> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pyai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Build your own Gong: conversation intelligence on PyAI

> Build a Gong-style conversation intelligence product on PyAI: Hear transcribes every call with speaker separation, Recap writes summaries and action items, and Trace scores compliance, all usage-based APIs.

You can build the core of a Gong-style conversation intelligence product on
three PyAI APIs: **Hear** batch jobs transcribe every call with speaker
separation, **Recap** turns each finished call into a summary with action items
and structured fields, and **Trace** scores each call against compliance rule
packs. You keep the transcript, the metrics, and the customer relationship in
your own product, and you pay per call instead of per seat.

Gong is the better buy when you want a finished revenue-intelligence suite:
deal boards, coaching workflows, forecast rollups, a polished UI your sales
team lives in. Building on PyAI makes sense when conversation data is a
*feature inside your product* (a CRM, a dialer, an agency platform, a vertical
tool) rather than a tool your team logs into.

## What you're building

```mermaid theme={null}
flowchart LR
  rec[Call recordings or live calls] --> hear[Hear batch jobs: diarized transcript]
  hear --> metrics[Your metrics: talk ratio, keywords, trackers]
  hear --> recap[Recap: summary + action items]
  hear --> trace[Trace: compliance scorecard]
  metrics --> app[(Your product)]
  recap --> app
  trace --> app
```

Every call becomes three artifacts you own: a diarized transcript (with SRT/VTT
if you want subtitles), a structured summary, and a compliance scorecard with a
tamper-evident audit hash.

## The PyAI stack for this build

| Piece                                      | PyAI product                               | Endpoints                                    | Scope                                 |
| ------------------------------------------ | ------------------------------------------ | -------------------------------------------- | ------------------------------------- |
| Transcription with speaker separation      | Hear (batch jobs, discounted vs. realtime) | `POST /v1/transcription/jobs`                | `hear:transcribe` + `transcribe:jobs` |
| Summaries, action items, structured fields | Recap                                      | `/v1/recap/config`, `/v1/recap/calls`        | `recap:configure`, `recap:read`       |
| Compliance scorecards per call             | Trace                                      | `/v1/trace/config`, `/v1/trace/interactions` | `trace:configure`, `trace:read`       |
| Calls that already run on PyAI             | Omni call records                          | `/v1/omni/calls`                             | `omni:read`                           |

## Build order

<Steps>
  <Step title="Transcribe your call archive">
    Submit recordings to `POST /v1/transcription/jobs` and receive results by
    signed webhook or polling. Stereo telephony recordings (one party per
    channel) give the cleanest speaker split; model-based diarization covers
    mono. Batch is billed at a discounted rate versus realtime, which is what
    makes processing a whole archive practical. The full, runnable pipeline is
    in the [conversation-intelligence guide](/guides/conversation-intelligence).
  </Step>

  <Step title="Compute the metrics your users ask for">
    Talk ratio, longest monologue, keyword and competitor mentions, question
    rate. These fall out of the diarized transcript with a few lines of code,
    and the conversation-intelligence guide shows the exact computations.
  </Step>

  <Step title="Turn on Recap for summaries">
    Recap fires on call end: summary, action items, and structured fields you
    define, with no bot to invite to anything. For custom extraction (deal
    stage, qualification fields, your own schema), follow the
    [post-call extraction guide](/guides/post-call-extraction).
  </Step>

  <Step title="Add Trace if your users are regulated">
    Sales calls in healthcare, finance, or collections need disclosure and
    PII handling. Trace scores every call against TCPA, HIPAA, PII, and
    brand-voice packs and gives you an exposure dashboard. See the
    [Trace guide](/guides/trace-guardrails).
  </Step>
</Steps>

## When to buy Gong instead

Buy Gong (or a similar suite) when the buyer is a sales leader who wants
coaching dashboards, deal inspection, and forecasting out of the box, and when
per-seat pricing is acceptable. Build on PyAI when you are embedding call
intelligence in a product you sell, when you need the raw transcripts and
metrics in your own database, or when per-seat pricing breaks your margins at
scale.

## FAQ

### Is PyAI a Gong alternative?

PyAI is not a finished revenue-intelligence suite; it is the API layer you
build one on. If you want the finished suite, buy it. If conversation
intelligence is a feature inside your product, PyAI gives you the
transcription, summarization, and compliance scoring primitives.

### Can I analyze calls that did not happen on PyAI?

Yes. Hear batch jobs accept recordings you upload or point to by URL, so calls
from any carrier, dialer, or conferencing tool can be transcribed and analyzed.

### Do I need my own LLM?

Not for summaries and action items: Recap produces those natively. For fully
custom metrics and free-form extraction you define, the pipeline hands you a
clean diarized transcript and you run your own model over it.

### What does it cost to build this on PyAI?

Every piece is usage-based (per minute of audio, per call), with no seats and
no platform fee. Current rates are on the
[pricing page](https://pyai.com/pricing).

## Start building

[Get an instant sandbox key](/quickstart) (no signup) and run the full pipeline
in the [conversation-intelligence guide](/guides/conversation-intelligence),
or scaffold it in one command:

```bash theme={null}
npm create pyai-app@latest recap-call-intelligence
```