chore: merge dev into v2 (#35591)
Co-authored-by: Frank <frank@anoma.ly> Co-authored-by: Aarav Sareen <96787824+arvsrn@users.noreply.github.com> Co-authored-by: Brendan Allan <git@brendonovich.dev> Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com> Co-authored-by: Jack <jack@anoma.ly> Co-authored-by: Brendan Allan <14191578+Brendonovich@users.noreply.github.com> Co-authored-by: Shoubhit Dash <shoubhit2005@gmail.com> Co-authored-by: opencode-agent[bot] <219766164+opencode-agent[bot]@users.noreply.github.com> Co-authored-by: James Long <longster@gmail.com> Co-authored-by: Dustin Deus <deusdustin@gmail.com> Co-authored-by: starptech <starptech@starptechs-MBP.fritz.box> Co-authored-by: Luke Parker <10430890+Hona@users.noreply.github.com> Co-authored-by: 𝓛𝓲𝓽𝓽𝓵𝓮 𝓕𝓻𝓪𝓷𝓴 <little-frank@opencord.local> Co-authored-by: Dax <mail@thdxr.com> Co-authored-by: usrnk1 <7547651+usrnk1@users.noreply.github.com> Co-authored-by: Jay <53023+jayair@users.noreply.github.com> Co-authored-by: runvip <164729189+runvip@users.noreply.github.com> Co-authored-by: opencode <opencode@sst.dev> Co-authored-by: Julian Coy <julian@ex-machina.co> Co-authored-by: Vladimir Glafirov <vglafirov@gitlab.com> Co-authored-by: Adam <2363879+adamdotdevin@users.noreply.github.com> Co-authored-by: Kit Langton <kit.langton@gmail.com> Co-authored-by: Simon Klee <hello@simonklee.dk> Co-authored-by: Jay <air@live.ca> Co-authored-by: David Hill <1879069+iamdavidhill@users.noreply.github.com>
This commit is contained in:
co-authored by
Frank
Aarav Sareen
Brendan Allan
opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>
Jack
Brendan Allan
Shoubhit Dash
opencode-agent[bot] <219766164+opencode-agent[bot]@users.noreply.github.com>
James Long
Dustin Deus
starptech
Luke Parker
𝓛𝓲𝓽𝓽𝓵𝓮 𝓕𝓻𝓪𝓷𝓴
Dax
usrnk1
Jay
runvip
opencode
Julian Coy
Vladimir Glafirov
Adam
Kit Langton
Simon Klee
Jay
David Hill
parent
f87998f37f
commit
9e0d3976e1
@@ -1280,7 +1280,7 @@ describe("SessionRunnerLLM", () => {
|
||||
currentModel = compactModel
|
||||
requests.length = 0
|
||||
responses = [
|
||||
fragmentFixture("text", "text-summary", ["## Goal\n- Preserve the task"]).completeEvents,
|
||||
fragmentFixture("text", "text-summary", ["## Objective\n- Preserve the task"]).completeEvents,
|
||||
fragmentFixture("text", "text-final", ["Continued"]).completeEvents,
|
||||
]
|
||||
yield* session.prompt({
|
||||
@@ -1291,22 +1291,22 @@ describe("SessionRunnerLLM", () => {
|
||||
yield* session.resume(sessionID)
|
||||
|
||||
expect(requests).toHaveLength(2)
|
||||
expect(userTexts(requests[0])[0]).toContain("## Goal")
|
||||
expect(userTexts(requests[0])[0]).toContain("## Objective")
|
||||
expect(userTexts(requests[1])).toHaveLength(1)
|
||||
expect(userTexts(requests[1])[0]).toContain("<summary>\n## Goal\n- Preserve the task\n</summary>")
|
||||
expect(userTexts(requests[1])[0]).toContain("<summary>\n## Objective\n- Preserve the task\n</summary>")
|
||||
expect(userTexts(requests[1])[0]).toContain(`[User]: ${"Recent exact request ".repeat(180)}`)
|
||||
|
||||
const context = yield* (yield* SessionStore.Service).context(sessionID)
|
||||
expect(context.map((message) => message.type)).toEqual(["compaction", "assistant"])
|
||||
expect(context[0]).toMatchObject({
|
||||
type: "compaction",
