Constraints

Constraints are persistent rules that shape how OpenBee behaves. There are two types: global constraints that the Bee reads and writes to accumulate experience across sessions, and worker constraints that define fixed behavioral guidelines for individual workers.

Global Constraints

Global constraints are a persistent key-value store the Bee uses to remember preferences, workflow insights, and learned patterns. They are organized by scope: global for system-wide knowledge, or a session key for user-specific preferences.

How It Works

Each time you interact with the Bee, it loads relevant constraints before responding. When it learns something worth persisting — a preference, a pattern, a workflow insight — it saves it automatically.

You can also manage constraints directly through conversation:

Save a Constraint

"Remember that I prefer concise replies"

"Save a note: our API uses snake_case naming"

View Constraints

"What rules do you have on record about me?"

"Show all saved constraints"

Delete a Constraint

"Forget my language preference"

"Delete the constraint about API naming"

Constraint Structure

Each global constraint entry has:

The Bee uses descriptive keys that act as self-documentation, making constraints easy to understand and manage.

Global vs. Session-Scoped

• Global constraints apply across all users and sessions — useful for universal insights and patterns
• Session-scoped constraints are tied to one user or conversation — useful for personal preferences

The Bee chooses the appropriate scope automatically based on context.

Managing via CLI

You can also manage global constraints directly with the CLI:

# List all global constraints
openbee ctl constraint get --scope global

# Get a specific constraint
openbee ctl constraint get --scope global --key user_language_preference

# Save or update a constraint
openbee ctl constraint save --scope global --key task_routing_insight \
  --value "prefer CodeBot for Python tasks"

# Delete a constraint
openbee ctl constraint delete --scope global --key outdated_preference

Worker Constraints

Worker constraints are a text field on each worker that gets injected into their system prompt at the start of every task. They define the worker's persistent behavioral guidelines — expertise, rules, communication style, and limitations.

How It Works

When a task is assigned to a worker, their constraints are formatted directly into the system prompt:

You are a Worker in an AI team.
Name: CodeBot
Description: Handles code review tasks

## Work Constraints
Always follow PEP 8. Prefer type hints. Write tests for every function.

Worker constraints persist across all tasks and don't change between conversations unless you explicitly update them.

Set Worker Constraints

Set constraints when creating or updating a worker:

"Create a worker named CodeBot with constraints: always follow PEP 8, prefer type hints, write tests for every function"

"Update CodeBot's constraints to include: use pytest instead of unittest"

View Worker Constraints

"What are CodeBot's constraints?"

"Show me CodeBot's details"

Update Worker Constraints

"Change CodeBot's constraints to: focus on Go development, follow Effective Go guidelines"

"Add to CodeBot's constraints: always run linting before committing"

What to Put in Worker Constraints

Worker constraints are well-suited for:

Domain expertise:

"You are an experienced Python developer. Always follow PEP 8 and prefer type hints."

Communication style:

"Respond concisely in bullet points. Avoid long paragraphs."

Safety rules:

"Never make breaking API changes without confirming with the user first."

Audience context:

"You are helping a non-technical user. Explain everything in simple terms."

Global vs. Worker Constraints

Best Practices

Let the Bee Learn Naturally

The Bee proactively saves constraints when it notices preferences or patterns. Interact naturally and it will capture recurring themes on its own.

Use Explicit Constraints for Critical Preferences

For preferences you don't want missed, tell the Bee directly:

"Remember: always assign code review tasks to CodeBot"

"Remember: I prefer detailed explanations over brief answers"

Keep Worker Constraints Focused

Worker constraints should be concise and directly relevant to the worker's role. Avoid overloading them — focus on rules that meaningfully affect output quality.

Review and Clean Up Periodically

Constraints can become outdated. Review and remove stale entries occasionally:

"Show all constraints"

"Delete the constraint about the old API endpoint"