Overnight: v0.1 completion + v0.2 durable-run pipeline (umbrella) - #101

Closed

leon0399 wants to merge 24 commits into

masterfrom

experiment/overnight-loop

Commits on Jul 1, 2026

chore(api): bump default OpenAI model to gpt-5.4-mini

leon0399
and
claude
committed
docs: add VISION.md — north star, current focus, and non-goals

leon0399
and
claude
committed
feat(api): configurable OpenAI-compatible provider base URL + model (#88)

leon0399
and
claude
committed

Commits on Jul 2, 2026

feat(api): lineage-based conversation context compaction (#57)

leon0399
and
claude
committed
test(api): opt-in Q&A eval suite against the real model (#58)

leon0399
and
claude
committed
feat(api): generate chat titles on the server (#78)

leon0399
and
claude
committed
test(api): verify the Q&A eval set green live — v0.1 complete (#58)

leon0399
and
claude
committed
feat(infra): pg-boss queue + scheduler on Postgres behind a Queue interface (#47)

leon0399
and
claude
committed
feat(api): durable runs + append-only run-event store (#48, first slice)

leon0399
and
claude
committed
feat(api): model.delta persistence + run-event SSE replay by cursor (#48/#49)

leon0399
and
claude
committed
feat(infra): test CI — gate PRs on lint/build/test + rls-test.sh (#70)

leon0399
and
claude
committed
refactor(api): extract run execution from the HTTP path (stages #50)

leon0399
and
claude
committed
feat(api): flag-gated worker execution with run-event stream bridge (#48/#50)

leon0399
and
claude
committed
feat(api): run cancellation — PATCH /runs/:id {status: cancelled} (#48)

leon0399
and
claude
committed
feat(api): run heartbeat + per-run deadman timeout (#48)

leon0399
and
claude
committed
feat(api): per-chat single-flight runs via partial unique index (#48)

leon0399
and
claude
committed
feat(api): close the #55 streaming-loop hardening deferrals (#73)

leon0399
and
claude
committed
feat(api): auth hardening — global fail-closed guard, atomic sessions, TRUST_PROXY (#68)

leon0399
and
claude
committed
feat(api): auth hardening tranche 2 — rate limiting + session cleanup cron (#68)

leon0399
and
claude
committed
feat(api): chat stream-resume endpoint — GET /chats/:id/stream (#49)

leon0399
and
claude
committed
feat(web): resume-on-refresh chat client via reconnectToStream (#49)

leon0399
and
claude
committed
test(e2e): browser chat-flow + mid-run refresh resume in worker mode (#49, #80)

leon0399
and
claude
committed
fix(api): masked lint errors + redelivery double-execution gate

leon0399
and
claude
committed
fix(api): close three races found by adversarial review (#48)

leon0399
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Overnight: v0.1 completion + v0.2 durable-run pipeline (umbrella) - #101

Overnight: v0.1 completion + v0.2 durable-run pipeline (umbrella)#101
leon0399 wants to merge 24 commits into
masterfrom
experiment/overnight-loop

Commits on Jul 1, 2026

chore(api): bump default OpenAI model to gpt-5.4-mini

docs: add VISION.md — north star, current focus, and non-goals

feat(api): configurable OpenAI-compatible provider base URL + model (#88)

Commits on Jul 2, 2026

feat(api): lineage-based conversation context compaction (#57)

test(api): opt-in Q&A eval suite against the real model (#58)

feat(api): generate chat titles on the server (#78)

test(api): verify the Q&A eval set green live — v0.1 complete (#58)

feat(infra): pg-boss queue + scheduler on Postgres behind a Queue interface (#47)

feat(api): durable runs + append-only run-event store (#48, first slice)

feat(api): model.delta persistence + run-event SSE replay by cursor (#48/#49)

feat(infra): test CI — gate PRs on lint/build/test + rls-test.sh (#70)

refactor(api): extract run execution from the HTTP path (stages #50)

feat(api): flag-gated worker execution with run-event stream bridge (#48/#50)

feat(api): run cancellation — PATCH /runs/:id {status: cancelled} (#48)

feat(api): run heartbeat + per-run deadman timeout (#48)

feat(api): per-chat single-flight runs via partial unique index (#48)

feat(api): close the #55 streaming-loop hardening deferrals (#73)

feat(api): auth hardening — global fail-closed guard, atomic sessions, TRUST_PROXY (#68)

feat(api): auth hardening tranche 2 — rate limiting + session cleanup cron (#68)

feat(api): chat stream-resume endpoint — GET /chats/:id/stream (#49)

feat(web): resume-on-refresh chat client via reconnectToStream (#49)

test(e2e): browser chat-flow + mid-run refresh resume in worker mode (#49, #80)

fix(api): masked lint errors + redelivery double-execution gate

fix(api): close three races found by adversarial review (#48)