fix: apply_plan() flattens multiple tool call results by khushiiagrawal · Pull Request #155 · mesa/mesa-llm

khushiiagrawal · 2026-03-05T18:28:49Z

Summary

apply_plan() and aapply_plan() flatten multiple tool call results into a single dict, causing all but the last result to be silently lost.

Bug / Issue

Fixes #142

Each tool result has the same keys (name, response), so the flat dict comprehension overwrites earlier entries. If the LLM calls both move_one_step and arrest_citizen, only the arrest is stored in memory.

Implementation

mesa_llm/llm_agent.py — Replaced the flat dict comprehension in apply_plan() and aapply_plan() with a "tool_calls" list that preserves all results.
mesa_llm/memory/memory.py — Added list handling to MemoryEntry.__str__() so the new structure renders correctly.
tests/test_llm_agent.py — Updated existing test assertion; added test_apply_plan_preserves_multiple_tool_calls (sync) and test_aapply_plan_preserves_multiple_tool_calls (async).

Testing

tests/test_llm_agent.py  — 17 passed 
tests/test_memory/       — 40 passed

Signed-off-by: khushiiagrawal <khushisaritaagrawal@gmail.com>

for more information, see https://pre-commit.ci

coderabbitai · 2026-03-05T18:29:03Z

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: ae66decc-503c-48d4-b7f0-f05a9aa47c25

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment

Tip

Try Coding Plans. Let us write the prompt for your AI agent so you can ship faster (with fewer bugs).
Share your feedback on Discord.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

codecov · 2026-03-05T18:29:56Z

Codecov Report

❌ Patch coverage is 92.85714% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 90.66%. Comparing base (4ea91ed) to head (8096e0e).
⚠️ Report is 20 commits behind head on main.

Files with missing lines	Patch %	Lines
mesa_llm/memory/memory.py	92.85%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #155      +/-   ##
==========================================
+ Coverage   90.08%   90.66%   +0.58%     
==========================================
  Files          19       19              
  Lines        1503     1554      +51     
==========================================
+ Hits         1354     1409      +55     
+ Misses        149      145       -4

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Signed-off-by: khushiiagrawal <khushisaritaagrawal@gmail.com>

colinfrisch · 2026-03-07T14:23:09Z

I like this PR and what it solves. I'm going to let copilot review to make sure I didn't miss stuff, but looks quite good to me :)

Copilot

Pull request overview

This PR fixes a bug (#142) where apply_plan() and aapply_plan() in LLMAgent used a flat dict comprehension to merge multiple tool call results, causing all but the last result to be silently lost (since every tool result has the same name/response keys). The fix wraps the tool call results in a "tool_calls" list to preserve all results.

Changes:

Replaced the flat dict comprehension in both apply_plan() and aapply_plan() with a "tool_calls" list structure that preserves all tool call results.
Added list handling to MemoryEntry.__str__() for rendering list values in the display output.
Added comprehensive tests for both sync and async variants with multiple tool calls, plus memory display tests for list values.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
`mesa_llm/llm_agent.py`	Core fix: changed tool call result storage from flat dict to `"tool_calls"` list in both sync and async `apply_plan` methods.
`mesa_llm/memory/memory.py`	Added `list` handling to `MemoryEntry.__str__()` for rendering list values in rich display output.
`tests/test_llm_agent.py`	Updated existing test assertion for new structure; added sync and async tests verifying multiple tool calls are preserved.
`tests/test_memory/test_memory_parent.py`	Added tests for `MemoryEntry.__str__()` with list-of-dicts and list-of-strings content.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.

Copilot · 2026-03-07T14:26:05Z

mesa_llm/memory/memory.py

+            elif isinstance(value, list):
+                for i, item in enumerate(value):
+                    if isinstance(item, dict):
+                        lines.append(f"   [blue]├──[/blue] [cyan]({i + 1})[/cyan]")
+                        lines.extend(format_nested_dict(item, 2))
+                    else:
+                        lines.append(f"   [blue]├──[/blue] [cyan]{item}[/cyan]")


The list handling added here only applies when a list is a direct value at the top level of MemoryEntry.content. However, in the actual data path from apply_plan(), the memory entry content will be {"action": {"tool_calls": [...]}} — the list is nested inside the "tool_calls" key of a dict. When format_nested_dict encounters this list, it falls through to the else branch (line 39-41) and renders it as a raw Python list string, not using the new tree formatting.

To properly handle this, format_nested_dict should also handle list values, similar to the handling added at lines 53-59. Otherwise the nice tree display only works for the synthetic test case, not the real data structure produced by the fix in llm_agent.py.

