dqe-dedup

DQE One — Create a Deduplication Process (Guided)

Use the dqe-one MCP server to guide the user through creating a check_duplicate process step by step.

Prerequisite: The dqe-one MCP server must be configured. If it is not connected, say so and stop.

---

STEP 0 — Parse arguments

Check $ARGUMENTS for --source=<type>. If present, pre-select that source type and skip the source question. Otherwise ask.

---

STEP 1 — Load available resources (parallel)

Before asking any questions, call these three tools in parallel so you have the data ready:

list_files        → available CSV files
list_rulesets     → available deduplication rulesets
list_credentials  → available data source credentials

If list_rulesets returns an empty list, warn the user:

"No rulesets are configured yet. A ruleset defines the matching and merging rules for deduplication. Please create one in the DQEOne interface first, then come back." Stop here.

---

STEP 2 — Collect configuration (one section at a time)

Ask questions section by section. Do not proceed to the next section until the current one is answered.

Section A — Process name

"What would you like to name this deduplication process?"

Section B — Data source

"Where is your data?

• csv — an uploaded file (shown below)
• salesforce — Salesforce org
• dynamics — Microsoft Dynamics 365
• postgres — PostgreSQL database
• bigquery — Google BigQuery
• sftp — SFTP server"

Show the list of available files (from list_files) and credentials (from list_credentials) to help the user decide.

Depending on the answer:

• csv: ask which file ID to use (show the file list)
• salesforce / dynamics / postgres / bigquery / sftp: ask which credential ID to use (show credential list), then ask for the object/table name
• sftp: additionally ask for input file path and output directory path

Section C — Ruleset

"Which ruleset should govern the matching and merging rules?"

Show the ruleset list (ID, name, source). If only one ruleset exists, propose it automatically.

Section D — Primary key

"Which field is the unique identifier for each record? (e.g. Id, email, customer_id)"

If the source is a CSV file, call get_file with the chosen file ID to show the available column names.

Section E — Field mapping

"Which fields should be included in the deduplication? List the field names, or confirm to use all columns."

Suggest the fields from get_file if CSV; otherwise ask the user to provide them.

Build the fields object: { "FieldName": "Display Label", … }.

---

STEP 3 — Confirmation

Show a complete summary of the configuration:

Process name : <name>
Source       : <source>
File / Object: <file name or object name>
Credential   : <credential name> (if applicable)
Ruleset      : <ruleset name>
Primary key  : <field>
Fields       : <field list>

Ask: "Does this look correct? I'll create the process now."

If the user requests changes, go back to the relevant section.

---

STEP 4 — Create the process

Call create_process with the collected parameters:

{
  "name": "<name>",
  "kind": "check_duplicate",
  "source": "<source>",
  "primaryKey": "<field>",
  "fields": { … },
  "ruleSet": <ruleset_id>,
  "file": <file_id>,          // if source=csv
  "credential": <cred_id>,    // if not csv
  "object": "<table_name>"    // if not csv and not sftp
}

On success: "✅ Process created (ID: )." On error: report the error and ask whether to retry with different parameters.

---

STEP 5 — Offer immediate execution

Ask: "Would you like to run this process now?"

If yes: call run_process with the new process ID, then call list_runs filtered by that process ID to show initial status.

Report:

• If status is running or queued: "⏳ Process is running. Use /dqe-quality:dqe-list in a few minutes to check the final status."
• If status is completed: show record count and any errors.
• If status is failed: show the error message and suggest checking credentials or source connectivity.