O guia definitivo da API do desenvolvedor
A superfície de API servidor a servidor revisada para criar integrações Chirps seguras. Fluxos de faturamento e OAuth do painel, callbacks públicos do widget, webhooks de provedores, cron e rotas internas são excluídos intencionalmente.
1. Autenticação e princípios básicos
URL base: https://chirps.cc
Todos os endpoints de API exigem uma chave de API Chirps no cabeçalho Authorization. Você pode gerar uma chave de API no API Manager na seção Desenvolvedor do seu painel.
Authorization: Bearer <YOUR_CHIRPS_API_KEY>Exemplo de solicitação:
curl -X GET https://chirps.cc/api/billing/usage \
-H "Authorization: Bearer <YOUR_CHIRPS_API_KEY>" \
-H "Content-Type: application/json"Tratamento de erros padrão
| Código de status | Descrição |
|---|---|
200 OK | A solicitação foi bem-sucedida. |
400 Bad Request | A carga JSON falhou na validação. Inclui uma matriz de detalhes. |
401 Unauthorized | Token ausente ou expirado. |
402 Payment Required | O espaço de trabalho atingiu os limites de cobrança. |
403 Forbidden | O usuário não tem as permissões de RBAC necessárias. |
404 Not Found | O recurso solicitado não existe. |
500 Server Error | Erro inesperado da plataforma. |
409 Conflict | The request conflicts with current state, such as a booking slot that is no longer available. |
413 Payload Too Large | The request exceeds the endpoint's documented size limit. |
429 Too Many Requests | The API-key request limit was reached; retry after the current window resets. |
503 Service Unavailable | A required safety or rate-limit dependency is unavailable. Retry with backoff; the request is rejected rather than processed without protection. |
Escopos e privilégio mínimo
Cada chave fica limitada aos escopos selecionados. Use chaves somente leitura sempre que possível e uma chave separada para cada integração. Um escopo de escrita não concede automaticamente o escopo de leitura correspondente.
assistants:read
assistants:write
analytics:read
artifacts:read
artifacts:write
billing:read
bookings:read
bookings:write
chat:write
connectors:read
connectors:write
conversations:read
conversations:write
knowledge:read
knowledge:write
leads:read
leads:write
telephony:read
telephony:write
widget:writeAPI-key rate limit
Each developer key is limited to 300 requests per rolling minute by default. The limit is atomic across all application instances and can be changed by the platform with API_KEY_RATE_LIMIT_PER_MINUTE. A rejected request returns 429; retry only after the window resets.
2. Ciclo de vida do assistente (criação)
GET /api/assistants
Listar assistentes
Busca uma lista de todos os assistentes associados ao espaço de trabalho da chave de API.
Crie um assistente
POST /api/assistants
Cria um novo assistente com todas as configurações básicas.
{
"businessName": "Acme Corp",
"businessUrl": "https://acme.com",
"knowledgeBaseType": "both",
"knowledgeBaseUrl": "https://acme.com/help",
"knowledgeBaseText": "Raw text content of a document...",
"knowledgeBaseFileName": "company_handbook.pdf",
"language": "en",
"isMultiLanguage": false,
"tonePersonality": "professional",
"isCustomPersonality": false,
"supportEmail": "[email protected]",
"maxPages": 5,
"maxDepth": 2,
"welcomeMessageDelay": 2
}| Parâmetro | Tipo | Obrigatório | Descrição |
|---|---|---|---|
businessName | Corda | Sim | Nome da empresa (2 a 100 caracteres). |
businessUrl | Corda | Sim | URL válido do site comercial (máximo de 500 caracteres). |
knowledgeBaseType | Enum | Sim | Um de: "url", "document", "both", "none". |
knowledgeBaseUrl | Corda | Não | Deve ser fornecido se o tipo for "url" ou "both". |
knowledgeBaseText | Corda | Não | Texto bruto se o tipo for "document" ou "both". |
knowledgeBaseFileName | Corda | Não | Nome do arquivo enviado se fornecer texto. |
language | Corda | Sim | Código do idioma (por exemplo, "en", "es"). |
isMultiLanguage | Booleano | Não | Ative a tradução automática. Padrão false. |
tonePersonality | Corda | Sim | Por exemplo, "professional", "friendly". Máximo de 100 caracteres. |
isCustomPersonality | Booleano | Não | Ative o tom personalizado. Padrão false. |
customPersonalityDescription | Corda | Não | Instruções de tom personalizado, se habilitadas (máximo de 2.000 caracteres). |
additionalInstructions | Corda | Não | Instruções básicas de prompt (máximo de 10.000 caracteres). |
supportEmail | Corda | Sim | Endereço de e-mail válido para escalonamento. |
maxPages | Número | Não | Páginas para rastrear. Mínimo 1, Máx. 100. Padrão 5. |
maxDepth | Número | Não | Profundidade de rastreamento. Mínimo 1, Máx. 20. Padrão 2. |
welcomeMessageDelay | Número | Não | Atraso em segundos antes da saudação. Mínimo 0, Máx. 6. Padrão 2. |
Excluir assistente
DELETE /api/assistants/:id
Exclui permanentemente o assistente e exclui em cascata todos os seus nós de conhecimento e histórico. Nenhuma carga útil necessária.
