Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
18 commits
Select commit Hold shift + click to select a range
3d803f8
remove per-task wall-clock cap default
ZmeiGorynych Jun 11, 2026
62f7d16
autopsy: generate tool schema from Pydantic model + retry on validati…
ZmeiGorynych Jun 11, 2026
397493c
cascade N1 = upstream ex_base (mini-interact + livesqlbench) + embedd…
ZmeiGorynych Jun 11, 2026
7983a15
process PR #46 reviews: thread conditions, mutation-safety, loader is…
ZmeiGorynych Jun 12, 2026
4e847b2
process PR #46 review round 2: plumb conditions end-to-end, fix sys.p…
ZmeiGorynych Jun 12, 2026
fa41c6a
process PR #46 review round 3: Postgres N1 path, commented-mutation d…
ZmeiGorynych Jun 12, 2026
8a87572
process PR #46 review round 4: own conn lifecycle, plumb conditions t…
ZmeiGorynych Jun 12, 2026
1ab2f87
process PR #46 review round 5: CTE-prefixed mutations, tiktoken encod…
ZmeiGorynych Jun 12, 2026
4dada19
DEV-1550 stage 2: bump slayer to 0.7.4, delete bird-side embedding tr…
ZmeiGorynych Jun 14, 2026
5c35e99
process PR #46 review round 6: bake upstream graders, kill author-lap…
ZmeiGorynych Jun 19, 2026
d8b7ead
process PR #46 review round 6 (continued): grader prereq guard, skip …
ZmeiGorynych Jun 19, 2026
874e705
process PR #46 review round 7: require db_utils marker, validate clou…
ZmeiGorynych Jun 19, 2026
c6edc13
process PR #46 review round 8: validate every resolver branch against…
ZmeiGorynych Jun 19, 2026
f49b901
process PR #46 review round 9: surface the upstream-grader unavailabl…
ZmeiGorynych Jun 19, 2026
ee8e3a5
process PR #46 review round 10: lock the N1 warn-once dedup against t…
ZmeiGorynych Jun 19, 2026
81cc06e
process PR #46 review round 11: lock the upstream-module loader again…
ZmeiGorynych Jun 19, 2026
988aa3b
process PR #46 review round 12: roll back loader sys.path + sys.modul…
ZmeiGorynych Jun 19, 2026
ad7f018
Merge remote-tracking branch 'origin/main' into egor/dev-1550-bird-ag…
ZmeiGorynych Jun 19, 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
18 changes: 18 additions & 0 deletions Dockerfile.cloud
Original file line number Diff line number Diff line change
Expand Up @@ -90,5 +90,23 @@ COPY --from=audited-gold . /app/bird-interact-agents/audited_gold/
# inside the worker (via git's `--git-common-dir`).
COPY --from=annotations . /app/bird-interact-agents/annotations/

# DEV-1550: bake the upstream BIRD-Interact and livesqlbench grader
# subtrees so `eval.upstream_ex_base` can load `test_utils.py` +
# `db_utils.py` in the cloud actor. Without this the cascade-tier-N1
# dispatch silently falls back to legacy `_set_equal` (the loader's
# `FileNotFoundError` is caught upstream and downgraded), so the PR's
# N1=ex_base promise never engages in production.
#
# Build contexts (`bird-interact-evaluation`, `livesqlbench-evaluation`)
# are wired by `cloud.image.build_and_push` and point at the evaluation
# subdirs of `paths.bird_interact_root()` / `paths.livesqlbench_root()`.
# The deeper rel paths preserved here match the host-tree layout so
# `_MINI_INTERACT_REL` / `_LIVESQLBENCH_REL` in `upstream_ex_base` resolve
# identically on cloud and on the developer laptop.
COPY --from=bird-interact-evaluation . \
/app/upstream_graders/bird-interact/mini_interact/knowledge_based/mini_interact_conv/evaluation/
COPY --from=livesqlbench-evaluation . \
/app/upstream_graders/livesqlbench/evaluation/src/

ENV BIRD_RESULTS_ROOT=/tmp/results \
BIRD_INTERACT_AGENTS_CLOUD=1
4 changes: 2 additions & 2 deletions pyproject.toml
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,7 @@ pydantic-ai = [
smolagents = ["smolagents>=1.0"]
agno = ["agno>=2.0"]
mcp-agent = ["mcp-agent>=0.0.16"]
slayer = ["motley-slayer[embedding-search]>=0.7.3"]
slayer = ["motley-slayer[embedding-search]>=0.7.4"]
# `original` brings in the upstream BIRD-Interact mini_interact_agent harness
# so we can call its main.py directly via `uv run python -m batch_run_bird_interact.main`.
original = ["mini-interact-agent"]
Expand All @@ -51,7 +51,7 @@ all = [
"smolagents>=1.0",
"agno>=2.0",
"mcp-agent>=0.0.16",
"motley-slayer[embedding-search]>=0.7.3",
"motley-slayer[embedding-search]>=0.7.4",
"mini-interact-agent",
"ray[default]>=2.30",
"google-cloud-storage>=2.18",
Expand Down
Loading