fix(cli): harden dictionary iteration and HITL fallback handling#1151
Merged
Mason Daugherty (mdrxy) merged 10 commits intolangchain-ai:masterfrom Feb 9, 2026
Conversation
Wrap dictionary .items() and .values() calls with list() to prevent 'dictionary changed size during iteration' errors when parallel tool calls modify the dictionaries concurrently. This fixes a race condition in textual_adapter.py where multiple concurrent tool calls (e.g., 8 parallel SQL queries) would cause the application to crash because dictionaries were being modified while being iterated over. The fix applies list() wrapper to 6 unprotected iterations: - _build_interrupted_ai_message: current_tool_messages.items() - execute_task_textual: pending_interrupts.items() - execute_task_textual: adapter._current_tool_messages.values() (4 locations) This follows the existing safe pattern already used elsewhere in the same file (lines 568, 705, 739). Fixes langchain-ai#956
Saakshi Gupta (saakshigupta2002)
requested a review
from Mason Daugherty (mdrxy)
as a code owner
github-actions
Bot
added
cli
Extract list() calls to local variables to stay within 88 char limit.
Member
|
Saakshi Gupta (@saakshigupta2002) could you add a test that fails without these changes? |
Contributor
Author
sure! will do that soon |
…chain-ai#956) Add TestDictIterationSafety class with 6 tests that verify: - Direct dict iteration fails with RuntimeError when the dict is modified concurrently (simulated via MutatingDict subclasses) - list() snapshot protects against concurrent modification - _build_interrupted_ai_message handles dict mutation safely These tests would fail if the list() wrappers added in the prior commit were removed, proving the fix is both necessary and sufficient.
Contributor
Author
Just pushed a commit with regression tests. I added a The key idea: I created Two of the tests ( Let me know if you'd like any changes! :) |
Mason Daugherty (mdrxy)
changed the title
Mason Daugherty (mdrxy)
pushed a commit
that referenced
this pull request
Feb 10, 2026
🤖 I have created a release *beep* *boop* --- ## [0.0.20](deepagents-cli==0.0.19...deepagents-cli==0.0.20) (2026-02-10) ### Features * **cli:** `--quiet` flag to suppress non-agent output w/ `-n` ([#1201](#1201)) ([3e96792](3e96792)) * **cli:** add docs link to `/help` ([#1098](#1098)) ([8f8fc98](8f8fc98)) * **cli:** built-in skills, ship `skill-creator` as first ([#1191](#1191)) ([42823a8](42823a8)) * **cli:** enrich built-in skill metadata with license and compatibility info ([#1193](#1193)) ([b8179c2](b8179c2)) * **cli:** implement message queue for CLI ([#1197](#1197)) ([c4678d7](c4678d7)) * **cli:** model switcher & arbitrary chat model support ([#1127](#1127)) ([28fc311](28fc311)) * **cli:** non-interactive mode w/ shell allow-listing ([#909](#909)) ([433bd2c](433bd2c)) * **cli:** support custom working directories and LangSmith sandbox templates ([#1099](#1099)) ([21e7150](21e7150)) ### Bug Fixes * **cli:** `-m` initial prompt submission ([#1184](#1184)) ([a702e82](a702e82)) * **cli:** align skill-creator example scripts with agent skills spec ([#1177](#1177)) ([199d176](199d176)) * **cli:** harden dictionary iteration and HITL fallback handling ([#1151](#1151)) ([8b21fc6](8b21fc6)) * **cli:** per-subcommand help screens, short flags, and skills enhancements ([#1190](#1190)) ([3da1e8b](3da1e8b)) * **cli:** port skills behavior from SDK ([#1192](#1192)) ([ad9241d](ad9241d)), closes [#1189](#1189) * **cli:** rewrite skills create template to match spec guidance ([#1178](#1178)) ([f08ad52](f08ad52)) * **cli:** terminal virtualize scrolling to stop perf issues ([#965](#965)) ([5633c82](5633c82)) * **cli:** update splash thread ID on `/clear` ([#1204](#1204)) ([23651ed](23651ed)) * **deepagents:** refactor summarization middleware ([#1138](#1138)) ([e87001e](e87001e)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
james8814
pushed a commit
to james8814/deepagents
that referenced
this pull request
Mar 1, 2026
