Use the Zentrik ChatGPT app

Use the app to save customer notes, support summaries, interview snippets, or research observations into Zentrik as evidence records.

The write path is idempotent. ChatGPT provides a stable idempotency key for the source interaction, and repeated calls with the same key return the existing evidence record instead of creating duplicates.

Use search to find relevant evidence, findings, opportunities, and ideas. Use fetch on a search result to retrieve the full context that ChatGPT should summarize.

The search and fetch shapes follow the standard ChatGPT company-knowledge pattern: search returns fetchable IDs, and fetch returns text plus string metadata.

Ask what evidence supports a Zentrik opportunity when you need to understand why it exists before planning, prioritizing, or drafting.

The app returns compact opportunity context, linked findings, and relationship counts. It does not expose broad product graph editing or destructive update tools.

New evidence may continue processing after it is created. Use the processing-status tool when ChatGPT returns a follow-up status handle.

Status responses are compact: they include the evidence record, processing state, and normalized result IDs without raw queue payloads or internal debug fields.

The review account should choose the ChatGPT App Review workspace during OAuth.

That workspace contains synthetic evidence records, generated findings, and OPPORTUNITY-1 so reviewers can test evidence capture, repeated-problem summaries, search/fetch, and opportunity-evidence review without seeing customer data.

Useful review prompts:

• Create a Zentrik evidence record from these customer interview notes.
• What customer problems repeat across recent evidence in Zentrik?
• Search Zentrik for customer evidence about export workflows, then fetch the most relevant result.
• What evidence supports OPPORTUNITY-1?

The app should not trigger for unrelated calendar questions, public posting requests, or destructive product graph changes.