langchain-ai/deepagents
1
0
Fork 0
mirror of https://github.com/langchain-ai/deepagents.git synced 2026-07-21 01:05:27 -04:00
Code Issues 169 Packages Projects Releases 16 Wiki Activity

[PR #1327] feat(cli): Add stateless loop detection middleware with tests #1310

New Issue
Open
opened 2026-02-16 09:18:42 -05:00 by yindo · 0 comments
yindo commented 2026-02-16 09:18:42 -05:00
Owner
Copy Link
Copy Source

📋 Pull Request Information

Original PR: https://github.com/langchain-ai/deepagents/pull/1327
Author: @vtrivedy
Created: 2/13/2026
Status: 🔄 Open

Base: main ← Head: vivek/loop-detection

📝 Commits (1)

  • 4c76cb9 Add stateless loop detection middleware with tests

📊 Changes

3 files changed (+763 additions, -0 deletions)

View changed files

➕ libs/cli/deepagents_cli/loop_detection.py (+241 -0)
📝 libs/cli/pyproject.toml (+3 -0)
➕ libs/cli/tests/unit_tests/test_loop_detection.py (+519 -0)

📄 Description

  • Rewrites the loop detection middleware proposed in #1295 to be fully stateless, addressing review feedback from @eyurtsev and @ccurme
  • Edit counts and warning dedup are derived by scanning state["messages"] on each hook invocation — no mutable instance state
  • Thread-safe, sub-agent safe (shared instance, different state dicts), and summarization-aware (old messages compress → counts naturally reset)

How it works

  • wrap_tool_call: after each edit_file/write_file, counts edits from messages. At soft threshold (8), appends a nudge to the ToolMessage content. Agent can acknowledge or ignore.
  • after_model: counts edits from messages. At hard threshold (12) with no pending tool calls, injects a HumanMessage telling the agent to stop and ask the user, then forces a new model turn via jump_to: "model".
  • Warning dedup: checks if marker text already exists in message history. No sets or dicts on the instance.
  • Path resolution: uses Path.resolve() so relative and absolute paths to the same file are counted together.
  • Only instance attrs are soft_threshold and hard_threshold (immutable config).

Not yet wired into agent.py — kept as a standalone proposal with tests.

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/langchain-ai/deepagents/pull/1327 **Author:** [@vtrivedy](https://github.com/vtrivedy) **Created:** 2/13/2026 **Status:** 🔄 Open **Base:** `main` ← **Head:** `vivek/loop-detection` --- ### 📝 Commits (1) - [`4c76cb9`](https://github.com/langchain-ai/deepagents/commit/4c76cb93b71722f79048c1645696b2af09d834bb) Add stateless loop detection middleware with tests ### 📊 Changes **3 files changed** (+763 additions, -0 deletions) <details> <summary>View changed files</summary> ➕ `libs/cli/deepagents_cli/loop_detection.py` (+241 -0) 📝 `libs/cli/pyproject.toml` (+3 -0) ➕ `libs/cli/tests/unit_tests/test_loop_detection.py` (+519 -0) </details> ### 📄 Description - Rewrites the loop detection middleware proposed in #1295 to be fully stateless, addressing review feedback from @eyurtsev and @ccurme - Edit counts and warning dedup are derived by scanning `state["messages"]` on each hook invocation — no mutable instance state - Thread-safe, sub-agent safe (shared instance, different state dicts), and summarization-aware (old messages compress → counts naturally reset) ## How it works - `wrap_tool_call`: after each `edit_file`/`write_file`, counts edits from messages. At soft threshold (8), appends a nudge to the ToolMessage content. Agent can acknowledge or ignore. - `after_model`: counts edits from messages. At hard threshold (12) with no pending tool calls, injects a HumanMessage telling the agent to stop and ask the user, then forces a new model turn via `jump_to: "model"`. - Warning dedup: checks if marker text already exists in message history. No sets or dicts on the instance. - Path resolution: uses `Path.resolve()` so relative and absolute paths to the same file are counted together. - Only instance attrs are `soft_threshold` and `hard_threshold` (immutable config). Not yet wired into `agent.py` — kept as a standalone proposal with tests. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
yindo added the pull-request label 2026-02-16 09:18:42 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
MCP
backends
bug
cli
config
customization
deepagents
documentation
external
feature
fix
good first issue
help wanted
integration
internal
invalid
middleware
model
needs investigation
p0
p1
p2
p3
performance
pull-request
Mirrored from GitHub Pull Request
 question
sdk
subagents
unable to reproduce
windows
No Label pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langchain-ai/deepagents#1310
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?