Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
37 commits
Select commit Hold shift + click to select a range
bbd6e1d
feat: add regression eval for finding Edge Function console output
claude Aug 24, 2026
15a71c3
chore: refresh eval results
github-actions[bot] Aug 24, 2026
e3b620b
chore: bump MCP_SERVER_VERSION to 0.11.0 and adapt log evals to query…
claude Aug 25, 2026
febffdf
fix: score the console-output eval from real query_logs calls
claude Aug 25, 2026
3e37545
chore: refresh eval results
github-actions[bot] Aug 25, 2026
21aa9f0
fix: gate the console-output eval on query_logs result evidence
claude Aug 25, 2026
382191d
chore: refresh eval results
github-actions[bot] Aug 25, 2026
d82b8a7
fix: harden console log result evidence
barryroodt Aug 25, 2026
703356b
chore: refresh eval results
github-actions[bot] Aug 25, 2026
85b1011
fix: read log rows through the untrusted-data boundary
claude Aug 25, 2026
8dd3b1a
docs: fix stale helper name in row-extractor comment
claude Aug 25, 2026
b7f0e26
chore: refresh eval results
github-actions[bot] Aug 25, 2026
e3cda5b
Merge branch 'feat/eval-edge-function-console-output' into chore/bump…
claude Aug 25, 2026
d7164d7
chore: drop MCP 0.11 version-isolation scaffolding
claude Aug 25, 2026
5e824d7
docs: drop stale version-pinned experiment reference from eval README
claude Aug 25, 2026
b6e5f60
feat: add regression eval for finding Edge Function console output
claude Aug 24, 2026
aac44c2
chore: refresh eval results
github-actions[bot] Aug 24, 2026
7199bc6
fix: score the console-output eval from real query_logs calls
claude Aug 25, 2026
4900b09
chore: refresh eval results
github-actions[bot] Aug 25, 2026
302bb6c
fix: gate the console-output eval on query_logs result evidence
claude Aug 25, 2026
4e047d4
chore: refresh eval results
github-actions[bot] Aug 25, 2026
4c76b92
fix: harden console log result evidence
barryroodt Aug 25, 2026
782f302
chore: refresh eval results
github-actions[bot] Aug 25, 2026
6146668
fix: read log rows through the untrusted-data boundary
claude Aug 25, 2026
6da2e05
docs: fix stale helper name in row-extractor comment
claude Aug 25, 2026
00c17bf
chore: refresh eval results
github-actions[bot] Aug 25, 2026
6cb9e56
Merge branch 'feat/eval-edge-function-console-output' into chore/bump…
claude Aug 25, 2026
5f2bde3
Revert "Merge branch 'feat/eval-edge-function-console-output' into ch…
claude Aug 25, 2026
ebabfe0
Revert "docs: drop stale version-pinned experiment reference from eva…
claude Aug 25, 2026
571d0d3
Revert "chore: drop MCP 0.11 version-isolation scaffolding"
claude Aug 25, 2026
5bf7cda
Revert "Merge branch 'feat/eval-edge-function-console-output' into ch…
claude Aug 25, 2026
8553f0a
chore: restore #228 interface-frontmatter work after revert reconcili…
claude Aug 25, 2026
aaabf74
chore: drop hardcoded mcp-server versions from log fixture comments
claude Aug 25, 2026
ce22934
Revert "fix: scope investigate-logs-001 counts to invocation rows"
mattrossman Aug 25, 2026
055dcb9
chore: trim log fixture comments touched by the bump
mattrossman Aug 25, 2026
2fe1cd4
chore: keep only necessary fixture comment edits from the version bump
mattrossman Aug 25, 2026
7bcc664
fix: restore mcp#333 citation dropped by the previous trim
mattrossman Aug 25, 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
2 changes: 1 addition & 1 deletion packages/core/src/index.ts
Original file line number Diff line number Diff line change
Expand Up @@ -1235,7 +1235,7 @@ async function getAvailablePort(): Promise<number> {
}