Copilot · 2026-03-07T14:26:06Z

tests/test_memory/test_memory_parent.py

+        str_repr = str(entry)
+        assert "move_one_step" in str_repr
+        assert "arrest_citizen" in str_repr
+


This test uses content = {"action": [list of dicts]} where the list is a direct value of the "action" key. However, the actual memory entry produced by apply_plan() will have the structure {"action": {"tool_calls": [list of dicts]}} — the list is nested one level deeper. This means the test exercises the new elif isinstance(value, list) branch in __str__() but does not cover the actual data path, where the list is inside format_nested_dict and would fall through to the else (plain value) branch. Consider adding a test with the real structure {"action": {"tool_calls": [...]}} to verify the display works end-to-end.

Suggested change

def test_memory_entry_str_with_nested_tool_calls_list(self):

"""Test MemoryEntry string representation with nested tool_calls list under action."""

mock_agent = Mock()

content = {

"action": {

"tool_calls": [

{"name": "move_one_step", "response": "moved"},

{"name": "arrest_citizen", "response": "arrested"},

]

}

}

entry = MemoryEntry(content=content, step=1, agent=mock_agent)

str_repr = str(entry)

assert "move_one_step" in str_repr

assert "arrest_citizen" in str_repr

Signed-off-by: khushiiagrawal <khushisaritaagrawal@gmail.com>

khushiiagrawal · 2026-03-07T16:39:20Z

@colinfrisch thanks for the feedback. i have addressed all the copilot reviews.
could you please take a look once
thank you!

khushiiagrawal · 2026-03-10T19:52:29Z

@colinfrisch @wang-boyu could you please take a look. thanks.

khushiiagrawal and others added 2 commits March 5, 2026 23:54

fix apply_plan flattening tool calls

6792576

Signed-off-by: khushiiagrawal <khushisaritaagrawal@gmail.com>

[pre-commit.ci] auto fixes from pre-commit.com hooks

83fc482

for more information, see https://pre-commit.ci

test: add coverage for MemoryEntry list rendering

a35c4fe

Signed-off-by: khushiiagrawal <khushisaritaagrawal@gmail.com>

colinfrisch requested a review from Copilot March 7, 2026 14:23

colinfrisch self-assigned this Mar 7, 2026

Copilot started reviewing on behalf of colinfrisch March 7, 2026 14:23 View session

Copilot AI reviewed Mar 7, 2026

View reviewed changes

fix: handle lists in format_nested_dict; add nested tool_calls test

8096e0e

Signed-off-by: khushiiagrawal <khushisaritaagrawal@gmail.com>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix: apply_plan() flattens multiple tool call results#155

fix: apply_plan() flattens multiple tool call results#155
khushiiagrawal wants to merge 4 commits intomesa:mainfrom
khushiiagrawal:bug/142

khushiiagrawal commented Mar 5, 2026

Uh oh!

coderabbitai bot commented Mar 5, 2026 •

edited

Loading

Review skipped

Uh oh!

codecov bot commented Mar 5, 2026 •

edited

Loading

Uh oh!

colinfrisch commented Mar 7, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 7, 2026

Uh oh!

Copilot AI Mar 7, 2026

Uh oh!

khushiiagrawal commented Mar 7, 2026

Uh oh!

khushiiagrawal commented Mar 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

+    def test_memory_entry_str_with_nested_tool_calls_list(self):
+        """Test MemoryEntry string representation with nested tool_calls list under action."""
+        mock_agent = Mock()
+        content = {
+            "action": {
+                "tool_calls": [
+                    {"name": "move_one_step", "response": "moved"},
+                    {"name": "arrest_citizen", "response": "arrested"},
+                ]
+            }
+        }
+        entry = MemoryEntry(content=content, step=1, agent=mock_agent)
+        str_repr = str(entry)
+        assert "move_one_step" in str_repr
+        assert "arrest_citizen" in str_repr

Uh oh!

Conversation

khushiiagrawal commented Mar 5, 2026

Summary

Bug / Issue

Implementation

Testing

Uh oh!

coderabbitai bot commented Mar 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

codecov bot commented Mar 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

colinfrisch commented Mar 7, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 7, 2026

Choose a reason for hiding this comment

Uh oh!

khushiiagrawal commented Mar 7, 2026

Uh oh!

khushiiagrawal commented Mar 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

coderabbitai bot commented Mar 5, 2026 •

edited

Loading

codecov bot commented Mar 5, 2026 •

edited

Loading