* fix(llm): cap the tool calls one assistant response may queue
* fix(llm): cap the subscription backend's responses too
---------
Co-authored-by: Ahmed Allam <ahmed39652003@gmail.com>
An exception escaping a non-interactive cycle re-raised before the status
handling, so a dying child stayed 'running' and its parent waited out the
timeout on a completion report the child could no longer send. Set the
terminal status and wake the parent on the way out too.
A tool call for a name Strix does not register raised ModelBehaviorError
from the SDK turn resolver, which nothing retries: the root agent's raise
tore down the whole scan and a sub-agent died before its status was set.
Opt into the SDK's tool_not_found_behavior="return_error_to_model" so the
unknown call comes back as a tool result and the agent self-corrects.
The setting landed in openai-agents 0.19.0, which requires openai>=2.45,
so both pins move.
One tool was doing three jobs (wait on the user, wait on other agents, and
- wrongly - wait for a long-running command), so the driver had to guess which
one an agent meant and used parent_id as the proxy: the root waits for a human,
everyone else waits for agents. That proxy is wrong, since the user can message
any agent from the TUI's agent tree.
Tool identity now carries the intent, and the coordinator records it as a
wait_kind that survives snapshot/restore:
respond_to_user -> wait_kind="user", never auto-resumed (root or not)
wait_for_agents -> wait_kind="agents", auto-resumed on a 300s timer
recovery exhaust -> wait_kind="stalled"
respond_to_user fuses the message and the yield into one call, so there is no
way to answer and then forget to stop - the two-step that gpt-4o-mini skipped
2/2 in live testing. Plain text still renders as before.
Auto-resume is also bounded now: an agent that re-parks after every timeout
burned a model turn every 300s for the rest of the scan (and, since parked
children notify their parent, spammed the parent's inbox on the same cycle).
After _MAX_IDLE_AUTO_RESUMES it stays parked until a real message arrives.
The user can message any agent from the TUI, not only the root, so the
justification is that the parent is an agent with no other way to learn
the child parked - not that the child has no human resumer.
Parking is self-service only for the root, which the user is watching.
A parked child owes its parent a report it can no longer send, so the
parent would wait out its full timeout for nothing.
An exhausted agent parked in 'waiting' got a fresh nudge budget on every
600s auto-resume, so a wedged agent could nudge-park-nudge indefinitely.
Track the count on the coordinator, snapshot it, and reset it only on
real input or an explicit lifecycle tool.
Interactive turns ended by plain text left the agent parked in 'waiting'
forever. Require an explicit lifecycle tool in both modes and nudge a
text-only turn back into a tool call, bounded by a recovery limit.
LiteLLM treats provider-qualified metadata lookups as an auth path.
Use the underlying model slug so context sizing cannot block the scan
loop in a device-code poll.
Adds an integration test that drives Runner.run_streamed against a
non-streaming gateway through _NonStreamingModel: the synthetic terminal
event feeds the runner, which executes the tool call and continues to a
final answer over two non-streaming turns. Removes the README env-var note.
Some OpenAI-compatible gateways don't support Server-Sent Events (or
deliver them unreliably), but the SDK run loop Strix uses only issues
streamed requests, so such a gateway fails every turn. Add an opt-in
LLM_DISABLE_STREAMING setting that wraps the resolved model in
_NonStreamingModel: each turn makes one non-streaming get_response and
replays the completed result as a single terminal stream event, so tool
calls, usage, and the rest of the agent loop are unchanged. Subscription
(ChatGPT) models are always streamed and are not wrapped.
A configured tool_output_max_bytes smaller than the truncation notice
itself can't fit a bounded preview, so a persisted result could exceed the
ceiling. Enforce a config floor (ge=1024) so nonsensical values are
rejected at load time instead of being worked around at runtime.
The head+tail slices could each take half of max_bytes, then the
truncation notice and its separators were appended on top, so the value
persisted to history could exceed the configured maximum. Reserve an
upper bound for the notice (and separators) out of the byte budget before
slicing so the whole joined result stays within max_bytes.
Chat-completions mode converts filesystem CustomTools to FunctionTools
(which bounds their result), but the Responses-API path kept them native
and unbounded, so a large read_file could still exhaust the context
window. Always configure the Filesystem capability to head+tail bound
tool output in both modes.
Treat tool_output_max_tokens as a ceiling so an explicit model-supplied
cap can't exceed it, and derive the truncation notice's dropped-line
count from the lines actually kept after the byte-trim pass. Also cast
the pygments fallback lexer so it satisfies the resolve_lexer return
type under the pre-commit mypy hook.
Cap the size of every tool result so a single verbose command (recursive
find, noisy scanner, full page dump) can't pin the conversation near the
model's context window for the rest of a scan.
- New ContextSettings config group with env-tunable caps.
- Default the SDK shell tools' max_output_tokens so exec_command /
write_stdin truncate head+tail instead of returning unbounded output.
- Bound Strix's own FunctionTool/CustomTool results (line + UTF-8 byte
head+tail preview with a truncation notice) and cap error strings.