Creation validation and quota safety
| Input | Contract |
|---|---|
knowledgeBaseUrl | For url or both: 1–20 comma-separated public HTTP(S) URLs, with at most 2,000 characters in total. |
knowledgeBaseText | Required for document or both; maximum 2,000,000 characters. |
knowledgeBaseFileName | Optional source label; maximum 255 characters. |
maxPages / maxDepth | Requested values are bounded by both the documented hard limits and the workspace plan. |
Creation allowance is reserved atomically before insertion, so concurrent requests cannot exceed the monthly quota. Database insertion failures refund that reservation. A created assistant starts inactive in developer mode; call the setup endpoint and wait for its successful terminal event before publishing it.
3. Análise e estatísticas do assistente
Imediatamente após a criação e o uso, você pode buscar programaticamente as métricas de desempenho do assistente.
GET /api/analytics/overview?assistantId=uuid&range=30
Obtenha estatísticas de uso, total de mensagens, leads capturados e métricas gerais de engajamento para o assistente especificado. O parâmetro range é opcional (número de dias, mínimo 7, máximo 365, padrão 30).
Resposta Esperada
{
"totalConversations": 1245,
"totalMessages": 8420,
"totalLeads": 156,
"totalBookings": 42,
"satisfactionRate": 94,
"avgMessagesPerSession": 6.7,
"avgSessionsPerDay": 41.5,
"liveNow": 3,
"onlineNow": 12,
"dailySeries": [
{
"date": "2026-07-15",
"conversations": 45,
"messages": 310,
"leads": 5,
"bookings": 2
}
],
"agentStats": [
{
"name": "Jane Doe",
"messagesSent": 120,
"csat": 98
}
]
}| Propriedade | Tipo | Descrição |
|---|---|---|
totalConversations | Número | Número total de sessões de chat gerenciadas pelo assistente. |
totalMessages | Número | Número total de mensagens (usuário + IA) enviadas em todas as conversas. |
totalLeads | Número | Número total de leads únicos (nome, e-mail, telefone) extraídos. |
totalBookings | Número | Número total de compromissos do calendário agendados com sucesso. |
satisfactionRate | Número | Porcentagem (0-100) de avaliações de feedback positivo em relação ao total de avaliações. |
avgMessagesPerSession | Número | Número médio de mensagens trocadas por conversa. |
avgSessionsPerDay | Número | Número médio de conversas por dia dentro do intervalo solicitado. |
liveNow | Número | Número de conversas ativas atualmente controladas por um agente humano. |
onlineNow | Número | Número de usuários conversando ativamente com a IA nos últimos 2 minutos. |
dailySeries | Variedade | Matriz de objetos diários contendo contagens de conversations, messages, leads e bookings por data. |
agentStats | Variedade | Métricas de desempenho para agentes humanos, incluindo messagesSent e csat (pontuação de satisfação do cliente). |
4. Conhecimento e treinamento
Depois que um assistente é criado, você gerencia seu cérebro por meio de "Nós de conhecimento". Cada nó representa um pedaço de dados (um URL rastreado, um documento carregado ou um snippet de texto manual).
Gerenciando nós de conhecimento
GET /api/assistants/:id/nodes - Lista todos os nós de conhecimento deste assistente.
POST /api/assistants/:id/nodes - Injete manualmente um novo bloco de conhecimento.
{
"title": "Return Policy",
"refined_content": "We offer a 30-day return policy for all unused items...",
"type": "text",
"metadata": {}
}PATCH /api/assistants/:id/nodes/:nodeId - Modifique o conteúdo de um nó existente ou alterne seu estado ativo.
