falaai-api — Python SDK for Conversation Intelligence, Speech Analytics & Compliance

Official Python SDK for the FalaAI API — transcribe audio, analyze conversations and audit compliance (COPC CX, ISO 18295-1). Use each API independently or combine them into your own pipeline.

Analyze calls, contact-center recordings, voice notes, chat and email. Get speaker-separated transcripts, summaries, reasons, actions, sentiment and a compliance risk score.

Use any FalaAI API independently

FalaAI is a set of independent REST APIs. You do not need FalaAI Transcription to use FalaAI analysis or compliance auditing. If your application already has a transcript (your own speech-to-text, a chatbot transcript, CRM history or messaging), send that text straight to the analysis APIs.

Your STT                             ->  FalaAI Diagnostic  ->  FalaAI Risk Audit
Telegram voice -> your STT           ->  FalaAI Risk Audit
3CX / Asterisk / Genesys transcript  ->  FalaAI Diagnostic  ->  FalaAI Risk Audit
CRM conversation                     ->  FalaAI Risk Audit

Integrate FalaAI at any point of your pipeline — not only at capture/transcription.

Install

pip install falaai-api

Requires Python 3.11+.

Quickstart

1. Get an API key

Create a free account and copy your fai_ key: https://falaai.action.tec.br/api/auth (or the Dashboard).

2. Set environment variables

FALAAI_BASE_URL=https://api01-falaai.action.tec.br
FALAAI_API_KEY=fai_xxxxxxxx

3. Transcribe a call (audio -> text)

import os
from falaai_api import ApiClient, Configuration, SpeechApi

config = Configuration(host=os.environ["FALAAI_BASE_URL"])
config.access_token = os.environ["FALAAI_API_KEY"]
speech = SpeechApi(ApiClient(configuration=config))

with open("call.mp3", "rb") as fh:
    transcription = speech.create_transcription_v1_audio_transcriptions_post(
        file=("call.mp3", fh.read()),
        model="falaai-transcribe-1",
        language="pt",
        client_reference_id="call_202609271408",
    )

print(transcription.model_dump_json(indent=2))

Expected response (abridged):

{
  "id": "tr-...",
  "object": "transcription",
  "model": "falaai-transcribe-1",
  "language": "por",
  "duration_seconds": 25.0,
  "text": "...",
  "dialog": "Speaker 1: [...] ...",
  "usage": { "audio_seconds": 25.0, "credits_consumed": 25, "processing_ms": 951 }
}

Only need analysis? Skip step 3 and call AnalysisApi with your own transcript (use text for a plain transcript; dialog + audio_events for a diarized one).

4. Analyze or audit an existing transcript (no transcription needed)

import os
from falaai_api import ApiClient, Configuration, AnalysisApi
from falaai_api.models.diagnostic_request import DiagnosticRequest
from falaai_api.models.risk_audit_request import RiskAuditRequest

config = Configuration(host=os.environ["FALAAI_BASE_URL"])
config.access_token = os.environ["FALAAI_API_KEY"]
analysis = AnalysisApi(ApiClient(configuration=config))

transcript = "Good morning, how can I help? I need to cancel my subscription."

# 5 analyses in one call: summary, reason, action, topic, sentiment
diagnostic = analysis.create_diagnostic_v1_analyze_diagnostic_post(
    diagnostic_request=DiagnosticRequest(text=transcript, language="pt-BR", duration_seconds=81.46)
)

# Compliance risk score + violations + auditable report
audit = analysis.create_risk_audit_v1_analyze_risk_audit_post(
    risk_audit_request=RiskAuditRequest(
        text=transcript, language="pt-BR", response_language="pt-BR", duration_seconds=81.46
    )
)

What is FalaAI API?

FalaAI API is an AI conversation-intelligence API for analyzing customer-service, contact-center, sales, messaging and other business conversations. It combines speech-to-text (with speaker diarization and audio-event detection), conversation analysis (summary, contact reason, action taken, topic classification, sentiment) and a compliance/risk audit against COPC CX and ISO 18295-1. Conversation content is processed and discarded (zero-storage).

