff452fc38bd9ede
|
||||
|
||||
@@ -24,7 +24,7 @@ The TUI rebuilds the transcript from the active `session.surface` and reprojects
|
||||
|
||||
Editor input calls `agent.send()` while idle and `agent.steer()` while a turn is running. Cancellation, reasoning visibility, tool-card expansion, redraw, transcript clearing, and exit are terminal-only controls. The idle footer derives context occupancy from `tokenMeter` and shows the selected model and explicit reasoning effort; during a run, elapsed activity and the Escape interrupt hint replace that summary. `/status` remains available in either state and appends a detailed terminal-only snapshot: session identity and timestamps, selected model, reasoning effort/default state and reasoning visibility, lifecycle counts folded from the event log, the same deduplicated usage buckets and KV-cache rate as the footer, and context use from `tokenMeter` plus the selected model's advertised capacity. The plugin registers the shared `userInteraction` provider and presents queued questions in a wide bottom-left keyboard panel with batch progress, numbered options, and aligned descriptions; agent behavior and answer logging remain owned by their existing services.
|
||||
|
||||
The `/model` command presents the advisory `ctx.llm` catalog as a keyboard selector and changes only this TUI session's target; argument forms remain available for direct selection. Each model row owns the adapter-advertised reasoning-effort order and default: Shift+Tab cycles that row's efforts, while a model without selectable metadata keeps default behavior. Agent-scoped prompt-assembly and request waterfalls snapshot one provider/model/reasoning-effort target per step, so `{{provider}}` / `{{model}}` interpolation and request routing cannot split when a command arrives during assembly. The latest logged request header restores a used target; a selection that never reaches a request remains process-local.
|
||||
The `/model` command presents the advisory `ctx.llm` catalog as a keyboard selector and changes only this TUI session's target; argument forms remain available for direct selection. Each model row owns the adapter-advertised reasoning-effort order and default: Shift+Tab cycles that row's efforts, includes provider-default behavior when the adapter advertises no default, and leaves models without selectable metadata unchanged. Agent-scoped prompt-assembly and request waterfalls snapshot one provider/model/reasoning-effort target per step, so `{{provider}}` / `{{model}}` interpolation and request routing cannot split when a command arrives during assembly. The latest logged request header restores a used target; a selection that never reaches a request remains process-local.
|
||||
|
||||
### Terminal ownership
|
||||
|
||||
|
||||
@@ -24,7 +24,7 @@ TUI 从活跃的 `session.surface` 重建 transcript（文本记录），并在
|
||||
|
||||
agent 空闲时，编辑器输入调用 `agent.send()`；轮次运行中则调用 `agent.steer()`。取消、推理显隐、工具卡片展开、重绘、清空 transcript 和退出都只是终端控制。空闲态页脚根据 `tokenMeter` 得出上下文占用率，并显示所选模型和显式选定的推理强度；agent 运行期间，该摘要会替换为带已用时长的活动指示和 Escape 中断提示。`/status` 在这两种状态下均可用，并会追加一份仅在终端显示的详细快照，其中包括会话标识与时间戳、所选模型、推理强度（或默认状态）及推理显隐状态、从事件日志归并得出的生命周期计数、与页脚一致的去重用量分项和 KV 缓存命中率，以及 `tokenMeter` 给出的上下文用量和所选模型公布的容量。插件注册共享的 `userInteraction` 提供方，在左下角宽幅键盘操作面板中呈现排队的问题，面板显示批次进度、带编号的选项和对齐的描述；agent 行为和答案日志仍由既有服务负责。
|
||||
|
||||
