Commit graph

67 commits

Author SHA1 Message Date
Yiorgis Gozadinos
721acbcf38
Address review on PR #524
- _budget_notice no longer names the cite tool after prepare_tools has
  withdrawn it; the post-grace state gets the plain no-tools text back.
- Split search-budget rejections from any failed tool call: the code tool
  raises ToolFailed for every error in model-written Python, so
  budget_spent was true for a ZeroDivisionError.
- docs/capabilities/rag.md described the old single-turn removal.
- Drop the rationale clause from the CHANGELOG entry.
2026-07-30 15:50:06 +03:00
Yiorgis Gozadinos
68238f0b6a
Name the readlines workaround when sandbox code iterates a file
50 executions across 40 cases in an 822-case run died on
'_io.TextIOWrapper' object is not iterable, and those cases scored 37.5%
judged against 60.1% and cited 12.5% against 55.2%.
2026-07-30 14:56:17 +03:00
Yiorgis Gozadinos
5752f61f2c
Recover chunk ids mistyped from search results
Resolve a cited id that misses exactly to the nearest id the run
retrieved, above a 0.75 similarity cutoff.
2026-07-30 14:55:43 +03:00
Yiorgis Gozadinos
685a7c393d
Keep the cite tool past a capability's request limit 2026-07-30 14:55:08 +03:00
Yiorgis Gozadinos
221c90af72
Bound a compute-only call at code_timeout
The pool watchdog counts only the time a worker spends running code, so a read
that blocks the worker never trips it. The margin above code_timeout therefore
guarded a race that cannot happen, and only bought a runaway call 90s where 60s
was configured. Pass code_timeout straight through. The two limits are disjoint:
the watchdog bounds a call that computes, and the read deadline bounds a call
that reads.

Drop the ordering test, which was true for any positive margin. The containment
test no longer sets code_timeout to zero, because that value now also disables
the watchdog and kills the worker before the read guard can refuse. It patches
the guard instead.

Monty 0.0.19 runs a plain class and a class with __enter__ and __exit__. Only
inheritance and metaclasses raise. Say that in the instructions.

MemoryFile is no longer constructed, so drop the import, the annotation, and the
stale references in the sandbox docstring and CLAUDE.md. That CLAUDE.md line
also still claimed a ThreadPoolExecutor, _run_async, and a fresh interpreter per
call.
2026-07-28 19:11:19 +03:00
Yiorgis Gozadinos
feaca386d3
Document that file objects cannot be iterated
Monty 0.0.19 supports open() and with blocks, but a file object is still not
iterable. See pydantic/monty#490, which is still open. The instructions now
state the limitation in three places and give the alternative next to each
one: readlines() or read().split("\n").

Replace chr(10) with "\n" in the prose and in the example. Both work on
0.0.19, and chr(10) implies that the escape is broken.

Add a test that pins the limitation. The test fails when Monty gains
iteration support, which is the signal to relax the instructions.
2026-07-28 19:10:53 +03:00
Yiorgis Gozadinos
5c8df37af1
Support open()/with in the analysis sandbox
Document files can be read with open() and with-blocks (.read(),
.readline(), .readlines()); writes raise PermissionError. File objects
remain non-iterable and the collections module is still unavailable.
2026-07-28 19:10:53 +03:00
Yiorgis Gozadinos
53084d6fdd
Bump pydantic-monty to 0.0.19
0.0.19 runs sandboxed code in a subprocess worker pool (AsyncMonty /
AsyncMontySession) and drops MontyRepl. The sandbox checks out a session,
drives it with feed_run, and closes it to return the worker; close() is
now async. Worker crashes surface as a failed SandboxResult. The document
VFS (OSAccess/CallbackFile/MemoryFile) is unchanged.
2026-07-28 19:10:53 +03:00
Yiorgis Gozadinos
f13a3fb677
Report tool failures with ToolFailed 2026-07-27 18:26:42 +03:00
Yiorgis Gozadinos
081856c6b1
Instruct capabilities to judge user-attached images 2026-07-25 14:40:51 +03:00
Yiorgis Gozadinos
f87fabe556
Make the capability image-attachment gate a vision bool 2026-07-24 16:00:01 +03:00
Yiorgis Gozadinos
6c5bc0aae1
Gate combined-chat capability vision on the driving model 2026-07-24 15:27:20 +03:00
Yiorgis Gozadinos
b8dcb066dc
Port cite partial-success feedback to capabilities 2026-07-24 15:27:19 +03:00
Yiorgis Gozadinos
77585ef26a
fix per-question capability limits and tool isolation 2026-07-24 15:26:18 +03:00
Yiorgis Gozadinos
43c17a6777
fix capability execution limits and chat loading 2026-07-24 15:26:17 +03:00
Yiorgis Gozadinos
22ffd5e92e
Add max retries=3 back 2026-07-24 15:26:17 +03:00
Yiorgis Gozadinos
9deb1f2bd4
replace haiku.skills with native Pydantic AI capabilities 2026-07-24 15:26:17 +03:00