{
"title": "Updated Return Policy",
"refined_content": "We offer a 60-day return policy now...",
"is_active": true
}DELETE /api/assistants/:id/nodes/:nodeId - Exclua totalmente um bloco de conhecimento.
POST /api/assistants/:id/nodes/:nodeId/re-crawl - Força o scraper a rastrear novamente o URL associado a este nó e atualizar seu conteúdo. Carga útil: {}.
Treinamento de IA
POST /api/assistants/:id/train - Compile todos os nós de conhecimento is_active e use um orquestrador de IA para sintetizar dinamicamente um novo prompt do Master System. Carga útil: {}.
Upload, crawl, and index consistency
| Operation | Limits and guarantees |
|---|---|
| Manual node | Title: 1–200 characters; content: 1–200,000 characters; metadata: object up to 20,000 serialized characters. |
| File upload | PNG, JPEG, WebP, PDF, DOCX, TXT, or Markdown; maximum 10 MB. Failed extraction creates a disabled repairable node and returns vector_sync_pending: true. |
| Deep crawl | maxPages is a positive integer capped by the plan (100 hard maximum); depth is capped at 20. Only public HTTP(S) destinations are allowed, including redirects and browser subrequests. |
| Enable / disable | Disabled nodes are removed from the vector index and are also filtered at retrieval time. Re-enable or edit an active node to rebuild its vectors. |
vector_sync_pending response means the record was saved but is not yet guaranteed searchable; call the sync endpoint or retry the edit.POST /api/knowledge/urls is retained only to return an explicit 410 Gone; it never stores data. Migrate to POST /api/assistants/:id/nodes/deep-crawl.5. Aparência e comportamento do widget
Every accepted PATCH field, its real default, bounds, plan behavior, and nested shape.
PATCH /api/assistants/:id/widget-config
Send only the fields you want to change. The API rejects unknown keys and invalid nested data instead of silently storing configuration the widget cannot use.
{
"designPreset": "corporate-soft",
"primaryColor": "#1972F5",
"theme": "dark",
"bubbleStyle": "branded-drawer",
"bubbleText": "Talk to our team",
"welcomeMessages": [
{
"text": "Hello! How can I help?",
"delay": 1.5
}
],
"welcomeSuggestions": [
"View pricing",
"Book a demo"
],
"pageVisibility": {
"mode": "exclude",
"paths": [
"/checkout/*"
]
},
"voiceEnabled": true,
"voiceName": "Aoede",
"copilotEnabled": true,
"gdprConsentEnabled": true,
"privacyPolicyUrl": "https://example.com/privacy"
}| Field | Type / bounds | Default | Effect | Plan behavior |
|---|---|---|---|---|
primaryColor | hex color | #000000 | Código de cor hexadecimal (por exemplo, "#000000"). | — |
designPreset | enum | default | Visual system: default, legacy, corporate, corporate-soft, editorial. | — |
position | enum | bottom-right | "bottom-right", "bottom-left", "top-right", "top-left". | — |
theme | enum | light | "light" ou "dark". | — |
fontFamily | enum | default | "default" (Inter) ou "inherit". | — |
bubbleStyle | enum | circle | "circle", "bar", "drawer", "branded-circle", "branded-bar", "branded-drawer". | Branded launcher styles require custom-branding access. |
bubbleText | string (1–80) | Chat with us | Texto a ser exibido próximo ao balão (para estilos de barra/gaveta). | — |
customLogoUrl | HTTP(S) URL or null | null | URL para imagem do logotipo do cabeçalho personalizado. | Requires custom-branding access. Prefer HTTPS or the widget-logo upload endpoint. |
title | string (1–50) | business name | O título no cabeçalho do widget. | — |
headerSubtitle | string (0–100) | We typically reply instantly | Legenda no cabeçalho do widget. | — |
welcomeMessage | string (0–300) | Hi! How can I help you today? | A mensagem de saudação inicial. | — |
welcomeMessages | array (1–5) | one greeting | Ordered greeting objects with text (1–300) and delay in seconds (0–10). | — |
