Skip to content

Commit b9e2bda

Browse files
committed
Merge remote-tracking branch 'upstream/main' into fix/pr-727-review
# Conflicts: # src/scripts/verify-coverage-contract.mjs
2 parents ef7a941 + 500152b commit b9e2bda

42 files changed

Lines changed: 1290 additions & 186 deletions

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

‎.coderabbit.yaml‎

Lines changed: 13 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -48,17 +48,19 @@ reviews:
4848
instructions: >-
4949
Act as an adversarial second-opinion reviewer. Verify PR claims against implementation and
5050
contracts. Trace changed inputs through normal, boundary, error, cancellation, retry, and
51-
default paths and their consumers. Seek plausible counterexamples and regressions from removed
52-
safeguards. Identify assumptions in changed code that depend on facts outside the diff. First
53-
verify repository conventions, tests, and related implementations. When a potential finding
54-
depends on external behavior, use web search and prefer official documentation, specifications,
55-
or upstream repositories. Report only concrete, actionable conflicts or failure modes, citing
56-
the relevant repository location or external source. Prioritize correctness, security, data loss,
57-
lifecycle, and test gaps. Do not report generic best practices, unsupported concerns, speculative
58-
style comments, or unrelated refactors. When changed code introduces a local implementation of a
59-
cross-cutting concern, check whether it bypasses or duplicates an established repository abstraction
60-
or nearby convention. Report only a concrete inconsistency with behavioral or maintenance impact,
61-
and allow intentional deviations.
51+
default paths and their direct test counterparts. Do not flag defects in files not changed by
52+
this PR unless the defect is directly triggered by changed code and cannot be detected in the
53+
changed file alone. Seek plausible counterexamples and regressions from removed safeguards.
54+
Identify assumptions in changed code that depend on facts outside the diff. First verify
55+
repository conventions, tests, and related implementations. When a potential finding depends on
56+
external behavior, use web search and prefer official documentation, specifications, or upstream
57+
repositories. Report only concrete, actionable conflicts or failure modes, citing the relevant
58+
repository location or external source. Prioritize correctness, security, data loss, lifecycle,
59+
and test gaps. Do not report generic best practices, unsupported concerns, speculative style
60+
comments, or unrelated refactors. When changed code introduces a local implementation of a
61+
cross-cutting concern, check whether it bypasses or duplicates an established repository
62+
abstraction or nearby convention. Report only a concrete inconsistency with behavioral or
63+
maintenance impact, and allow intentional deviations.
6264
6365
- path: "**/*.{ts,tsx,js,jsx,mts,mjs,cts,cjs}"
6466
instructions: >-

‎.github/workflows/code-qa.yml‎

Lines changed: 13 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -173,6 +173,14 @@ jobs:
173173
node src/scripts/verify-lcov.mjs src/coverage/services/lcov.info
174174
node src/scripts/verify-lcov.mjs src/coverage/misc/lcov.info
175175
node src/scripts/verify-lcov.mjs src/coverage/tree-sitter/lcov.info
176+
- name: Merge extension coverage reports
177+
run: |
178+
mkdir -p src/coverage/merged
179+
pnpm --dir src run merge:coverage
180+
node src/scripts/verify-lcov.mjs src/coverage/merged/lcov.info
181+
# Validate cache boundaries before publishing any new Turbo entries.
182+
- name: Verify coverage cache inputs
183+
run: pnpm --dir src run verify:coverage-cache-inputs
176184
- name: Save Turbo cache
177185
if: steps.turbo-cache.outputs.cache-hit != 'true'
178186
uses: actions/cache/save@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
@@ -184,21 +192,16 @@ jobs:
184192
# there mostly adds Codecov overhead without changing pass/fail
185193
# behavior.
186194
# Coverage is uploaded in separate steps so each LCOV gets the
187-
# correct flag set. Codecov double-counts overlapping lines when a
188-
# single upload carries multiple flags whose paths overlap, so the
189-
# core lanes and webview lane must be uploaded individually with
190-
# their own flag.
195+
# correct flag set. Extension lanes instrument the same sources, so
196+
# union them before upload; a line is covered when any lane executes
197+
# it. Core and webview reports retain their independent flags.
191198
# See https://docs.codecov.com/docs/flags
192199
- name: Upload non-core coverage to Codecov
193200
if: matrix.upload-coverage
194201
uses: codecov/codecov-action@fb8b3582c8e4def4969c97caa2f19720cb33a72f # v7.0.0
195202
with:
196203
files: >-
197-
src/coverage/api/lcov.info,
198-
src/coverage/core/lcov.info,
199-
src/coverage/services/lcov.info,
200-
src/coverage/misc/lcov.info,
201-
src/coverage/tree-sitter/lcov.info,
204+
src/coverage/merged/lcov.info,
202205
packages/cloud/coverage/lcov.info,
203206
packages/telemetry/coverage/lcov.info,
204207
apps/cli/coverage/lcov.info
@@ -240,6 +243,7 @@ jobs:
240243
src/coverage/services/lcov.info
241244
src/coverage/misc/lcov.info
242245
src/coverage/tree-sitter/lcov.info
246+
src/coverage/merged/lcov.info
243247
webview-ui/coverage/lcov.info
244248
packages/cloud/coverage/lcov.info
245249
packages/telemetry/coverage/lcov.info