`/model` 命令将建议性的 `ctx.llm` 目录呈现为键盘选择器，并且只更改当前 TUI 会话的目标；带参数的形式仍可直接选择目标。每个模型行都持有适配器公布的推理强度顺序和默认值：按 Shift+Tab 可循环切换该行的推理强度；没有可选元数据的模型保持默认行为。agent 作用域内的 prompt 组装和请求两条 waterfall（瀑布式事件）会为每个步骤快照一次同一个提供方/模型/推理强度目标，因此即使命令在组装期间到达，`{{provider}}` / `{{model}}` 插值与请求路由也不会分裂。系统通过日志中最新的请求头恢复已经使用过的目标；未被请求使用的选择只保留在当前进程中。
|
||||
`/model` 命令将建议性的 `ctx.llm` 目录呈现为键盘选择器，并且只更改当前 TUI 会话的目标；带参数的形式仍可直接选择目标。每个模型行都持有适配器公布的推理强度顺序和默认值：按 Shift+Tab 可循环切换该行的推理强度；如果适配器没有公布默认值，循环中还会包含提供方默认行为；没有可选元数据的模型则保持不变。agent 作用域内的 prompt 组装和请求两条 waterfall（瀑布式事件）会为每个步骤快照一次同一个提供方/模型/推理强度目标，因此即使命令在组装期间到达，`{{provider}}` / `{{model}}` 插值与请求路由也不会分裂。系统通过日志中最新的请求头恢复已经使用过的目标；未被请求使用的选择只保留在当前进程中。
|
||||
|
||||
### 终端所有权
|
||||
|
||||
|
||||
@@ -272,7 +272,7 @@ export class BasicCompactService extends CompactService {
|
||||
return this.compactRegion(range.start, range.end, agent, signal)
|
||||
}
|
||||
|
||||
const context = (await this.ctx.llm.resolveModelInfo(target.provider, target.model)).context
|
||||
const context = (await this.ctx.llm.resolveModelInfo(target.provider, target.model, signal)).context
|
||||
const targetKey = `${target.provider}/${target.model}`
|
||||
if (context === undefined) {
|
||||
throw new TargetPressureConfigError(
|
||||
|
||||
@@ -463,6 +463,18 @@ describe('pressure measurement and retention', () => {
|
||||
.resolves.not.toBeNull()
|
||||
})
|
||||
|
||||
it('forwards turn cancellation to proactive model metadata resolution', async () => {
|
||||
const ctx = createContext()
|
||||
const resolveModelInfo = vi.spyOn(ctx.llm, 'resolveModelInfo')
|
||||
const compact = service(compactConfig, ctx)
|
||||
const session = conversation()
|
||||
const signal = new AbortController().signal
|
||||
|
||||
await expect(compact.compactIfNeeded(agent(session, MODEL), 'pressure', signal))
|
||||
.resolves.not.toBeNull()
|
||||
expect(resolveModelInfo).toHaveBeenCalledWith(MODEL, MODEL, signal)
|
||||
})
|
||||
|
||||
it('re-resolves capacity after a same-model-id provider switch in one session', async () => {
|
||||
const ctx = new Context()
|
||||
void new LlmService(ctx)
|
||||
|
||||
@@ -22,7 +22,7 @@ When optional `ctx.sessionReferences` is mounted, the same `@` menu also offers
|
||||
|
||||
While the agent is running, ordinary editor submissions call `agent.steer()`; otherwise they call `agent.followup()`. A slash at the start of the submitted line enters `ctx.commands` instead: known commands execute directly, unknown commands produce a warning, and neither path automatically reaches the model. A command producer may explicitly schedule agent work; [`dsh-plan-mode`](../../plan/plan-mode/README.md#model-and-human-surfaces) uses that contract for `/plan [message]`. The TUI registers `/help`, `/model`, `/clear`, `/reasoning`, `/tools`, `/redraw`, `/reload`, `/resume`, `/status`, and `/exit` as agent-scoped definitions; every other effective command joins autocomplete and `/help` dynamically, as do `/skill:` completions. A status line above the editor reports the turn phase the TUI derives from session events — waiting for the first token, thinking, responding, or executing tools — with the elapsed time in that phase and the running step total, refreshed each second, and ends with the `Enter sends steering, Esc cancels` hint; while steering messages wait to reach the model it inserts a `N queued ·` badge before the hint that clears as each drains. Ctrl+C or Escape cancels a running turn. Tool cards collapse long bodies into a configurable head/tail preview; Ctrl+O toggles every card between its preview and full output. Ctrl+R toggles reasoning, Ctrl+L redraws, and Ctrl+D exits while idle.
