Search and filters

AuditorIA indexes every call by dozens of criteria. This guide covers the three search paths (quick, card-based, and full history) and how to combine them to find exactly the call you need.

Quick search by card

From the Dashboard, the quick-access cards launch pre-filtered searches:

Card	Applied filter
Search by campaign	specific campaign_id
Search by operator	operator_id or name
Search by date	Date range with date-picker
Search by direction	Phone direction or geography
View problematic audios	Negativity or hate speech > threshold

Clicking a card prompts for the specific criterion. Example — Search by campaign:

Enter the ID and click Confirm. The system builds the query using the composite index (campaign_id, date) for fast responses even on millions of records.

---

Result list

Results are shown in a dense table with server-side pagination.

Columns

Column	What it shows
☐	Checkbox for bulk selection
Ticket	Internal ticket ID (e.g., 16722762)
Campaign	Numeric campaign ID
User	Agent operator_id
Date	Call date
Time	Local time
File name	Original .mp3 name
Status	Badge: Ready to process, Transcribe audio, Audio not found

Ticket states

State	Meaning
Ready to process	Audio available, task not yet created
Transcribe audio	Hover → CTA to launch manual transcription
Audio not found	File is not in MinIO (deleted, never uploaded, rotated)

Pagination controls

• Recordings per page: 10, 25, 50, or 100.
• Navigation: < > + page-number input.
• Counter: "0 of 1816 rows selected" + "Page 179 of 182" in the lower corner.

info

Performance

The backend uses range queries (date >= :start AND date < :end) over composite indexes — date filtering is O(log n) even on tables with tens of millions of rows. See the architecture note.

---

Full history and combined filters

From the sidebar → Dashboard > History you access the advanced view with every available filter.

Available filters

Filter	Type	Detail
Status	Multi-select	pending, processing, completed, error, approved, rejected
Date range	Date picker	Start / end, presets (today, 7d, 30d, custom)
Campaign	Multi-select	Respects user permissions
Operator	Autocomplete	By name or operator_id
Engine	Select	whisperx / openai / deepgram
Direction	Radio	inbound / outbound / any
Duration	Range slider	0s to 2h
Score	Range	0 to 100
Sentiment	Multi-select	positive / neutral / negative
Tags	Chips	Filter by GPT-generated tags
Free text	Input	Full-text search within the transcription

Combining filters

Example: completed tasks from campaign 162, by operator 1897, in the last 30 days, with negative sentiment, sorted by descending score.

tip

Persistence

Filters are saved in the URL (?status=completed&campaign=162&...) so you can share the view by link.

---

Full-text search

The Text search field searches inside the transcribed content.

Supported operators

Operator	Example	Behavior
Implicit AND	cancel service	Both words in the same segment
OR	refund OR return	Either of the two
"..."	"customer number"	Exact phrase
-	sales -spam	Excludes the term
~	invoice~	Fuzzy search (tolerates typos)

info

Search engine

Texts are indexed in PostgreSQL using tsvector + GIN indexes for sub-second search up to ~10M segments.

---

Sorting

Click a column header to toggle asc/desc. Compound sorting is supported by holding Shift while clicking.

Column	Order	Typical use
Date	DESC by default	See the most recent first
Score	ASC	Detect failed audits
Duration	DESC	Identify abnormally long calls
Operator	Alphabetical ASC	Group by agent

---

Bulk selection and batch actions

1. Use the header checkbox to Select all on the page.
2. Or check tasks individually.
3. A floating action bar appears:

Action	Behavior
Approve all	Marks selected tasks as approved
Re-run	Reprocesses the selected tasks (same or different engine)
Export	Downloads CSV with the selected rows
Assign campaign	Re-associates the tasks to another campaign
Archive	Moves to historical archive

warning

Limits

• Max batch: 500 tasks per action (configurable per tenant).
• Re-run in batch re-queues in Redis and can saturate workers: do it outside peak hours.

---

Export

From the filtered listing:

1. Apply the desired filters.
2. Click Export.
3. Choose format:
   • CSV (Excel-ready, with UTF-8 BOM).
   • XLSX (native Excel, with styles).
   • JSON (object per task with full metadata).
4. The file downloads as auditoria-export-YYYY-MM-DD-HHmm.ext.

info

The export respects active filters. If you select rows, only those are exported; otherwise all that match filters are exported.

---

Performance

• Key composite indexes: (operator_id, date), (campaign_id, date) on call_logs.
• Range queries for dates: date >= :start AND date < :end (not DATE(col) = :val, which doesn't use indexes).
• Server-side pagination: LIMIT/OFFSET with a bounded WHERE, avoiding millions of rows on the client.
• Full-text search in PostgreSQL with tsvector, refreshed on transcription INSERT/UPDATE.

---

Troubleshooting

Problem	Cause	Solution
"0 results" with reasonable filters	Incorrect date range time zone	Verify TZ in profile (es-AR = GMT-3)
Pagination always returns the same page	Aggressive browser cache	Hard refresh (Ctrl + Shift + R)
Export > 100k rows takes a long time	Backend processes in background	Wait for the notification when ready
Full-text doesn't find obvious terms	Transcription not yet indexed (async)	Retry in 1–2 min

---

Next steps

• Transcription viewer
• Reports and analytics
• Real-time notifications
• Tutorial: Quality monitoring