06.6 parallel

Parallel Execution

Parallel execution allows multiple branches of a workflow to run independently before synchronizing at a common point.

Instead of executing every step sequentially, BindAI can perform several tasks simultaneously, improving workflow throughput and reducing overall execution time.

Parallel execution is especially useful when branches do not depend on one another.

---

What is Parallel Execution?

A parallel node starts multiple execution paths at the same time.

Conceptually:

Parallel

↙      ↓      ↘

Branch A

Branch B

Branch C

Each branch proceeds independently until synchronization.

---

Why Use Parallel Execution?

Many workflows contain independent operations.

Examples include:

• calling multiple agents
• querying different APIs
• generating several reports
• processing independent datasets
• validating information from multiple sources

Running these operations sequentially increases execution time unnecessarily.

---

Parallel Workflow

A typical parallel workflow looks like this.

Start

↓

Parallel

↙          ↘

Agent A    Agent B

↓

Join

↓

End

The workflow continues only after both branches complete.

---

Parallel Node

The Parallel Node creates multiple execution branches.

Parallel

↓

Branch 1

Branch 2

Branch 3

Each outgoing connection represents one independent execution path.

---

Join Node

Parallel execution is usually followed by a Join Node.

Branch A

↓

Join

↑

Branch B

The Join node waits until every required branch has finished before allowing execution to continue.

---

Synchronization

The Join node acts as a synchronization barrier.

Branch A Complete

✓

Branch B Complete

✓

↓

Continue Workflow

If one branch is still running, the workflow remains paused at the join.

---

Example

Suppose two independent agents gather different information.

Parallel

↙            ↘

Sales Agent  Support Agent

↓

Join

↓

Summary Agent

The Summary Agent receives outputs from both branches.

---

Shared Workflow Context

All branches operate on the same workflow context.

Workflow Context

↙           ↘

Branch A   Branch B

↓

Join

Branches should write to different variables whenever possible to avoid conflicts.

---

Variable Management

A common pattern is assigning each branch its own output variable.

Example:

Branch A

↓

result_a

Branch B

↓

result_b

The Join node allows later steps to access both results.

---

Parallel vs Sequential

Sequential	Parallel
One task at a time	Multiple tasks simultaneously
Simpler execution	Higher throughput
Longer execution time	Faster for independent tasks
No synchronization required	Join node required

Parallel execution improves performance when tasks do not depend on one another.

---

Independent Branches

Branches should remain independent whenever possible.

Good examples:

• multiple AI agents
• multiple API calls
• multiple document searches

Poor examples:

• one branch requiring data produced by another branch

Dependent operations should remain sequential.

---

Error Handling

If one parallel branch fails, workflow behavior depends on the workflow design.

Common strategies include:

• stop the entire workflow
• retry the failed branch
• continue with partial results
• route to an error-handling workflow

Choose the strategy appropriate for your application.

---

Performance Benefits

Parallel execution is especially valuable for:

• AI pipelines with multiple agents
• external service integrations
• document retrieval
• distributed processing
• large-scale automation

Independent work can be completed concurrently rather than waiting for each step to finish.

---

Best Practices

• Use parallel execution only for independent tasks.
• Synchronize branches using a Join node.
• Give each branch separate workflow variables.
• Avoid modifying the same variable from multiple branches.
• Keep branches balanced in complexity when possible.
• Handle branch failures explicitly.

---

Summary

Parallel execution enables BindAI workflows to perform multiple independent tasks simultaneously.

By combining Parallel and Join nodes, workflows can improve performance while maintaining deterministic execution and synchronization across branches.