|
||||
|
||||
`/model` opens the advisory `ctx.llm` catalog as a keyboard selector: Up/Down moves, Shift+Tab cycles the focused model's adapter-advertised reasoning efforts in display order, Enter selects the model and effort, and Escape closes it. Models without selectable effort metadata ignore Shift+Tab; the selector renders the exact advertised list—including `off` when present—and does not synthesize, clamp, or transfer an effort between models. `/model <model>` still selects an unambiguous model id directly, while `/model <provider>/<model>` selects an exact target and uses its adapter default when one exists. The configured target or latest logged request header initializes the selector, and an unlisted current model remains visible because catalogs are advisory. Selection is local to this TUI session. Prompt assembly snapshots the target for one step, replaces `{{provider}}` and `{{model}}`, and applies the same provider/model/reasoning-effort target through `agent/request`; a switch during assembly therefore starts with a later step. The request header durably records targets that reach the model, while an unused selection remains process-local.
|
||||
`/model` opens the advisory `ctx.llm` catalog as a keyboard selector: Up/Down moves, Shift+Tab cycles the focused model's adapter-advertised reasoning efforts in display order, Enter selects the model and effort, and Escape closes it. When an adapter does not advertise a default effort, the cycle also includes `provider default`, which clears an explicit selection; models without selectable effort metadata ignore Shift+Tab. The selector renders the exact advertised effort list—including `off` when present—and does not synthesize, clamp, or transfer an effort between models. `/model <model>` still selects an unambiguous model id directly, while `/model <provider>/<model>` selects an exact target and uses its adapter default when one exists. The configured target or latest logged request header initializes the selector, and an unlisted current model remains visible because catalogs are advisory. Selection is local to this TUI session. Prompt assembly snapshots the target for one step, replaces `{{provider}}` and `{{model}}`, and applies the same provider/model/reasoning-effort target through `agent/request`; a switch during assembly therefore starts with a later step. The request header durably records targets that reach the model, while an unused selection remains process-local.
|
||||
|
||||
`/reload` (EXPERIMENTAL, dev-only) re-reads every file-backed loader config tree and applies the diff to the running app — the HMR watcher's config path, invoked manually; it needs the cordis Loader in the context and degrades to a warning without one, runs only while the agent is idle, and refuses re-entry while a reload is in flight. Module-source hot reload remains watcher-owned. When a `skills` service is mounted, `/skill:<name> [instructions]` loads that skill's instructions into the conversation as a user turn; autocomplete lists the model-invocable skills, and any skill (including a model-disabled one) is loadable by its exact name.
