AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.
ci(backend): raise the AutoPilot tool-schema budget to 67,651 chars for the in-flight expert PRs (#14484)
Reinier van der Leer67897d0eb9f6
diff --git a/autogpt_platform/backend/backend/copilot/tools/tool_schema_test.py b/autogpt_platform/backend/backend/copilot/tools/tool_schema_test.pyindex d85e4d1a4f..b5654d1325 100644--- a/autogpt_platform/backend/backend/copilot/tools/tool_schema_test.py+++ b/autogpt_platform/backend/backend/copilot/tools/tool_schema_test.py@@ -121,6 +121,23 @@ from backend.copilot.tools import TOOL_REGISTRY# ten merged; 2,253 headroom.# list_expert_chats / read_expert_chat (SECRT-2581) add 1,706 chars and fit# under 65_000 without a bump of their own; merged registry measures 62,694.+# Bumped 65_000 -> 67_651 on 2026-09-09: dev's 62,747 plus the six expert PRs+# then in flight, which add 4,903 between them and blow the old ceiling by 2,650+# while each branch's own CI, measuring only its own delta, stays green.+# Measured per branch, not estimated:+# #14443 fix-expert-credential-grant-paths +3,744+# #14207 multi-expert-teams +981+# #11220 input-blocks-alongside-trigger +178+# #14244 agent-collab-architecture +0+# #14209 autopilot-auto-mode-v2 +0+# #14432 secrt-2593-publish +0+# There is NO margin on top, deliberately. This limit is a brake: it exists to+# make every increase in what AutoPilot pays per turn a decision someone took,+# so slack for growth nobody has measured is the one thing it must not carry.+# The assertion below is a strict <, so the ceiling is the measured total plus+# one — 67,651 admits exactly that aggregate and nothing beyond it.+# The next tool that does not fit raises this line itself, with its own measured+# number and its own row above.## ON CONFLICT, KEEP THE HIGHER VALUE. Two branches tuning this line independently# both look correct: each one's CI only measures its own delta against dev, while@@ -130,7 +147,7 @@ from backend.copilot.tools import TOOL_REGISTRY# Measure it the way this test does — one json.dumps over the whole list —# not by summing per-tool lengths, which misses ~142 chars of array# separators and overstates the headroom.-_CHAR_BUDGET = 65_000+_CHAR_BUDGET = 67_651@pytest.fixture(scope="module")