‎.github/workflows/mutation-testing.yml‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -16,6 +16,7 @@ concurrency:
1616
jobs:
1717
mutation-diff:
1818
name: mutation-diff
19+
if: github.event_name == 'merge_group' || github.event.pull_request.draft == false
1920
runs-on: ubuntu-latest
2021
timeout-minutes: 30
2122
steps:

‎package.json‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -13,7 +13,7 @@
1313
"check-types": "turbo check-types --log-order grouped --output-logs new-only",
1414
"test": "turbo test --log-order grouped --output-logs new-only",
1515
"test:mutation-ci": "node --test scripts/stryker-diff.test.mjs",
16-
"lifecycle:model-check": "tsx scripts/check-task-lifecycle.ts && tsx scripts/check-task-store-concurrency.ts && tsx scripts/check-provider-handoff-scheduler.ts && pnpm cleanup-protocol:model-check && pnpm parser-scope:model-check && tsx scripts/check-completion-persistence.ts",
16+
"lifecycle:model-check": "tsx scripts/check-task-lifecycle.ts && tsx scripts/check-task-store-concurrency.ts && tsx scripts/check-provider-handoff-scheduler.ts && pnpm cleanup-protocol:model-check && pnpm parser-scope:model-check && tsx scripts/check-completion-persistence.ts && tsx scripts/check-delegated-mode-readers.ts",
1717
"cleanup-protocol:model-check": "tsx scripts/check-task-cleanup-protocol.ts",
1818
"parser-scope:model-check": "node scripts/run-native-tool-call-parser-scoping.mjs",
1919
"mcp:integration-check": "tsx scripts/check-mcp-oauth-integration.ts",

‎packages/types/src/__tests__/deepseek-v4-pro.test.ts‎

Lines changed: 15 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -10,12 +10,18 @@ describe("DeepSeek V4 Pro 0813 provider catalogs", () => {
1010
expect(model?.contextWindow).toBeGreaterThanOrEqual(1_000_000)
1111
})
1212

13-
it("uses peak first-party pricing and unchanged OpenCode Go pricing", () => {
13+
it("uses current peak first-party pricing and unchanged OpenCode Go pricing", () => {
14+
expect(deepSeekModels["deepseek-flash"]).toMatchObject({
15+
supportsImages: true,
16+
outputPrice: 1.2,
17+
cacheWritesPrice: 0.3,
18+
cacheReadsPrice: 0.006,
19+
})
1420
expect(deepSeekModels["deepseek-v4-flash"]).toMatchObject({
15-
supportsImages: false,
16-
outputPrice: 1.32,
17-
cacheWritesPrice: 0.44,
18-
cacheReadsPrice: 0.014,
21+
supportsImages: true,
22+
outputPrice: 1.2,
23+
cacheWritesPrice: 0.3,
24+
cacheReadsPrice: 0.006,
1925
})
2026
expect(deepSeekModels["deepseek-v4-pro"].supportsImages).toBe(false)
2127
expect(deepSeekModels["deepseek-v4-pro"]).toMatchObject({
@@ -42,6 +48,10 @@ describe("DeepSeek V4 Pro 0813 provider catalogs", () => {
4248
expect(model.supportsPromptCache).toBe(true)
4349
expect(model.contextWindow).toBeGreaterThanOrEqual(1_000_000)
4450
expect(model.supportsReasoningEffort).toEqual(["disable", "low", "high", "max"])
51+
expect(model).toMatchObject({ outputPrice: 1.2, cacheWritesPrice: 0.3, cacheReadsPrice: 0.006 })
52+
expect(model.description).toContain("Legacy model name")
53+
expect(model).not.toHaveProperty("supportsTemperature")
54+
expect(model).not.toHaveProperty("defaultTemperature")
4555
})
4656

4757
// Self-hosted providers retain separate IDs for the preview weights and 0813 checkpoint.