|
||||
|
||||
|
||||
@@ -586,6 +586,11 @@ interface ModelChoice extends AgentLlmTarget {
|
||||
reasoning?: LlmModelReasoningInfo
|
||||
}
|
||||
|
||||
interface ModelDialogSelection {
|
||||
choice: ModelChoice
|
||||
reasoningEffort: ReasoningEffortId | undefined
|
||||
}
|
||||
|
||||
function targetLabel(target: AgentLlmTarget): string {
|
||||
return `${target.provider}/${target.model}`
|
||||
}
|
||||
@@ -1262,7 +1267,7 @@ class ModelDialog implements Component {
|
||||
current: AgentLlmTarget | undefined,
|
||||
maxVisible: number,
|
||||
private readonly palette: Palette,
|
||||
done: (choice: ModelChoice) => void,
|
||||
done: (selection: ModelDialogSelection) => void,
|
||||
cancel: () => void,
|
||||
) {
|
||||
this.items = new Map()
|
||||
@@ -1294,10 +1299,9 @@ class ModelDialog implements Component {
|
||||
const selected = choices.find(choice => targetLabel(choice) === item.value)
|
||||
/* v8 ignore next -- SelectList only returns values built from `choices`. */
|
||||
if (selected === undefined) return
|
||||
const effort = this.efforts.get(item.value)
|
||||
done({
|
||||
...selected,
|
||||
...effort === undefined ? {} : { reasoningEffort: effort },
|
||||
choice: selected,
|
||||
reasoningEffort: this.efforts.get(item.value),
|
||||
})
|
||||
}
|
||||
this.list.onCancel = cancel
|
||||
@@ -1324,11 +1328,13 @@ class ModelDialog implements Component {
|
||||
const choice = this.choices.get(selectedItem.value)
|
||||
if (choice?.reasoning === undefined) return
|
||||
const current = this.efforts.get(selectedItem.value)
|
||||
const currentIndex = choice.reasoning.efforts.findIndex(effort => effort.id === current)
|
||||
const next = choice.reasoning.efforts[(currentIndex + 1) % choice.reasoning.efforts.length]
|
||||
/* v8 ignore next -- validated reasoning metadata always carries at least one effort. */
|
||||
if (next === undefined) return
|
||||
this.efforts.set(selectedItem.value, next.id)
|
||||
const efforts: Array<ReasoningEffortId | undefined> = [
|
||||
...choice.reasoning.defaultEffort === undefined ? [undefined] : [],
|
||||
...choice.reasoning.efforts.map(effort => effort.id),
|
||||
]
|
||||
const currentIndex = efforts.indexOf(current)
|
||||
const next = efforts[(currentIndex + 1) % efforts.length]
|
||||
this.efforts.set(selectedItem.value, next)
|
||||
const item = this.items.get(selectedItem.value)
|
||||
/* v8 ignore next -- items and choices are constructed from the same values. */
|
||||
if (item === undefined) return
|
||||
@@ -2118,10 +2124,14 @@ export function createTuiChat(
|
||||
}
|
||||
resolveContextWindow(target.current)
|
||||
|
||||
const selectModel = (selected: ModelChoice): void => {
|
||||
const selectModel = (
|
||||
selected: ModelChoice,
|
||||
explicitReasoning?: { effort: ReasoningEffortId | undefined },
|
||||
): void => {
|
||||
const sameRoute = target.current?.provider === selected.provider && target.current.model === selected.model
|
||||
const reasoningEffort = selected.reasoningEffort
|
||||
?? (sameRoute ? target.current?.reasoningEffort ?? selected.reasoning?.defaultEffort : selected.reasoning?.defaultEffort)
|
||||
const reasoningEffort = explicitReasoning === undefined
|
||||
? (sameRoute ? target.current?.reasoningEffort ?? selected.reasoning?.defaultEffort : selected.reasoning?.defaultEffort)
|
||||
: explicitReasoning.effort
|
||||
