Ahmed Allam 7028b54a84 fix(context): bound the complete retrieval response by the caller's limit ...

The continuation hint was reserved only against the global page cap, so a
small caller-supplied limit bounded content alone and the appended hint pushed
the complete response past the requested maximum. Reserve the hint out of the
effective ceiling (min of limit and the global cap) so content plus hint always
honours the caller's limit, flooring content so pages still make progress.

2026-07-26 00:43:15 +00:00

..

agents

fix(context): keep tool-output retrieval reliable and stream pages

2026-07-26 00:16:02 +00:00

config

fix(context): reject tool-output byte ceilings below the notice size

2026-07-25 23:46:11 +00:00

core

feat(context): spill oversized tool output to disk with a retrieval tool

2026-07-26 00:16:02 +00:00

interface

refactor(reports): fold fence helpers into report writer

2026-07-25 13:09:57 -07:00

llm

fix(context): cap summary output at model limit; safe token upper bound

2026-07-26 00:15:49 +00:00

report

fix(context): clamp shell output cap and count byte-trimmed dropped lines

2026-07-25 22:32:56 +00:00

runtime

runtime: resolve staged local-dir path to avoid symlink rejection on macOS (#857)

2026-07-25 05:45:31 -07:00

skills

fix(prompt): make root agent orchestrate-only and fold fixing into reporting

2026-07-22 02:56:49 -07:00

telemetry

Sign in with a ChatGPT subscription for inference (#854)

2026-07-24 15:41:19 -07:00

tools

fix(context): bound the complete retrieval response by the caller's limit

2026-07-26 00:43:15 +00:00

utils

fix(agent): fix tool schemas not retrieved on pyinstaller binary and validate tool call args

2026-01-14 10:57:32 -08:00

viewer

fix(reports): safe-fence markdown PoC export and share fence helpers

2026-07-25 13:09:57 -07:00

__init__.py

Open-source release for Alpha version

2025-08-08 20:36:44 -07:00