|
||||
summary: "## Goal\n- Preserve the task",
|
||||
summary: "## Objective\n- Preserve the task",
|
||||
})
|
||||
|
||||
requests.length = 0
|
||||
executions.length = 0
|
||||
responses = [
|
||||
fragmentFixture("text", "text-summary-2", ["## Goal\n- Preserve the updated task"]).completeEvents,
|
||||
fragmentFixture("text", "text-summary-2", ["## Objective\n- Preserve the updated task"]).completeEvents,
|
||||
fragmentFixture("text", "text-final-2", ["Continued again"]).completeEvents,
|
||||
]
|
||||
yield* session.prompt({
|
||||
@@ -1318,12 +1318,12 @@ describe("SessionRunnerLLM", () => {
|
||||
|
||||
expect(requests).toHaveLength(2)
|
||||
expect(userTexts(requests[0])[0]).toContain(
|
||||
"<previous-summary>\n## Goal\n- Preserve the task\n</previous-summary>",
|
||||
"<previous-summary>\n## Objective\n- Preserve the task\n</previous-summary>",
|
||||
)
|
||||
expect(userTexts(requests[0])[0]).toContain("Recent exact request")
|
||||
expect((yield* (yield* SessionStore.Service).context(sessionID))[0]).toMatchObject({
|
||||
type: "compaction",
|
||||
summary: "## Goal\n- Preserve the updated task",
|
||||
summary: "## Objective\n- Preserve the updated task",
|
||||
})
|
||||
}),
|
||||
)
|
||||
@@ -1336,17 +1336,17 @@ describe("SessionRunnerLLM", () => {
|
||||
LLMEvent.stepStart({ index: 0 }),
|
||||
LLMEvent.providerError({ message: "prompt too long", classification: "context-overflow" }),
|
||||
],
|
||||
fragmentFixture("text", "text-summary", ["## Goal\n- Recover overflow"]).completeEvents,
|
||||
fragmentFixture("text", "text-summary", ["## Objective\n- Recover overflow"]).completeEvents,
|
||||
fragmentFixture("text", "text-final", ["Recovered"]).completeEvents,
|
||||
]
|
||||
yield* session.prompt({ sessionID, prompt: PromptInput.Prompt.make({ text: "Continue" }), resume: false })
|
||||
yield* session.resume(sessionID)
|
||||
|
||||
expect(requests).toHaveLength(3)
|
||||
expect(userTexts(requests[1])[0]).toContain("## Goal")
|
||||
expect(userTexts(requests[2])[0]).toContain("<summary>\n## Goal\n- Recover overflow\n</summary>")
|
||||
expect(userTexts(requests[1])[0]).toContain("## Objective")
|
||||
expect(userTexts(requests[2])[0]).toContain("<summary>\n## Objective\n- Recover overflow\n</summary>")
|
||||
expect(yield* session.context(sessionID)).toMatchObject([
|
||||
{ type: "compaction", summary: "## Goal\n- Recover overflow" },
|
||||
{ type: "compaction", summary: "## Objective\n- Recover overflow" },
|
||||
{ type: "assistant", finish: "stop" },
|
||||
])
|
||||
yield* replaySessionProjection(sessionID)
|
||||
@@ -1366,7 +1366,7 @@ describe("SessionRunnerLLM", () => {
|
||||
]
|
||||
responses = [
|
||||
overflow(),
|
||||
fragmentFixture("text", "text-summary", ["## Goal\n- Recover once"]).completeEvents,
|
||||
fragmentFixture("text", "text-summary", ["## Objective\n- Recover once"]).completeEvents,
|
||||
overflow(),
|
||||
]
|
||||
yield* session.prompt({ sessionID, prompt: PromptInput.Prompt.make({ text: "Continue" }), resume: false })
|
||||
@@ -1394,7 +1394,7 @@ describe("SessionRunnerLLM", () => {
|
||||
}),
|
||||
)
|
||||
responses = [
|
||||
fragmentFixture("text", "text-summary", ["## Goal\n- Recover raw overflow"]).completeEvents,
|
||||
