Record Managers
Learn about record managers and how to implement it effectively.
2 min read
🆕Recently updated
Last updated: 12/9/2025
Record Managers
Record Manager nodes in InnoSynth-Forjinn give workflows and agents powerful, flexible control over reading, updating, and persisting structured data—whether for session context, incrementing counters, custom audit/log, or propagating state across distributed runs.
Why Use Record Managers?
- Data Persistence: Store and update per-session or global values (user profile, run stats, totals).
- Advanced State Tracking: Retain and update state between workflow executions or agent/LLM steps.
- Audit & Custom Logging: Record decisions, important actions, and operation traces for later analysis or rollback.
- Integration: Bridge to external DB for hybrid use with platform’s internal state.
Key Functions
- Set/Get State Value: Save & retrieve variables (e.g.,
userLastIntent,score,docCounter). - Increment/Decrement: Convenient for counters, usage metering.
- Conditional Update: Change state only if criteria are met (e.g., consensus, approval, threshold).
- History/Audit Storage: Save a log of actions, node outputs, or errors for traceability.
Example Workflow Scenarios
- Session Memory: Remember choices through a multi-turn dialog.
- Evaluating Flows: Tally pass/fail scores while evaluating dataset samples.
- Rate Limiting: Store and validate run counts or quota per user/session.
- Audit Trail: Keep rolling record of agent/tool actions for security/compliance.
Configuration
- Key Field: Name of the state variable or DB record
- Value Field: The value to write (may be variable/output)
- Operation:
set,get,increment,append, ordelete - Scope:
session,user,globalor custom/external store
Using with External DB
- Record Managers can interface with platform storage, Redis, Postgres, or cloud DBs.
- For custom: Add DB connection in Credentials, configure node’s connection/field mapping.
Troubleshooting
- Value not updating: Ensure keys/names match exactly; check for overwrites later in flow.
- Concurrency loss: For high-scale jobs, use atomic/increment ops or DB with transaction support.
- No record found: Check scope; global/user/session separation can hide values not in the selected context.
Best Practices
- Use descriptive keys and document purpose to avoid state collisions
- Audit and routinely clean state store for privacy/compliance
- For critical ops, enable backup/export of record DB
Record Managers are the backbone of stateful/advanced flows—track, debug, and persist anything your workflow or agent needs to remember.