‎packages/types/src/providers/deepseek.ts‎

Lines changed: 27 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -6,23 +6,38 @@ import type { ModelInfo } from "../model.js"
66
// continuation within the same turn. See: https://api-docs.deepseek.com/guides/thinking_mode
77
export type DeepSeekModelId = keyof typeof deepSeekModels
88

9-
export const deepSeekDefaultModelId: DeepSeekModelId = "deepseek-v4-flash"
9+
export const deepSeekDefaultModelId: DeepSeekModelId = "deepseek-flash"
1010

1111
export const deepSeekModels = {
12+
"deepseek-flash": {
13+
maxTokens: 384_000,
14+
contextWindow: 1_000_000,
15+
supportsImages: true,
16+
supportsPromptCache: true,
17+
supportsReasoningEffort: ["disable", "low", "high", "max"], // Updated 2026-09-10
18+
preserveReasoning: true,
19+
reasoningEffort: "high",
20+
inputPrice: 0, // the inputs are priced as cache read/write, so `inputPrice` should be 0
21+
// Static estimates use peak rates; off-peak rates are 50% lower. Effective 2026-09-10.
22+
outputPrice: 1.2,
23+
cacheWritesPrice: 0.3,
24+
cacheReadsPrice: 0.006,
25+
description: `DeepSeek-V4.1-Flash is DeepSeek's fast multimodal model with image understanding. It supports thinking and non-thinking modes, JSON output, tool calls, chat prefix completion (beta), and FIM completion (beta) in non-thinking mode.`,
26+
},
1227
"deepseek-v4-flash": {
1328
maxTokens: 384_000,
1429
contextWindow: 1_000_000,
15-
supportsImages: false,
30+
supportsImages: true,
1631
supportsPromptCache: true,
1732
supportsReasoningEffort: ["disable", "low", "high", "max"], // Updated 2026-08-13
1833
preserveReasoning: true,
1934
reasoningEffort: "high",
2035
inputPrice: 0, // the inputs are priced as cache read/write, so `inputPrice` should be 0
21-
// Static estimates use peak rates; off-peak rates are 50% lower. Effective 2026-08-16.
22-
outputPrice: 1.32,
23-
cacheWritesPrice: 0.44,
24-
cacheReadsPrice: 0.014,
25-
description: `DeepSeek-V4-Flash is DeepSeek's fast, cost-efficient V4 model. It supports thinking and non-thinking modes, JSON output, tool calls, chat prefix completion (beta), and FIM completion (beta) in non-thinking mode.`,
36+
// This retired ID is billed as the current Flash model.
37+
outputPrice: 1.2,
38+
cacheWritesPrice: 0.3,
39+
cacheReadsPrice: 0.006,
40+
description: `Legacy model name routed to the latest DeepSeek Flash model, which supports image input. Use deepseek-flash for new configurations.`,
2641
},
2742
"deepseek-v4-pro": {
2843
displayName: "DeepSeek V4 Pro 0813",
@@ -49,14 +64,12 @@ export const deepSeekModels = {
4964
supportsReasoningEffort: ["disable", "low", "high", "max"], // Updated 2026-08-13
5065
preserveReasoning: true,
5166
reasoningEffort: "high",
52-
supportsTemperature: true,
53-
defaultTemperature: 1.0,
5467
inputPrice: 0, // the inputs are priced as cache read/write, so `inputPrice` should be 0
55-
// Static estimates use peak rates; off-peak rates are 50% lower.
56-
outputPrice: 1.32,
57-
cacheWritesPrice: 0.44,
58-
cacheReadsPrice: 0.014,
59-
description: `DeepSeek-V4-Flash-Vision-Exp is DeepSeek's experimental multimodal V4 Flash model with image understanding. It supports thinking and non-thinking modes, JSON output, tool calls, chat prefix completion (beta), and image input through Chat Completions, Responses, and Anthropic-compatible APIs.`,
68+
// This retired ID is billed as the current Flash model.
69+
outputPrice: 1.2,
70+
cacheWritesPrice: 0.3,
71+
cacheReadsPrice: 0.006,
72+
description: `Legacy model name routed to the latest DeepSeek Flash model, which supports image input. Use deepseek-flash for new configurations.`,
6073
},
6174
} as const satisfies Record<string, ModelInfo>
6275

