fix: pass complete context to async tasks instead of truncating to empty/last output

[Om-Borse26]

Description

Currently, tasks configured with async_execution=True do not receive the correct context from prior tasks.
In _execute_tasks (and _aexecute_tasks), the context for async tasks is generated using [last_sync_output] if last_sync_output else []. However, last_sync_output is only populated when resuming from a skipped/conditional task (in prepare_task_execution) and is otherwise None during a normal execution run.
As a result, async tasks receive a completely empty context and are unable to access the outputs of previous tasks in the pipeline.

This PR fixes the issue by passing the full task_outputs list to _get_context() for async tasks, perfectly mirroring the behavior of synchronous tasks.

Root Cause

• In crew.py, sync tasks correctly use task_outputs for context generation.
• Async tasks were hardcoded to use [last_sync_output], which is never updated during standard task execution iterations, leading to an empty context string.

Fix

• Modified lines 1343-1345 and 1555-1557 in src/crewai/crew.py to use task_outputs.

Testing

• pytest suite passes locally with no regressions.
• Verified manually that async tasks now receive the expected concatenated output of all preceding tasks in their context string.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 3a903a3b-da12-4479-9261-65183cf34ce7

📥 Commits

Reviewing files that changed from the base of the PR and between 1452ee2 and 548ca1e.

📒 Files selected for processing (1)

• lib/crewai/src/crewai/crew.py

---

📝 Walkthrough

Walkthrough

This change modifies how context is built for tasks marked async_execution in both Crew._aexecute_tasks() and Crew._execute_tasks(). Instead of passing only the last synchronous output, the full accumulated task_outputs list is now passed as context.

Changes

Async Task Context Construction

Layer / File(s)	Summary
Accumulated context for async tasks lib/crewai/src/crewai/crew.py	Both _aexecute_tasks() and _execute_tasks() now construct async task context from the full accumulated task_outputs list rather than only the most recent synchronous output (last_sync_output).

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main fix: async tasks now receive full context instead of only the last/empty output.
Description check	✅ Passed	The description matches the change and explains the async context bug and fix accurately.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

• X
• Mastodon
• Reddit
• LinkedIn

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

[Om-Borse26]

(Note: As per CONTRIBUTING.md, this PR was authored with the assistance of an AI coding agent. Please apply the llm-generated label to this PR as I do not have permissions to add labels myself.)

Steps to Reproduce / Proof of Bug

To make it easier for reviewers to verify this bug, here is a minimal reproduction script and the execution logs proving that async tasks currently receive a completely empty context string, and that this PR successfully fixes it by passing the full task outputs.

Reproduction Script (demo.py)

import os
from crewai import Agent, Task, Crew, Process

agent = Agent(
    role="Data Compiler",
    goal="Compile specs",
    backstory="You compile specifications.",
    llm="openai/gpt-4o-mini"
)

# 1. Sync Task
task1 = Task(
    description="Output exactly: '[SPECS] V8 Engine'",
    expected_output="A spec string",
    agent=agent,
)

# 2. Sync Task
task2 = Task(
    description="Output exactly: '[SPECS] Carbon Fiber Chassis'",
    expected_output="A spec string",
    agent=agent,
)

# 3. Async Task that relies on context from 1 and 2
task3 = Task(
    description="Review the provided context. If it is empty or missing the V8 Engine and Chassis, output 'FAILED: Context is empty'. If you see them, output them.",
    expected_output="The final combined specs",
    agent=agent,
    async_execution=True, 
    context=[task1, task2] 
)

crew = Crew(
    agents=[agent],
    tasks=[task1, task2, task3],
    process=Process.sequential
)

crew.kickoff()

print("\n--- Context actually handed to Task 3 (Async): ---")
print(task3.output.raw)

Execution Logs (Before this PR - Main Branch)

# The context passed to Task 3 is entirely lost.
--- Context actually handed to Task 3 (Async): ---
FAILED: Context is empty

Execution Logs (After this PR)

# Task 3 successfully receives the accumulated output of all prior tasks in the pipeline.
--- Context actually handed to Task 3 (Async): ---
[SPECS] V8 Engine
[SPECS] Carbon Fiber Chassis

[Om-Borse26]

Fixes #6417