What can you do with FalaAI?

• Transcribe audio to text with speaker separation and audio events.
• Diagnose a conversation: summary, reason, action taken, topic and sentiment.
• Audit conversations: compliance risk score, detections/violations and an auditable HTML report.
• Track usage, manage webhooks and email alerts, and health/version checks.

Use cases

• Contact center / Quality - audit 100% of conversations instead of a sample.
• Compliance / Legal - auditable evidence for audits and disputes.
• CX / Operations - risk score, sentiment and reason per conversation.
• Sales - analyze sales calls and extract structured outcomes.
• BI / Data - typed JSON ready for your database or analytics stack.

Why FalaAI

• Audit 100%, not a sample - every conversation gets a score, not a random QA sample.
• Auditable by design - a compliance risk score with detections/violations and an HTML report you can hand to an auditor.
• Credits, not tokens - you know exactly what each call costs (per second of audio, per character of text, per conversation). Monthly plans + non-expiring top-up packs.
• Zero-storage - audio/text/results are processed and discarded; only usage/audit records remain. TLS in transit + at rest, per-account isolation (multitenancy), LGPD (you are the controller; Action Tec is the processor).
• 3 native languages (PT/EN/ES) · start free (no card).
• B2B ready - one Enterprise plan serves N clients, no per-user fee (ideal for ISVs).

Integrations

FalaAI is language-independent - integrate at any point of your pipeline (capture, transcription, analysis or audit). Common ecosystems:

These are integration examples, not certified native integrations. Any platform can integrate through REST / cURL - the standard, language-independent path. Native ingestion connectors are on the roadmap.

BYOT - Bring Your Own Transcription

Already have speech-to-text (Whisper, Deepgram, AssemblyAI, your own)? Skip Transcription and send the text straight to Diagnostic and/or Risk Audit. You only pay for what you use.

SDK surface (Python)

Model IDs: falaai-transcribe-1, falaai-diagnostic-1, falaai-risk-audit-1.

Examples

Runnable examples in examples/: health.py, transcribe.py, diagnose.py, audit.py, whatsapp.py.

Authentication

Every request requires Authorization: Bearer fai_<your_key> — except the public endpoints (GET/HEAD /v1/health, GET /api/version). Set the key with Configuration.access_token (or FALAAI_API_KEY).

Error handling

Non-2xx responses raise falaai_api.exceptions.ApiException.

import os
from falaai_api import ApiClient, Configuration, HealthApi
from falaai_api.exceptions import ApiException

config = Configuration(host=os.environ["FALAAI_BASE_URL"])
try:
    print(HealthApi(ApiClient(configuration=config)).health_check().status)
except ApiException as e:
    print(f"HTTP {e.status}: {e.reason}")

Production usage

• Store API keys in environment variables or a secret manager — never hard-code.
• Reuse a single ApiClient across requests.
• Handle ApiException explicitly.
• Set sensible timeouts for long-running requests.

Compatibility

Documentation

• SDK docs (this language): https://actiontecbr.github.io/falaai-api-python/
• API reference (Swagger UI): https://api01-falaai.action.tec.br/docs
• OpenAPI contract: https://api01-falaai.action.tec.br/openapi.json
• Sandbox: https://falaai.action.tec.br/api#playground
• Quickstart: https://falaai.action.tec.br/api/quickstart
• Product page: https://falaai.action.tec.br/api

Versioning

Semantic versioning; the SDK version tracks the API version (1.21.51). See CHANGELOG.md and Releases.

Security

See SECURITY.md. Never commit real keys — use environment variables.

Contributing

See CONTRIBUTING.md.

License

MIT.

Links

• Website: https://falaai.action.tec.br
• API base URL: https://api01-falaai.action.tec.br
• GitHub organization: https://github.com/ActionTecBr
• Other SDKs: Node.js, PHP, Go, Ruby, Java, .NET.

Platform documentation (orientation)

• Odoo — https://www.odoo.com/documentation/19.0/developer/reference/external_api.html
• Rasa — https://rasa.com/docs
• Telegram Bots — https://core.telegram.org/bots/api