13 lines
1.7 KiB
Plaintext
13 lines
1.7 KiB
Plaintext
## 13. Prompting: deep analytic experience
|
||
|
||
**Mode A — single transcript, JSON only**
|
||
|
||
> “For this single transcript, use **Mode A: Extraction** and respond only with a single JSON object that follows the top‑level structure in the ‘Single‑transcript JSON response schema’ and the field definitions in ‘Transcript analytics schemas (per‑field definitions)’. Follow the General extraction rules. Output only valid JSON.”
|
||
|
||
**Mode B — single transcript, narrative**
|
||
|
||
> “Using the schemas in this document (support issue categories, ticket tags, wiki categories and slug patterns, game‑specific topics, game detection, Broccolini team IDs, wiki suggestion & outcome analytics), summarize this ticket transcript. Include account & contact, issue (with game_detected and game_or_server), reproduction, environment, priority & impact, rules/abuse if applicable, suggested wiki slugs, and any staff mentions/requests/sentiment. Note whether any wiki article appeared to solve or not solve the issue, and whether the user wanted Broccolini to ‘do it’ or was walked through doing it themselves.”
|
||
|
||
**Mode C — batch analytics**
|
||
|
||
> “Using **Mode C: Batch analytics** over transcripts in [path] or a list of JSON objects from Mode A, compute per‑ticket and aggregate analytics from this document: issue categories, tags, game_detected and game_or_server distributions, wiki usage and success/failure, staff involvement and wiki‑linked outcomes, email analytics, frequency/impact distributions, resolution patterns, intake gaps, and all recurring analytics in the Broccolini support section. Output tables and a concise narrative per major dimension.” |