add_observations

Add text observations to existing entities and index them for full-text and semantic search within the Memento memory server.

Instructions

Add text observations to existing entities and index them for full-text and semantic search.

Input Schema

TableJSON Schema

Name	Required	Description	Default
`observations`	Yes	List of {entityName, contents} pairs.

Implementation Reference

src/knowledge-graph-manager.js:64-87 (handler)

Core handler function that adds observations to entities: resolves entity IDs, inserts new observations, generates embeddings using embedTexts, and stores observation vectors in the repository.

async addObservations(list) {
    const results = [];
    for (const { entityName, contents } of list) {
        const entityId = await this.#repository.getOrCreateEntityId(entityName, 'Unknown');
        const inserted = [];
        for (const content of contents) {
            const { inserted: wasInserted, observationId } = await this.#repository.insertObservation(entityId, content);
            if (wasInserted && observationId !== null && observationId !== undefined) {
                inserted.push({ observationId, content });
            }
        }
        if (inserted.length) {
            const embeddings = await this.embedTexts(inserted.map(row => row.content));
            const vectorRows = inserted.map((row, index) => ({
                observationId: row.observationId,
                entityId,
                embedding: embeddings[index]
            }));
            await this.#repository.insertObservationVectors(vectorRows);
        }
        results.push({ entityName, addedObservations: inserted.map(item => item.content) });
    }
    return results;
}

src/server.js:91-110 (registration)

Registers the 'add_observations' MCP tool, defines input schema using Zod, and provides a thin handler that delegates to KnowledgeGraphManager.addObservations and returns JSON-formatted results.

this.tool(
    'add_observations',
    'Add text observations to existing entities and index them for full-text and semantic search.',
    {
        observations: z.array(z.object({
            entityName: z.string().describe('Name of the entity to annotate.'),
            contents:   z.array(z.string()).describe('List of text observations to add.')
        })).describe('List of {entityName, contents} pairs.')
    },
    async ({ observations }) => ({
        content: [{
            type: 'text',
            text: JSON.stringify(
                await this.#knowledgeGraphManager.addObservations(observations),
                null,
                2
            )
        }]
    })
);

src/server.js:95-99 (schema)

Zod schema defining the input parameters for the add_observations tool: an array of objects with entityName (string) and contents (array of strings).

    observations: z.array(z.object({
        entityName: z.string().describe('Name of the entity to annotate.'),
        contents:   z.array(z.string()).describe('List of text observations to add.')
    })).describe('List of {entityName, contents} pairs.')
},

Memento