…gchain-ai#1151) Fixes langchain-ai#956 Defensive hardening of dictionary iterations in `textual_adapter.py`, plus fixes for incomplete cleanup in HITL fallback branches. **Dictionary iteration:** Wraps all `_current_tool_messages` and `pending_interrupts` iterations with `list()` to snapshot before iterating. **HITL fallback branches:** The `else` branches for unexpected decision types previously fell through to reject without updating tool widget state or clearing `_current_tool_messages`. These now properly call `set_rejected()` and `.clear()`, matching the explicit reject branch. **Logging:** Replaces `with suppress(Exception)` with `try/except` + `logger.debug` so cancellation-cleanup failures are visible. Adds `logger.warning` for unexpected HITL decision types that previously failed silently. --------- Co-authored-by: Mason Daugherty <mason@langchain.dev> Co-authored-by: Mason Daugherty <github@mdrxy.com>
james8814
pushed a commit
to james8814/deepagents
that referenced
this pull request
Mar 1, 2026
🤖 I have created a release *beep* *boop* --- ## [0.0.20](langchain-ai/deepagents@deepagents-cli==0.0.19...deepagents-cli==0.0.20) (2026-02-10) ### Features * **cli:** `--quiet` flag to suppress non-agent output w/ `-n` ([langchain-ai#1201](langchain-ai#1201)) ([3e96792](langchain-ai@3e96792)) * **cli:** add docs link to `/help` ([langchain-ai#1098](langchain-ai#1098)) ([8f8fc98](langchain-ai@8f8fc98)) * **cli:** built-in skills, ship `skill-creator` as first ([langchain-ai#1191](langchain-ai#1191)) ([42823a8](langchain-ai@42823a8)) * **cli:** enrich built-in skill metadata with license and compatibility info ([langchain-ai#1193](langchain-ai#1193)) ([b8179c2](langchain-ai@b8179c2)) * **cli:** implement message queue for CLI ([langchain-ai#1197](langchain-ai#1197)) ([c4678d7](langchain-ai@c4678d7)) * **cli:** model switcher & arbitrary chat model support ([langchain-ai#1127](langchain-ai#1127)) ([28fc311](langchain-ai@28fc311)) * **cli:** non-interactive mode w/ shell allow-listing ([langchain-ai#909](langchain-ai#909)) ([433bd2c](langchain-ai@433bd2c)) * **cli:** support custom working directories and LangSmith sandbox templates ([langchain-ai#1099](langchain-ai#1099)) ([21e7150](langchain-ai@21e7150)) ### Bug Fixes * **cli:** `-m` initial prompt submission ([langchain-ai#1184](langchain-ai#1184)) ([a702e82](langchain-ai@a702e82)) * **cli:** align skill-creator example scripts with agent skills spec ([langchain-ai#1177](langchain-ai#1177)) ([199d176](langchain-ai@199d176)) * **cli:** harden dictionary iteration and HITL fallback handling ([langchain-ai#1151](langchain-ai#1151)) ([8b21fc6](langchain-ai@8b21fc6)) * **cli:** per-subcommand help screens, short flags, and skills enhancements ([langchain-ai#1190](langchain-ai#1190)) ([3da1e8b](langchain-ai@3da1e8b)) * **cli:** port skills behavior from SDK ([langchain-ai#1192](langchain-ai#1192)) ([ad9241d](langchain-ai@ad9241d)), closes [langchain-ai#1189](langchain-ai#1189) * **cli:** rewrite skills create template to match spec guidance ([langchain-ai#1178](langchain-ai#1178)) ([f08ad52](langchain-ai@f08ad52)) * **cli:** terminal virtualize scrolling to stop perf issues ([langchain-ai#965](langchain-ai#965)) ([5633c82](langchain-ai@5633c82)) * **cli:** update splash thread ID on `/clear` ([langchain-ai#1204](langchain-ai#1204)) ([23651ed](langchain-ai@23651ed)) * **deepagents:** refactor summarization middleware ([langchain-ai#1138](langchain-ai#1138)) ([e87001e](langchain-ai@e87001e)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #956
Defensive hardening of dictionary iterations in
textual_adapter.py, plus fixes for incomplete cleanup in HITL fallback branches.Dictionary iteration: Wraps all
_current_tool_messagesandpending_interruptsiterations withlist()to snapshot before iterating.HITL fallback branches: The
elsebranches for unexpected decision types previously fell through to reject without updating tool widget state or clearing_current_tool_messages. These now properly callset_rejected()and.clear(), matching the explicit reject branch.Logging: Replaces
with suppress(Exception)withtry/except+logger.debugso cancellation-cleanup failures are visible. Addslogger.warningfor unexpected HITL decision types that previously failed silently.