fragmentFixture("text", "text-summary", ["## Objective\n- Recover raw overflow"]).completeEvents,
|
||||
fragmentFixture("text", "text-final", ["Recovered"]).completeEvents,
|
||||
]
|
||||
yield* session.prompt({ sessionID, prompt: PromptInput.Prompt.make({ text: "Continue" }), resume: false })
|
||||
@@ -1402,7 +1402,7 @@ describe("SessionRunnerLLM", () => {
|
||||
|
||||
expect(requests).toHaveLength(3)
|
||||
expect(yield* session.context(sessionID)).toMatchObject([
|
||||
{ type: "compaction", summary: "## Goal\n- Recover raw overflow" },
|
||||
{ type: "compaction", summary: "## Objective\n- Recover raw overflow" },
|
||||
{ type: "assistant", finish: "stop" },
|
||||
])
|
||||
}),
|
||||
@@ -1433,7 +1433,7 @@ describe("SessionRunnerLLM", () => {
|
||||
const session = yield* setupOverflowRecovery
|
||||
responses = [
|
||||
[LLMEvent.providerError({ message: "prompt too long", classification: "context-overflow" })],
|
||||
fragmentFixture("text", "text-summary", ["## Goal\n- Interrupted"]).completeEvents,
|
||||
fragmentFixture("text", "text-summary", ["## Objective\n- Interrupted"]).completeEvents,
|
||||
]
|
||||
const firstGate = yield* Deferred.make<void>()
|
||||
const summaryGate = yield* Deferred.make<void>()
|
||||
@@ -2780,6 +2780,148 @@ describe("SessionRunnerLLM", () => {
|
||||
}),
|
||||
)
|
||||
|
||||
it.effect("returns policy-blocked tools to the model and continues", () =>
|
||||
Effect.gen(function* () {
|
||||
yield* setup
|
||||
const session = yield* SessionV2.Service
|
||||
const registry = yield* ToolRegistry.Service
|
||||
yield* registry.register({
|
||||
blocked: Tool.make({
|
||||
description: "Fail because policy blocked execution",
|
||||
input: Schema.Struct({}),
|
||||
output: Schema.Struct({}),
|
||||
execute: () =>
|
||||
Effect.fail(new PermissionV2.BlockedError({ rules: [] })).pipe(
|
||||
Effect.mapError(() => new Tool.Failure({ message: "Permission blocked" })),
|
||||
),
|
||||
}),
|
||||
})
|
||||
yield* session.prompt({ sessionID, prompt: PromptInput.Prompt.make({ text: "Call blocked" }), resume: false })
|
||||
|
||||
requests.length = 0
|
||||
responses = [
|
||||
[
|
||||
LLMEvent.stepStart({ index: 0 }),
|
||||
LLMEvent.toolCall({ id: "call-blocked", name: "blocked", input: {} }),
|
||||
LLMEvent.stepFinish({ index: 0, reason: "tool-calls" }),
|
||||
LLMEvent.finish({ reason: "tool-calls" }),
|
||||
],
|
||||
[
|
||||
LLMEvent.stepStart({ index: 0 }),
|
||||
LLMEvent.stepFinish({ index: 0, reason: "stop" }),
|
||||
LLMEvent.finish({ reason: "stop" }),
|
||||
],
|
||||
]
|
||||
|
||||
yield* session.resume(sessionID)
|
||||
|
||||
expect(requests).toHaveLength(2)
|
||||
expect(yield* session.context(sessionID)).toMatchObject([
|
||||
{ type: "user", text: "Call blocked" },
|
||||
{
|
||||
type: "assistant",
|
||||
content: [
|
||||
{ type: "tool", id: "call-blocked", state: { status: "error", error: { message: "Permission blocked" } } },
|
||||
],
|
||||
},
|
||||
{ type: "assistant", finish: "stop" },
|
||||
])
|
||||
}),
|
||||
)
|
||||
|
||||
it.effect("interrupts runner continuation when permission approval is declined", () =>
|
||||
Effect.gen(function* () {
|
||||
yield* setup
|
||||
const session = yield* SessionV2.Service
|
||||
const registry = yield* ToolRegistry.Service
