Skip to content

fix(cli): defer threshold compaction during tool loops #12792

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
marius-kilocode merged 2 commits into from
Aug 3, 2026
Merged

fix(cli): defer threshold compaction during tool loops #12792

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f46aa6a
fix(cli): preserve Kilo prompt cache keys
marius-kilocode Jul 31, 2026
0cf3648
fix(cli): defer threshold compaction during tool loops
marius-kilocode Aug 3, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions .changeset/compaction-threshold-boundary.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"@kilocode/cli": patch
---

Prevent configured compaction thresholds from interrupting active tool sequences.
5 changes: 2 additions & 3 deletions packages/opencode/src/session/overflow.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,8 +30,7 @@ export function isOverflow(input: {
if (input.model.limit.context === 0) return false

const count = KiloSessionOverflow.count(input.tokens) // kilocode_change
// kilocode_change start
const cap = KiloSessionOverflow.limit({ cfg: input.cfg, model: input.model, usable: usable(input) })
return count >= cap
// kilocode_change start - post-step checks are safety-only; economic thresholds run in preflight
return count >= usable(input)
// kilocode_change end
}
17 changes: 9 additions & 8 deletions packages/opencode/test/kilocode/session-overflow.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,16 +47,17 @@ function tokens(count: number): MessageV2.Assistant["tokens"] {
return { input: count, output: 0, reasoning: 0, cache: { read: 0, write: 0 } }
}

describe("Kilo auto-compaction threshold", () => {
test("triggers at the configured context percentage", () => {
describe("Kilo post-step compaction safety", () => {
test("ignores the configured threshold after a provider step", () => {
const conf = cfg({ threshold_percent: 75 })
const mdl = model({ context: 200_000, output: 32_000 })

expect(isOverflow({ cfg: conf, model: mdl, tokens: tokens(149_999) })).toBe(false)
expect(isOverflow({ cfg: conf, model: mdl, tokens: tokens(150_000) })).toBe(true)
expect(isOverflow({ cfg: conf, model: mdl, tokens: tokens(167_999) })).toBe(false)
expect(isOverflow({ cfg: conf, model: mdl, tokens: tokens(168_000) })).toBe(true)
})

test("keeps the reserved safety trigger when it is lower", () => {
test("uses the usable context limit when the threshold is high", () => {
const conf = cfg({ threshold_percent: 95 })
const mdl = model({ context: 200_000, output: 32_000 })

Expand All @@ -68,8 +69,8 @@ describe("Kilo auto-compaction threshold", () => {
const conf = cfg({ threshold_percent: 75 })
const mdl = model({ context: 400_000, input: 200_000, output: 32_000 })

expect(isOverflow({ cfg: conf, model: mdl, tokens: tokens(149_999) })).toBe(false)
expect(isOverflow({ cfg: conf, model: mdl, tokens: tokens(150_000) })).toBe(true)
expect(isOverflow({ cfg: conf, model: mdl, tokens: tokens(179_999) })).toBe(false)
expect(isOverflow({ cfg: conf, model: mdl, tokens: tokens(180_000) })).toBe(true)
})

test("ignores a cleared threshold", () => {
Expand Down Expand Up @@ -114,14 +115,14 @@ describe("Kilo auto-compaction threshold", () => {
const conf = cfg({ threshold_percent: 75 })
const mdl = model({ context: 200_000, output: 32_000 })

expect(isOverflow({ cfg: conf, model: mdl, tokens: { ...tokens(149_999), reasoning: 1 } })).toBe(true)
expect(isOverflow({ cfg: conf, model: mdl, tokens: { ...tokens(167_999), reasoning: 1 } })).toBe(true)
})

test("falls back to provider total when normalized usage is unavailable", () => {
const conf = cfg({ threshold_percent: 75 })
const mdl = model({ context: 200_000, output: 32_000 })

expect(isOverflow({ cfg: conf, model: mdl, tokens: { ...tokens(0), total: 150_000 } })).toBe(true)
expect(isOverflow({ cfg: conf, model: mdl, tokens: { ...tokens(0), total: 168_000 } })).toBe(true)
})

test("uses the output cap as the reserve for single-window gateway models", () => {
Expand Down
Loading