welcomeMessageDelay | number (0–10) | 2 | Atraso em segundos antes da saudação. | — |
welcomeSuggestions | string[] (max 3) | [] | Matriz de strings para chips iniciais de resposta rápida. | — |
inputPlaceholder | string (1–100) | Type your message... | Texto de espaço reservado para a entrada da mensagem. | — |
language | language code | assistant language | Código de idioma do widget (por exemplo, "en", "fr"). | — |
showBranding | boolean | true | Show the Powered by Chirps attribution. | Plans without branding removal always force true. |
voiceEnabled | boolean | true | Ative chamadas de voz. | Plans without voice access always force false. |
voiceModel | string | platform default | Realtime voice model identifier. Normally leave unchanged so Chirps can use the supported platform model. | — |
voiceName | enum | Fenrir | A identidade de voz para respostas de áudio. Veja as vozes disponíveis abaixo. | — |
fileUploadEnabled | boolean | true | Permitir que os usuários carreguem anexos. | Plans without file uploads always force false. |
dictationEnabled | boolean | true | Ative o ditado de voz (fala para texto). | — |
copilotEnabled | boolean | plan default | Allow the DOM-navigation Copilot experience. | Plans without Copilot always force false. |
pageVisibility | object | { mode: all, paths: [] } | Determina em quais páginas o widget aparece (mode, paths). | — |
humanizerEnabled | boolean | true | Apply the conversational response-style pass. | — |
promptInjectionProtectionEnabled | boolean | true | Screen visitor and uploaded-file prompt injection attempts and keep lower-trust content from overriding assistant instructions. | — |
rateLimitEnabled | boolean | true | Ative a limitação de taxa de chat. Padrão verdadeiro. | — |
rateLimitMaxMessages | integer (1–500) | 30 | Máximo de mensagens na janela. Padrão 30. | — |
rateLimitTimeWindow | integer minutes (1–1440) | 60 | Janela de tempo em minutos. Padrão 60. | — |
privacyPolicyUrl | HTTP(S) URL | empty | URL para sua política de privacidade. | — |
gdprConsentEnabled | boolean | false | Show the compact chat privacy notice before the visitor sends data. | — |
gdprDataMaskingEnabled | boolean | false | Mascare PII antes de salvar. Padrão falso. | — |
cookieConsentEnabled | boolean | false | Enable widget cookie preferences and consent UI. | — |
cookieConsentAnalyticsEnabled | boolean | true | Offer optional analytics in cookie preferences. Analytics events remain blocked until the visitor affirmatively consents. | — |
Page visibility matching
Use exact paths such as /pricing, or append /* to include descendants such as /docs/*. include shows only matches; exclude hides matches; all ignores paths. Full URLs are normalized to their pathname by the widget.
Available voice names
Zephyr, Aoede, Kore, Leda, Callirrhoe, Despina, Erinome, Laomedeia, Achernar, Pulcherrima, Gacrux, Algenib, Sulafat, Puck, Charon, Fenrir, Orus, Enceladus, Iapetus, Umbriel, Algieba, Autonoe, Rasalgethi, Alnilam, Schedar, Sadachbia, Achird, Zubenelgenubi, Sadaltager
GET /api/assistants/:id/widget-config without an API key. That read is intentionally origin-bound to the authorized business domain in production mode and returns plan-normalized public configuration. It is not a server-to-server developer endpoint; never treat it as a way to read private assistant settings.6. Artefatos interativos
The complete configuration model for forms, bookings, cards, CSAT, and custom tools.
GET /api/assistants/:id/artifacts · PUT /api/assistants/:id/artifacts · PATCH /api/assistants/:id/artifacts
Artifacts are tool-driven UI. The model calls the enabled tool and Chirps renders it; integrations must send configuration JSON, never {{ARTIFACT: ...}} strings.