if (sameRoute && target.current?.reasoningEffort === reasoningEffort) {
|
||||
const reasoning = targetReasoningLabel(selected, reasoningEffort)
|
||||
appendNotice(`Model is already ${targetLabel(selected)}${reasoning === undefined ? '' : ` with reasoning effort ${displayText(reasoning)}`}.`)
|
||||
@@ -2154,9 +2164,9 @@ export function createTuiChat(
|
||||
target.current,
|
||||
resolved.maxModelOptions,
|
||||
palette,
|
||||
(selected) => {
|
||||
(selection) => {
|
||||
void session.close()
|
||||
selectModel(selected)
|
||||
selectModel(selection.choice, { effort: selection.reasoningEffort })
|
||||
},
|
||||
() => { void session.close() },
|
||||
),
|
||||
|
||||
@@ -26,9 +26,9 @@ buffer
|
||||
style 8-8 fg=bright-blue
|
||||
style 38-77 fg=bright-black
|
||||
style 83-83 fg=bright-blue
|
||||
15| " │ → deepseek/deepseek-v4-pro DeepSeek V4 Pro — Off │ "
|
||||
15| " │ → deepseek/deepseek-v4-pro DeepSeek V4 Pro — provider default │ "
|
||||
style 8-8 fg=bright-blue
|
||||
style 10-64 fg=bright-blue inverse
|
||||
style 10-77 fg=bright-blue inverse
|
||||
style 83-83 fg=bright-blue
|
||||
16| " │ │ "
|
||||
style 8-8 fg=bright-blue
|
||||
|
||||
@@ -26,9 +26,9 @@ buffer
|
||||
style 8-8 fg=bright-blue
|
||||
style 10-77 fg=bright-blue inverse
|
||||
style 83-83 fg=bright-blue
|
||||
15| " │ deepseek/deepseek-v4-pro DeepSeek V4 Pro — High │ "
|
||||
15| " │ deepseek/deepseek-v4-pro DeepSeek V4 Pro — provider default │ "
|
||||
style 8-8 fg=bright-blue
|
||||
style 36-65 fg=bright-black
|
||||
style 36-77 fg=bright-black
|
||||
style 83-83 fg=bright-blue
|
||||
16| " │ │ "
|
||||
style 8-8 fg=bright-blue
|
||||
|
||||
@@ -1,25 +1,27 @@
|
||||
terminal 92x32 buffer=normal length=32 base=0 viewport=0
|
||||
lifecycle started=1 stopped=0 progress=inactive
|
||||
title "DSH snapshot"
|
||||
cursor hidden column=1 viewportRow=6 bufferRow=6
|
||||
cursor hidden column=1 viewportRow=7 bufferRow=7
|
||||
buffer
|
||||
0| " DEEPSEEK HARNESS"
|
||||
style 1-8 fg=bright-blue bold
|
||||
style 10-16 bold
|
||||
1| " Snapshot agent ready."
|
||||
style 1-21 fg=bright-black
|
||||
2| " deepseek-v4-pro off • main-session"
|
||||
style 1-36 dim
|
||||
2| " deepseek-v4-pro • main-session"
|
||||
style 1-32 dim
|
||||
3| <blank>
|
||||
4| " Model selected: deepseek/deepseek-v4-pro. Reasoning effort: Off. New steps will use it. "
|
||||
style 1-87 fg=bright-black
|
||||
5| "────────────────────────────────────────────────────────────────────────────────────────────"
|
||||
4| " Model selected: deepseek/deepseek-v4-pro. Reasoning effort: provider default. New steps "
|
||||
style 1-91 fg=bright-black
|
||||
5| " will use it. "
|
||||
style 1-12 fg=bright-black
|
||||
6| "────────────────────────────────────────────────────────────────────────────────────────────"
|
||||
style 0-91 dim
|
||||
6| " "
|
||||
7| " "
|
||||
style 1-1 inverse
|
||||
7| "────────────────────────────────────────────────────────────────────────────────────────────"
|
||||
8| "────────────────────────────────────────────────────────────────────────────────────────────"
|
||||
style 0-91 dim
|
||||
8| "deepseek-v4-pro off /workspace/project ↑0 ↓0 0% context tools:collapsed"
|
||||
style 0-45 dim
|
||||
9| "deepseek-v4-pro /workspace/project ↑0 ↓0 0% context tools:collapsed"
|
||||
style 0-41 dim
|
||||
style 65-91 dim
|
||||
9-31| <blank>
|
||||
10-31| <blank>
|
||||
|
||||
