The Open-Source Unified Namespace Explorer for the AI Era. Controls the Korelate via Model Context Protocol, providing tools for data exploration, semantic analysis, publishing, file management, simulation, and administration.
Korelate provides 27 tools with consistent naming (verb_noun pattern) and descriptions present for all tools. However, significant gaps exist: parameter descriptions are minimal or absent in many tools, output schemas are undocumented, and error handling lacks recovery guidance. Tool names are clear and action-oriented, but parameter constraints (enums, ranges, formats) are largely missing. The codebase shows basic input validation (e.g., escapeSQL, path traversal prevention) but does not expose this in schema definitions. Average tool description length ~120 chars (below baseline 194), and most parameters lack type constraints or format specifications.
Adds a new mapping rule to the mapper engine.
Adds a semantic tag to associate with one or more topics.
Creates a new alert rule.
Deletes an alert rule by ID.
Describes the schema of the perennial storage system currently active.
Returns the application status including online state, event count, and database size limit.
Retrieves the status of all data connectors.
Retrieves the status of the Dead Letter Queue (DLQ), including total failed messages and the last 50 messages.
Output schemas undocumented. Tools return JSON responses but LLMs cannot infer field names, types, or structure. E.g., get_dlq_status returns {total_failed_messages, messages} but this is not declared in tool definition.
Parameter constraints missing. Tools accept string/number inputs without enums, ranges, or format specs. E.g., get_system_logs accepts 'lines' as number but no min/max bounds (code shows max 100, not declared). search_data accepts 'limit' with no range constraint.
Error handling lacks recovery guidance. Code returns generic error messages (e.g., 'Failed to describe storage: {e.message}') without actionable next steps. LLMs cannot determine if error is retryable, user-fixable, or fatal.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
Retrieves the content of a file from the project root or data directory. Prevents directory traversal attacks.
Retrieves the status of the mapper engine.
Retrieves the UNS (Unified Namespace) model definition, optionally filtered by concept.
Retrieves semantic tags and their associated topics from the semantic manager.
Retrieves the status of all active simulators.
Retrieves system logs. Restricted to admin users only. Returns the last N lines of the korelate.log file.
Retrieves historical messages for a specific topic with optional time window and limit.
Returns a list of all distinct source IDs and topics from the events database.
Lists all active alerts.
Lists files in the root and data directories with safe extensions (.js, .mjs, .json, .md, .svg).
Lists all distinct MQTT topics from the events database, limited to 200 results.
Publishes a message to an MQTT topic.
Executes a query against the perennial storage system. Results are truncated to 100 rows.
Restarts a connector by ID.
Saves a file to the data directory. Filename is sanitized to prevent directory traversal.
Searches the events database for messages matching query terms, with optional time window filtering.
Starts a simulator by ID.
Stops a simulator by ID.
Updates the UNS model definition with new namespaces, object types, or instances. Protects external_instances from being overwritten.
Destructive operations lack confirmation. delete_alert and update_uns_model modify state without dry-run or confirmation step. Agents can accidentally delete alerts or overwrite model definitions.
Duplicate/overlapping tools. get_topics_list and list_topics both return topic information with unclear distinction. LLMs waste reasoning cycles deciding between them. Similarly, get_simulator_status and start_simulator/stop_simulator could be consolidated.