{
"form": {
"enabled": true,
"title": "Request a quote",
"description": "Collect the details needed for a quote.",
"triggerRules": "When a visitor explicitly asks for a quote.",
"fields": [
{
"key": "name",
"label": "Full name",
"type": "text",
"required": true,
"placeholder": "Jane Doe"
},
{
"key": "email",
"label": "Work email",
"type": "email",
"required": true
}
],
"submitLabel": "Request quote"
},
"booking": {
"enabled": true,
"title": "Book a consultation",
"description": "Choose an available time.",
"triggerRules": "When a visitor asks to book a meeting.",
"services_enabled": false,
"meetings_enabled": true,
"requiresScheduling": true,
"timezone": "Europe/Paris",
"slotDuration": 30,
"bufferMinutes": 10,
"workingHours": {
"monday": {
"enabled": true,
"start": "09:00",
"end": "17:00"
},
"tuesday": {
"enabled": true,
"start": "09:00",
"end": "17:00"
},
"wednesday": {
"enabled": true,
"start": "09:00",
"end": "17:00"
},
"thursday": {
"enabled": true,
"start": "09:00",
"end": "17:00"
},
"friday": {
"enabled": true,
"start": "09:00",
"end": "17:00"
},
"saturday": {
"enabled": false,
"start": "10:00",
"end": "14:00"
},
"sunday": {
"enabled": false,
"start": "10:00",
"end": "14:00"
}
},
"blockedDates": [
"2026-12-25"
],
"meeting_intake_fields": [
{
"key": "email",
"label": "Email",
"type": "email",
"required": true
}
],
"submitLabel": "Book meeting"
},
"card": {
"enabled": false,
"ctaLabel": "View details",
"ctaUrl": "",
"description": "",
"triggerRules": ""
},
"csat_survey": {
"enabled": true,
"title": "How was your experience?",
"description": "One-click feedback.",
"triggerRules": "Only after the request is resolved."
},
"custom": []
}| Section | Accepted properties | Runtime behavior |
|---|---|---|
form | enabled, title, description, triggerRules, fields, submitLabel | Lead capture and structured intake through show_form. |
booking | enabled, title, description, triggerRules, services_enabled, meetings_enabled, requiresScheduling, timezone, slotDuration, bufferMinutes, workingHours, blockedDates, categories, services, meeting_intake_fields, submitLabel | Meeting or service scheduling through show_booking; supplied values can be prefilled and exact requested slots are verified. |
card | enabled, ctaLabel, ctaUrl, description, triggerRules | Transactional rich-card behavior when verified product data exists. |
csat_survey | enabled, title, description, triggerRules | Thumbs-up/down feedback through show_csat_survey. |
custom | Array of enabled, name, type, description, triggerRules, schema, isCustom | Plan-gated custom tool-backed artifacts. |
Field schema
| Field | Type / bounds | Effect |
|---|---|---|
key | string | Stable machine key used by prefilled_data and requested_fields. |
label | string | Visitor-facing label. |
type | enum | text, email, tel, textarea, number, date, url, dropdown, radio, checkbox, slider, select, or location. |
required | boolean | Blocks submission until a valid value is supplied. |
placeholder | string (max 200) | Optional hint shown inside the control. |
options | string[] | Choices for dropdown, radio, and select controls. |
min, max, step | number | Numeric and slider bounds. |
geo_scope | object | Location boundary: worldwide, continent, country, state, city, or custom; optional value and map bounds. |
Booking service schema
| Field | Type / bounds | Effect |
|---|---|---|
id, name, description | string | Stable service identity and visitor-facing copy. |
duration | minutes | Service duration; the configured slot duration is the fallback. |
category, price | string | Optional filtering and display metadata. |
routing.type | enum | virtual, physical, or phone. |
routing.platform | enum | google_meet, teams, or custom for virtual services. |
routing.link, routing.address | string | Destination appropriate to the selected routing type. |
intake_fields | field[] | Service-specific information requested during booking. |
Validation limits and scheduling rules
| Field | Type / bounds | Effect |
|---|---|---|
fields, meeting_intake_fields, intake_fields | max 50 each | Every field needs a unique machine key, label, and supported type. |
services / categories | max 100 each | Service IDs should remain stable after clients begin storing them. |
slotDuration | integer 5–480 minutes | Length of generated calendar slots. |
bufferMinutes | integer 0–1440 minutes | Unavailable gap inserted around bookings. |
timezone | IANA time-zone name | Business schedule zone, for example Europe/Paris; visitor times are converted for display and calendar invitations. |
workingHours | seven day objects | Each enabled day uses 24-hour HH:mm and must end after it starts. |
blockedDates | YYYY-MM-DD[] (max 366) | Dates unavailable in the business time zone. |
custom | max 20 | Custom tools are plan-gated; names and types are required. |
Tool-call prefill behavior
| Field | Type / bounds | Effect |
|---|---|---|
prefilled_data | object | Values already learned in chat, keyed by configured field key. |
requested_fields | string[] | Show only the configured fields relevant to this intent. |
mode | meeting or service | Select the general meeting flow or a configured service flow. |
selected_service_id | string | Preselect a configured service; never invent an ID. |
requested_categories | string[] | Restrict the visible service list to configured categories. |
preferred_start_time_iso | ISO-8601 | Ask the widget to verify one visitor-requested instant before showing approval. |
artifacts array; streaming responses emit event: artifacts with { artifacts: [...] }. Chirps also keeps compatibility tags in message text for existing widgets and stored history. New clients should consume the structured field/event and treat message text as display copy.7. Conectores e ferramentas
Discover provider schemas at runtime, store credentials securely, then activate only the tools an assistant needs.