@@ -632,7 +632,7 @@ describe('TUI terminal-state snapshots', () => {
|
||||
await harness.terminal.dispose()
|
||||
})
|
||||
|
||||
it('pins the model selector and selection notice', async () => {
|
||||
it('pins the model selector, effort cycling, and provider-default selection', async () => {
|
||||
const harness = await setupSnapshot({
|
||||
catalog: {
|
||||
providers: [{ id: 'deepseek', name: 'DeepSeek' }],
|
||||
@@ -640,7 +640,7 @@ describe('TUI terminal-state snapshots', () => {
|
||||
{ provider: 'deepseek', id: 'deepseek-v4-flash', name: 'DeepSeek V4 Flash' },
|
||||
{ provider: 'deepseek', id: 'deepseek-v4-pro', name: 'DeepSeek V4 Pro' },
|
||||
],
|
||||
resolveModelInfo: () => Promise.resolve({
|
||||
resolveModelInfo: (_provider, model) => Promise.resolve({
|
||||
context: { contextWindow: 128_000 },
|
||||
reasoning: {
|
||||
efforts: [
|
||||
@@ -648,7 +648,9 @@ describe('TUI terminal-state snapshots', () => {
|
||||
{ id: ReasoningEffortId('high'), name: 'High' },
|
||||
{ id: ReasoningEffortId('max'), name: 'Max' },
|
||||
],
|
||||
defaultEffort: ReasoningEffortId('high'),
|
||||
...model === 'deepseek-v4-flash'
|
||||
? { defaultEffort: ReasoningEffortId('high') }
|
||||
: {},
|
||||
},
|
||||
}),
|
||||
},
|
||||
@@ -662,6 +664,8 @@ describe('TUI terminal-state snapshots', () => {
|
||||
harness.terminal.send('\x1b[B')
|
||||
harness.terminal.send('\x1b[Z')
|
||||
harness.terminal.send('\x1b[Z')
|
||||
harness.terminal.send('\x1b[Z')
|
||||
harness.terminal.send('\x1b[Z')
|
||||
})
|
||||
await checkpoint('model-effort-switching', harness.terminal, { includeScrollback: true })
|
||||
await renderAfter(harness, () => {
|
||||
|
||||
@@ -2387,6 +2387,36 @@ describe('pi-tui chat lifecycle and transcript', () => {
|
||||
await tick()
|
||||
expect(result.terminal.output.slice(providerDefaultOutput)).toContain('Reasoning effort: Standard.')
|
||||
|
||||
const resetDefaultOutput = result.terminal.output.length
|
||||
result.terminal.send('/model')
|
||||
result.terminal.send('\r')
|
||||
await vi.waitFor(() => {
|
||||
expect(result.terminal.output.slice(resetDefaultOutput)).toContain('Alpha Shared — Standard — current')
|
||||
})
|
||||
result.terminal.send('\x1b[Z')
|
||||
await tick()
|
||||
expect(result.terminal.output.slice(resetDefaultOutput)).toContain('Alpha Shared — Ultra — current')
|
||||
result.terminal.send('\x1b[Z')
|
||||
await tick()
|
||||
expect(result.terminal.output.slice(resetDefaultOutput)).toContain('Alpha Shared — provider default')
|
||||
result.terminal.send('\r')
|
||||
await tick()
|
||||
expect(result.terminal.output.slice(resetDefaultOutput)).toContain('Reasoning effort: provider default.')
|
||||
const explicitResetSeed: LlmCallConfig = {
|
||||
provider: 'beta',
|
||||
model: 'b1',
|
||||
reasoningEffort: ReasoningEffortId('max'),
|
||||
}
|
||||
await result.ctx.systemPrompt.assemble(assembleContextFor(result.agent))
|
||||
await expect(agentEvents(result.ctx, result.agent).waterfall(
|
||||
'agent/request',
|
||||
0,
|
||||
0,
|
||||
explicitResetSeed,
|
||||
new AbortController().signal,
|
||||
() => Promise.resolve(explicitResetSeed),
|
||||
)).resolves.toEqual({ provider: 'alpha', model: 'shared' })
|
||||
|
||||
const nonReasoningOutput = result.terminal.output.length
|
||||
result.terminal.send('/model')
|
||||
result.terminal.send('\r')
|
||||
|
||||
Reference in New Issue
Block a user