|
||||
yield* registry.register({
|
||||
declined: Tool.make({
|
||||
description: "Fail because the user declined approval",
|
||||
input: Schema.Struct({}),
|
||||
output: Schema.Struct({}),
|
||||
execute: () => Effect.die(new PermissionV2.DeclinedError()),
|
||||
}),
|
||||
})
|
||||
yield* session.prompt({ sessionID, prompt: PromptInput.Prompt.make({ text: "Call declined" }), resume: false })
|
||||
|
||||
requests.length = 0
|
||||
response = [
|
||||
LLMEvent.stepStart({ index: 0 }),
|
||||
LLMEvent.toolCall({ id: "call-declined", name: "declined", input: {} }),
|
||||
LLMEvent.stepFinish({ index: 0, reason: "tool-calls" }),
|
||||
LLMEvent.finish({ reason: "tool-calls" }),
|
||||
]
|
||||
|
||||
const exit = yield* session.resume(sessionID).pipe(Effect.exit)
|
||||
|
||||
expect(exit._tag).toBe("Failure")
|
||||
if (exit._tag === "Failure") expect(Cause.hasInterruptsOnly(exit.cause)).toBe(true)
|
||||
expect(requests).toHaveLength(1)
|
||||
expect(yield* session.context(sessionID)).toMatchObject([
|
||||
{ type: "user", text: "Call declined" },
|
||||
{
|
||||
type: "assistant",
|
||||
content: [
|
||||
{
|
||||
type: "tool",
|
||||
id: "call-declined",
|
||||
state: { status: "error", error: { message: "Tool execution interrupted" } },
|
||||
},
|
||||
],
|
||||
},
|
||||
])
|
||||
}),
|
||||
)
|
||||
|
||||
it.effect("returns permission corrections to the model and continues", () =>
|
||||
Effect.gen(function* () {
|
||||
yield* setup
|
||||
const session = yield* SessionV2.Service
|
||||
const registry = yield* ToolRegistry.Service
|
||||
yield* registry.register({
|
||||
corrected: Tool.make({
|
||||
description: "Fail with user correction feedback",
|
||||
input: Schema.Struct({}),
|
||||
output: Schema.Struct({}),
|
||||
execute: () =>
|
||||
Effect.fail(new PermissionV2.CorrectedError({ feedback: "Use another tool" })).pipe(
|
||||
Effect.mapError(() => new Tool.Failure({ message: "Use another tool" })),
|
||||
),
|
||||
}),
|
||||
})
|
||||
yield* session.prompt({ sessionID, prompt: PromptInput.Prompt.make({ text: "Call corrected" }), resume: false })
|
||||
|
||||
requests.length = 0
|
||||
responses = [
|
||||
[
|
||||
LLMEvent.stepStart({ index: 0 }),
|
||||
LLMEvent.toolCall({ id: "call-corrected", name: "corrected", input: {} }),
|
||||
LLMEvent.stepFinish({ index: 0, reason: "tool-calls" }),
|
||||
LLMEvent.finish({ reason: "tool-calls" }),
|
||||
],
|
||||
[
|
||||
LLMEvent.stepStart({ index: 0 }),
|
||||
LLMEvent.stepFinish({ index: 0, reason: "stop" }),
|
||||
LLMEvent.finish({ reason: "stop" }),
|
||||
],
|
||||
]
|
||||
|
||||
yield* session.resume(sessionID)
|
||||
|
||||
expect(requests).toHaveLength(2)
|
||||
expect(yield* session.context(sessionID)).toMatchObject([
|
||||
{ type: "user", text: "Call corrected" },
|
||||
{
|
||||
type: "assistant",
|
||||
content: [
|
||||
{ type: "tool", id: "call-corrected", state: { status: "error", error: { message: "Use another tool" } } },
|
||||
],
|
||||
},
|
||||
{ type: "assistant", finish: "stop" },
|
||||
])
|
||||
}),
|
||||
)
|
||||
|
||||
it.effect("fails the drain when tool output persistence fails", () =>
|
||||
Effect.gen(function* () {
|
||||
yield* setup
|
||||
|
||||
Reference in New Issue
Block a user