export const ACCESS_TOKEN = 'eval-token';
export const MCP_SERVER_VERSION = '0.8.1';
export const MCP_SERVER_VERSION = '0.11.0';
// Well-formed but inert PAT used when a Supabase MCP server is docs-only: the
// server requires a token to boot but never authenticates without a platform.
const THROWAWAY_ACCESS_TOKEN = `sbp_${'0'.repeat(40)}`;
Expand Down
2 changes: 1 addition & 1 deletion packages/platform-lite/package.json
Original file line number Diff line number Diff line change
Expand Up @@ -28,7 +28,7 @@
"zod": "catalog:"
},
"devDependencies": {
"@supabase/mcp-server-supabase": "^0.8.1",
"@supabase/mcp-server-supabase": "^0.11.0",
"@types/node": "catalog:",
"openapi-typescript": "^7.8.0",
"tsx": "^4.21.0",
Expand Down
16 changes: 8 additions & 8 deletions packages/platform-lite/src/project/log-seeding.ts
Original file line number Diff line number Diff line change
Expand Up @@ -122,24 +122,24 @@ CREATE TABLE IF NOT EXISTS storage_logs (
--
-- Provenance (hand-modeled; nothing usable is importable):
-- * The 'logs' relation shape is the hosted platform's Logflare/ClickHouse
-- contract. Hosted serves the /analytics/endpoints/logs route today; the
-- capabilities are landing per-PR (both in review as of 2026-07-21):
-- contract. Hosted serves the /analytics/endpoints/logs route today; both
-- capability PRs have landed:
-- - supabase/platform#35096: logs.all.otel unified stream + ClickHouse
-- dialect for the getLogs PRESETS - what mcp main (post-#326) emits.
-- - supabase/platform#35970: custom-SQL passthrough (timestamps
-- normalized platform-side) - what mcp#333's query_logs (still open)
-- targets; current mcp main does not depend on it.
-- The fixture models both current main and the #333 validation branch
-- against that contract. It is
-- normalized platform-side) - what mcp#333's query_logs (shipped in
-- mcp 0.10.0) targets.
-- The fixture models both against that contract. It is
-- platform-internal either way: no npm package exports the schema, so
-- fixtures must model it by hand, exactly like every other platform-lite
-- emulation in this package.
-- * The 'source' names are the unified-stream sources referenced by mcp's
-- preset SQL and the query_logs sql description - not exported as data.
-- (mcp's /platform entrypoint DOES export logsServiceSchema, but that
-- enumerates service PRESETS, a different namespace; and the resolved
-- package (^0.8.1) can drift ahead of the MCP_SERVER_VERSION pin the
-- harness actually runs, so importing it would track the wrong artifact.)
-- package's devDependency range can drift ahead of the MCP_SERVER_VERSION
-- pin the harness actually runs, so importing it would track the wrong
-- artifact.)
-- Resync this view when the pinned MCP_SERVER_VERSION moves.
-- * UNMODELED preset sources: workflow_run_logs (branch-action) and
-- realtime_logs (realtime) have no backing table in platform-lite.
Expand Down
11 changes: 6 additions & 5 deletions packages/platform-lite/test/clickhouse-logs.test.ts
Original file line number Diff line number Diff line change
Expand Up @@ -193,11 +193,12 @@ limit 100`
).rejects.toThrow(/unknown log seed source/i);
});

// verbatim from mcp 0.9.0 getClickHouseLogQuery('edge-function-runtime'),
// limit interpolated to 100. This preset is why function_logs must stay
// queryable: the pin (0.8.1) predates it, 0.9.0 ships it, and rejecting the
// source would turn it into a hard error the moment the pin moves.
it('runs the mcp 0.9.0 edge-function-runtime preset against its own rows', async () => {
// verbatim from getClickHouseLogQuery('edge-function-runtime'), limit
// interpolated to 100:
// https://github.com/supabase/mcp/blob/c2826ee85a66d4fd6d4ccfa5086ec823d4a9fc88/packages/mcp-server-supabase/src/logs.ts#L41-L48
// This preset is why function_logs must stay queryable: the pinned server
// ships it, so rejecting the source would be a hard error on every run.
it('runs the edge-function-runtime preset against its own rows', async () => {
const result = await logsDb.query<{
function_id: string;
level: string;
Expand Down
63 changes: 27 additions & 36 deletions pnpm-lock.yaml

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.