Lines changed: 170 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,170 @@
1+
// check-delegated-mode-readers.ts
2+
//
3+
// Refinement check for the delegated-child mode-reader invariant (issue #1623).
4+
//
5+
// check-provider-handoff-scheduler.ts verifies the write side: that
6+
// selectHandoffExecutionContext stores the task-local mode correctly.
7+
// This script verifies the read side: that the mode observable by
8+
// tool-validation readers is the task-local mode, not the shared provider mode.
9+
//
10+
// The VS Code-dependent readers (getEnvironmentDetails,
11+
// presentAssistantMessage) are covered by their vitest regression tests.
12+
// This script covers the pure-TS parts of the invariant chain and proves
13+
// that the two sources of mode are observably different, so any reader
14+
// that uses the wrong source silently produces wrong behavior.
15+
//
16+
// Invariant: for any delegated child task C with taskMode = M,
17+
// toolAllowedForMode(tool, M) ≠ toolAllowedForMode(tool, providerMode)
18+
// whenever M ≠ providerMode and the two modes differ on the tool's group.
19+
20+
import assert from "node:assert/strict"
21+
22+
import { DEFAULT_MODES } from "../packages/types/src/mode"
23+
24+
import { TOOL_GROUPS, ALWAYS_AVAILABLE_TOOLS, TOOL_ALIASES } from "../src/shared/tools"
25+
import { selectHandoffExecutionContext, type TaskExecutionContext } from "../src/core/task/providerHandoff"
26+
27+
// ---------------------------------------------------------------------------
28+
// Minimal inline mode-allows-tool check.
29+
// Avoids importing src/shared/modes.ts, which pulls in VS Code.
30+
// Only covers built-in modes (no custom modes, no file-regex options).
31+
// That is enough to prove the behavioral divergence this check needs.
32+
// ---------------------------------------------------------------------------
33+
34+
type ModeConfig = (typeof DEFAULT_MODES)[number]
35+
type GroupEntry = ModeConfig["groups"][number]
36+
37+
function groupName(entry: GroupEntry): string {
38+
return Array.isArray(entry) ? entry[0] : (entry as string)
39+
}
40+
41+
function toolAllowedForMode(tool: string, modeSlug: string): boolean {
42+
const resolvedTool = (TOOL_ALIASES as Record<string, string>)[tool] ?? tool
43+
if ((ALWAYS_AVAILABLE_TOOLS as readonly string[]).includes(resolvedTool)) return true
44+
const mode = DEFAULT_MODES.find((m) => m.slug === modeSlug)
45+
if (!mode) return false
46+
for (const entry of mode.groups) {
47+
const groupTools = (TOOL_GROUPS as Record<string, { tools: readonly string[] }>)[groupName(entry)]?.tools ?? []
48+
if (groupTools.includes(resolvedTool)) return true
49+
}
50+
return false
51+
}
52+
53+
// ---------------------------------------------------------------------------
54+
// Scenario: parent in "orchestrator" mode delegates child to "code".
55+
// Regression behavior: both readers used providerMode ("orchestrator").
56+
// Correct behavior: readers use taskMode ("code").
57+
//
58+
// orchestrator groups: [] → apply_diff blocked
59+
// code groups: [...edit] → apply_diff allowed
60+
// ---------------------------------------------------------------------------
61+
62+
const parentCtx: TaskExecutionContext = {
63+
mode: "orchestrator",
64+
apiConfigName: undefined,
65+
apiConfiguration: { apiProvider: "anthropic", consecutiveMistakeLimit: 3 },
66+
}
67+
68+
// 1. Handoff stores the task-local mode, not the parent mode.
69+
const childCtx = selectHandoffExecutionContext(parentCtx, "code", parentCtx.mode, false, undefined)
70+
assert.equal(childCtx.mode, "code", "handoff must store the requested task-local mode")
71+
assert.notEqual(childCtx.mode, parentCtx.mode, "test scenario requires divergent provider and task modes")
72+
73+
// 2. The two modes produce observably different tool-validation outcomes.
