Skip to content

fix(cli): avoid full part-table scan in session model usage #11893

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 1 commit into from
Jul 2, 2026
Merged

fix(cli): avoid full part-table scan in session model usage #11893

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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/fix-model-usage-full-scan.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"kilo-code": patch
---

Fix slow message loading when opening or switching sessions. The per-model token usage breakdown scanned the entire message history on every session open, which blocked the transcript from rendering for several seconds on large histories.
31 changes: 14 additions & 17 deletions packages/opencode/src/kilocode/session/model-usage.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,19 +94,13 @@ export namespace ModelUsage {
FROM family
ORDER BY id`

const USAGE_SQL = `
WITH RECURSIVE family(id) AS (
SELECT id
FROM session
WHERE id = ? AND project_id = ?

UNION

SELECT child.id
FROM session AS child
JOIN family AS parent ON child.parent_id = parent.id
WHERE child.project_id = ?
), step AS (
// Scope aggregation to the already-resolved family session IDs via an IN list.
// Re-deriving the family with an inline recursive CTE prevents SQLite from
// using part_session_idx and forces a full scan of the entire part table
// (seconds on large histories), which blocks the single-threaded server on
// every session open. A concrete IN list lets the planner seek the index.
const usageSql = (placeholders: string) => `
WITH step AS (
SELECT
coalesce(json_extract(part.data, '$.model.providerID'), json_extract(message.data, '$.providerID')) AS providerID,
coalesce(json_extract(part.data, '$.model.modelID'), json_extract(message.data, '$.modelID')) AS modelID,
Expand All @@ -116,10 +110,10 @@ export namespace ModelUsage {
max(0, cast(coalesce(json_extract(part.data, '$.tokens.reasoning'), 0) AS INTEGER)) AS reasoning,
max(0, cast(coalesce(json_extract(part.data, '$.tokens.cache.read'), 0) AS INTEGER)) AS cache_read,
max(0, cast(coalesce(json_extract(part.data, '$.tokens.cache.write'), 0) AS INTEGER)) AS cache_write
FROM family
JOIN part ON part.session_id = family.id
FROM part
JOIN message ON message.id = part.message_id AND message.session_id = part.session_id
WHERE json_extract(part.data, '$.type') = 'step-finish'
WHERE part.session_id IN (${placeholders})
Comment thread
marius-kilocode marked this conversation as resolved.
AND json_extract(part.data, '$.type') = 'step-finish'
AND json_extract(message.data, '$.role') = 'assistant'
)
SELECT
Expand Down Expand Up @@ -163,7 +157,10 @@ export namespace ModelUsage {
.prepare<{ id: SessionID }, [string, string, string]>(FAMILY_SQL)
.all(...familyArgs)
.map((item) => item.id)
const rows = db.prepare<Row, [string, string, string]>(USAGE_SQL).all(...familyArgs)
const rows =
sessionIDs.length === 0
? []
: db.prepare<Row, string[]>(usageSql(sessionIDs.map(() => "?").join(","))).all(...sessionIDs)
const totals = empty()
const models = rows.map((row): Model => {
totals.steps += row.steps
Expand Down
Loading