Analyze your first calls

You will create an analyzer, upload two recordings, run the analysis and read the results. You need an API key (see Authentication) and two audio files, one.mp3 and two.mp3.

export BASE="https://api.60db.ai/deep-analytics"
export KEY="sk_your_api_key"

The key must have the Call analysis permission. New keys have it. For a key created earlier, open Developers → API keys, edit the key and switch on Call analysis. Without it every call returns 403 DA_API_KEY_SCOPE_MISSING. See Permissions.

The examples use jq to read IDs from responses. Every response is { "success": true, "data": ... }.

Create an analyzer

AGENT=$(curl -s -X POST "$BASE/analysis-agents" \
  -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"name":"Refund policy check","instructions":"Flag any call where the agent gives wrong refund information."}' \
  | jq -r .data.id)

Create a batch for your files

The CSV lists the files by file_name. Audio is not sent here.

BATCH=$(curl -s -X POST "$BASE/analysis-agents/$AGENT/batches" \
  -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"source_kind":"direct_upload","csv_text":"call_id,file_name\nA-1,one.mp3\nA-2,two.mp3\n"}' \
  | jq -r .data.id)

The response shows total, valid and invalid rows.

Find the row for each file

curl -s "$BASE/imports/batches/$BATCH/items?state=draft" \
  -H "Authorization: Bearer $KEY" | jq '.data[] | {id, file_name}'

Save the row id of one.mp3 as ITEM. Repeat steps 4 to 6 for each file.

Get an upload slot

SLOT=$(curl -s -X POST "$BASE/imports/batches/$BATCH/items/$ITEM/upload-slot" \
  -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d "{\"size_bytes\": $(wc -c < one.mp3 | tr -d ' '), \"content_type\": \"audio/mpeg\"}")
UPLOAD_ID=$(echo "$SLOT" | jq -r .data.upload_id)
UPLOAD_URL=$(echo "$SLOT" | jq -r .data.upload.url)

Upload the file to storage

Send every entry of upload.fields first, then the file as the last field, file. Do not send your API key to storage.

ARGS=()
while IFS= read -r kv; do ARGS+=(--form-string "$kv"); done < <(
  echo "$SLOT" | jq -r '.data.upload.fields | to_entries[] | "\(.key)=\(.value)"')

curl -s -X POST "$UPLOAD_URL" "${ARGS[@]}" -F "[email protected];type=audio/mpeg"

A 2xx reply means success. Storage answers 403 when the slot expired and 400 when the size or type does not match. The same flow in code is in the JavaScript section.

Confirm the upload

curl -s -X POST "$BASE/imports/batches/$BATCH/items/$ITEM/asset-receipt" \
  -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d "{\"receipt_ref\": \"$UPLOAD_ID\"}"

60db checks the stored file itself. Wrong type, empty or unreadable files are rejected with an ASSET_* code.

Analyze

The confirm_count must equal the number of recordings in status ready.

N=$(curl -s "$BASE/analysis-agents/$AGENT/recordings?status=ready" \
  -H "Authorization: Bearer $KEY" | jq .data.total)

curl -s -X POST "$BASE/analysis-agents/$AGENT/analyze" \
  -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d "{\"confirm_count\": $N}"

This uses credits. The response includes estimated_credits_usd.

Poll until done

curl -s "$BASE/analysis-agents/$AGENT/recordings" \
  -H "Authorization: Bearer $KEY" | jq '.data.items[] | {title, status, verdict, problem_count}'

Repeat every few seconds until each status is done or failed.

Read the results

# Summary of problems
curl -s "$BASE/analysis-agents/$AGENT/problems" -H "Authorization: Bearer $KEY"

# Full analysis of one call: use a call_id from the recordings list
curl -s "$BASE/calls/$CALL_ID/analysis" -H "Authorization: Bearer $KEY"

The same in JavaScript

const BASE = 'https://api.60db.ai/deep-analytics';
const headers = { Authorization: `Bearer ${process.env.SIXTYDB_API_KEY}`, 'Content-Type': 'application/json' };

async function api(method, path, body) {
  const res = await fetch(BASE + path, { method, headers, body: body && JSON.stringify(body) });
  const json = await res.json();
  if (!json.success) throw new Error(`${json.code}: ${json.message}`);
  return json.data;
}

// 1. Analyzer and batch
const analyzer = await api('POST', '/analysis-agents', {
  name: 'Refund policy check',
  instructions: 'Flag any call where the agent gives wrong refund information.',
});
const batch = await api('POST', `/analysis-agents/${analyzer.id}/batches`, {
  source_kind: 'direct_upload',
  csv_text: 'call_id,file_name\nA-1,one.mp3\n',
});

// 2. Upload one file
const rows = await fetch(`${BASE}/imports/batches/${batch.id}/items?state=draft`, { headers }).then((r) => r.json());
const row = rows.data.find((r) => r.file_name === 'one.mp3');
const file = await (await import('node:fs/promises')).readFile('one.mp3');

const slot = await api('POST', `/imports/batches/${batch.id}/items/${row.id}/upload-slot`, {
  size_bytes: file.length,
  content_type: 'audio/mpeg',
});
const form = new FormData();
for (const [k, v] of Object.entries(slot.upload.fields)) form.append(k, v);
form.append('file', new Blob([file], { type: 'audio/mpeg' }), 'one.mp3'); // file must be last
const up = await fetch(slot.upload.url, { method: 'POST', body: form }); // no Authorization header
if (!up.ok) throw new Error(`Storage upload failed: ${up.status}`);

await api('POST', `/imports/batches/${batch.id}/items/${row.id}/asset-receipt`, { receipt_ref: slot.upload_id });

// 3. Analyze, poll, read
const ready = await api('GET', `/analysis-agents/${analyzer.id}/recordings?status=ready`);
await api('POST', `/analysis-agents/${analyzer.id}/analyze`, { confirm_count: ready.total });

let items;
do {
  await new Promise((r) => setTimeout(r, 5000));
  ({ items } = await api('GET', `/analysis-agents/${analyzer.id}/recordings`));
} while (items.some((i) => !['done', 'failed', 'cancelled'].includes(i.status)));

console.log(await api('GET', `/analysis-agents/${analyzer.id}/problems`));

Next steps

Watch your AI agents or dialer numbers so new calls are analyzed automatically: Set sources
Add recordings by URL instead of upload: Create batch with csv_urls, then Fetch recordings
Improve your AI agent from a bad call: Suggest agent fix
All limits and errors: Call Analysis API

Get Started

SDKs

Core Features

Analyze your first calls

The same in JavaScript

Next steps