74+
assert.equal(toolAllowedForMode("apply_diff", "orchestrator"), false, "orchestrator has no edit group")
75+
assert.equal(toolAllowedForMode("apply_diff", "code"), true, "code has the edit group")
76+
77+
// 3. Regression claim: a reader that consumes providerMode rejects apply_diff;
78+
// a reader that consumes taskMode correctly allows it.
79+
const viaProviderMode = toolAllowedForMode("apply_diff", parentCtx.mode) // "orchestrator" — wrong source
80+
const viaTaskMode = toolAllowedForMode("apply_diff", childCtx.mode) // "code" — correct source
81+
assert.equal(viaProviderMode, false, "stale provider mode rejects apply_diff (regression behavior)")
82+
assert.equal(viaTaskMode, true, "task-local mode allows apply_diff (correct behavior)")
83+
84+
// 4. Additional mode pairs that show the same divergence.
85+
const DIVERGENT_PAIRS: Array<{
86+
label: string
87+
providerMode: string
88+
taskMode: string
89+
probe: string
90+
blockedInProvider: boolean
91+
allowedInTask: boolean
92+
}> = [
93+
// orchestrator → code: edit tools blocked at provider level, allowed at task level
94+
{
95+
label: "orchestrator→code apply_diff",
96+
providerMode: "orchestrator",
97+
taskMode: "code",
98+
probe: "apply_diff",
99+
blockedInProvider: true,
100+
allowedInTask: true,
101+
},
102+
// orchestrator → code: command tools blocked at provider level, allowed at task level
103+
{
104+
label: "orchestrator→code execute_command",
105+
providerMode: "orchestrator",
106+
taskMode: "code",
107+
probe: "execute_command",
108+
blockedInProvider: true,
109+
allowedInTask: true,
110+
},
111+
// code → ask: edit tools allowed at provider level, blocked at task level
112+
{
113+
label: "code→ask apply_diff",
114+
providerMode: "code",
115+
taskMode: "ask",
116+
probe: "apply_diff",
117+
blockedInProvider: false,
118+
allowedInTask: false,
119+
},
120+
// ask → code: edit tools blocked at provider level, allowed at task level
121+
{
122+
label: "ask→code write_to_file",
123+
providerMode: "ask",
124+
taskMode: "code",
125+
probe: "write_to_file",
126+
blockedInProvider: true,
127+
allowedInTask: true,
128+
},
129+
]
130+
131+
for (const pair of DIVERGENT_PAIRS) {
132+
const ctx = selectHandoffExecutionContext(
133+
{ ...parentCtx, mode: pair.providerMode },
134+
pair.taskMode,
135+
pair.providerMode,
136+
false,
137+
undefined,
138+
)
139+
assert.equal(ctx.mode, pair.taskMode, `${pair.label}: handoff must store task-local mode`)
140+
assert.equal(
141+
toolAllowedForMode(pair.probe, pair.providerMode),
142+
!pair.blockedInProvider,
143+
`${pair.label}: wrong provider-mode result`,
144+
)
145+
assert.equal(
146+
toolAllowedForMode(pair.probe, pair.taskMode),
147+
pair.allowedInTask,
148+
`${pair.label}: wrong task-mode result`,
149+
)
150+
// The two sources disagree, so using the wrong one is always observable.
151+
assert.notEqual(
152+
toolAllowedForMode(pair.probe, pair.providerMode),
153+
toolAllowedForMode(pair.probe, pair.taskMode),
154+
`${pair.label}: provider and task mode must differ on this probe tool`,
155+
)
156+
}
157+
158+
// 5. For every built-in mode as a delegation target: selectHandoffExecutionContext
159+
// always stores the requested mode, regardless of parent mode.
160+
for (const mode of DEFAULT_MODES) {
161+
const ctx = selectHandoffExecutionContext(parentCtx, mode.slug, parentCtx.mode, false, undefined)
162+
assert.equal(ctx.mode, mode.slug, `handoff must store ${mode.slug}, not parent mode ${parentCtx.mode}`)
163+
}
164+
165+
console.log(
166+
`Delegated mode reader check passed: ` +
167+
`regression scenario verified, ` +
168+
`${DIVERGENT_PAIRS.length} divergent-mode pairs checked, ` +
169+
`${DEFAULT_MODES.length}/${DEFAULT_MODES.length} built-in modes verified`,
170+
)

0 commit comments

Comments
 (0)