mastra-ai/mastra · error

Cannot set both `model` and `observation.model`. Use `model`

Error message

Cannot set both `model` and `observation.model`. Use `model` to set both agents, or set each individually.

What it means

ObservationalMemory lets you set one top-level model for both the observation and reflection agents, or configure each sub-agent individually. Setting the top-level model together with observation.model is ambiguous and rejected at construction.

Source

Thrown at packages/memory/src/processors/observational-memory/observational-memory.ts:470

      // Release the lock
      releaseLock!();
      // Clean up if this is still our lock
      if (this.locks.get(key) === lockPromise) {
        this.locks.delete(key);
      }
    }
  }

  constructor(config: ObservationalMemoryConfig) {
    if (!coreFeatures.has('request-response-id-rotation')) {
      throw new Error(
        'Observational memory requires @mastra/core support for request-response-id-rotation. Please bump @mastra/core to a newer version.',
      );
    }

    // Validate that top-level model is not used together with sub-config models
    if (config.model && config.observation?.model) {
      throw new Error(
        'Cannot set both `model` and `observation.model`. Use `model` to set both agents, or set each individually.',
      );
    }
    if (config.model && config.reflection?.model) {
      throw new Error(
        'Cannot set both `model` and `reflection.model`. Use `model` to set both agents, or set each individually.',
      );
    }

    this.shouldObscureThreadIds = config.obscureThreadIds || false;
    this.storage = config.storage;
    this.scope = config.scope ?? 'thread';
    this.retrieval = Boolean(config.retrieval);
    this.retrievalScope = typeof config.retrieval === 'object' ? (config.retrieval.scope ?? 'resource') : 'resource';
    this.retrievalInstructions = typeof config.retrieval === 'object' ? config.retrieval.instructions : undefined;
    this.retrievalSearch = typeof config.retrieval === 'object' && Boolean(config.retrieval.vector);
    this.onIndexObservations = config.onIndexObservations;
    this.hooks = config.hooks;

View on GitHub (pinned to 75dd419e61)

Solutions

1. Remove the top-level model and set model on observation and reflection individually, OR
2. Remove observation.model and rely on the top-level model for both agents
3. If configs are merged programmatically, strip the top-level model when any sub-config model is present

Example fix

// before
new ObservationalMemory({ model: gpt4o, observation: { model: claude } })
// after (per-agent)
new ObservationalMemory({ observation: { model: claude }, reflection: { model: gpt4o } })
// or (shared)
new ObservationalMemory({ model: gpt4o, observation: {} })

if (config.model && config.observation?.model) {
  throw new Error('Set either top-level model or observation.model, not both');
}

function hasConsistentModelConfig(c: ObservationalMemoryConfig): boolean {
  return !(c.model && (c.observation?.model || c.reflection?.model));
}

try {
  const mem = new ObservationalMemory(config);
} catch (err) {
  if (err instanceof Error && err.message.includes('Cannot set both `model` and `observation.model`')) {
    delete config.observation.model; // let top-level model apply
  } else throw err;
}

When it happens

Trigger: new ObservationalMemory({ model: m1, observation: { model: m2, ... } }) — both a top-level model and an observation.model in the same config.

Common situations: Merging a shared-config object (with top-level model) with a per-agent config that also specifies observation.model; incremental config refactor that left the old top-level model in place after adding sub-config models.

Related errors

• Cannot set both `model` and `reflection.model`. Use `model`
• Cookie password must be at least 32 characters. Set WORKOS_C
• Factory rule version is required.
• ${label} must be an object.
• Factory rules must be an object.

AI-assisted analysis of mastra-ai/mastra@75dd419e61 (2026-08-30). Data as JSON: /api/errors/b4d454d387a43093. Report an issue: GitHub.