Start with GET /api/marketplace/connectors. Its provider definitions are the source of truth for credential fields, configuration fields, available tools, defaults, and required values; this keeps custom and future providers discoverable without a documentation release.
{
"provider": "shopify",
"name": "Shopify",
"credentials": {}
}- Install the workspace connector record with
POST /api/assistants/:id/connectors;credentialsmust be omitted or{}. - Read its exact field schema from
GET /api/marketplace/connectorsand activate it withPOST /api/assistants/:id/connectors/:connectorId/activate. - Send
configValues, optionalenabledTools, and optionalinstructions. Chirps verifies non-custom connectors and discovers the tools before saving the integration. - Rotate secrets by calling the activation endpoint again.
PUTonly renames a named Custom API. - Use
DELETE /api/assistants/:id/connectorswith{ connectorId }to disable it for one assistant. Use the nestedDELETE .../:connectorIdonly to permanently remove it from the workspace and every assistant.
| Field | Type / bounds | Effect |
|---|---|---|
configValues | object (max 100 scalar values) | Exact keys from the marketplace schema. Values are bounded and secrets are write-only. |
enabledTools | string[] (max 200) | Optional allow-list of tools discovered from the verified connector. |
instructions | string (max 10,000) | Assistant-specific constraints for tool use. |
name | string (max 120) | Optional rename during activation, supported only for named Custom APIs. |
8. Integração e implantação
Depois que seu assistente estiver configurado e treinado, você terá duas maneiras principais de implantá-lo:
Opção A: incorporação na Web (frontend)
A maneira mais fácil de integrar o assistente em um site padrão (WordPress, Shopify, React, etc.). O código incorporado é puramente determinístico. Você não precisa de um endpoint para buscá-lo. Basta colocar isso no <head> do seu site:
<script>
window.chirpsConfig = { assistantId: "YOUR_ASSISTANT_ID" };
</script>
<script src="https://chirps.cc/embed.js" async></script>Opção B: Integrações Headless (Backend/CRMs/Bots)
Se você estiver construindo um cliente personalizado (como um bot Discord, uma integração Telegram ou um aplicativo iOS nativo), você NÃO usará o Web Embed. Em vez disso, você interagirá diretamente com o endpoint /api/chat.
POST /api/chat ou /api/chat?stream=true
Importante: chamadas headless e servidor a servidor devem incluir Authorization: Bearer <YOUR_CHIRPS_API_KEY>. Somente o widget instalado no navegador usa o fluxo público vinculado à origem. Nunca coloque uma chave de API no JavaScript do frontend.
{
"assistantId": "123e4567-e89b-12d3-a456-426614174000",
"sessionId": "unique-session-id-for-user",
"conversationId": "optional-uuid",
"message": "Hello, I need help with my billing."
}*(Também suporta multipart/form-data com um campo file para anexos se fileUploadEnabled for verdadeiro).*
| Parâmetro | Tipo | Obrigatório | Descrição |
|---|---|---|---|
assistantId | Corda | Sim | O UUID do assistente. |
sessionId | Corda | Sim | Você gera isso. Um identificador exclusivo do seu próprio sistema para agrupar o histórico de bate-papo (por exemplo, um ID de usuário do Discord, um ID de bate-papo do Telegram ou um UUID aleatório). |
message | Corda | Sim | O conteúdo de texto da mensagem (máximo de 5.000 caracteres). |
conversationId | Corda | Não | O UUID da conversa (retornado na sua primeira solicitação). |
id | Corda | Não | ID de mensagem opcional do lado do cliente para evitar desduplicação. |
Modo 1: resposta JSON padrão (padrão)
Ao enviar uma solicitação POST para /api/chat (sem o parâmetro stream), você aguarda a resposta completa da IA e recebe um objeto JSON padrão.
{
"message": "Here is the information you requested...",
"conversationId": "uuid-of-conversation",
"messageId": "uuid-of-message"
}*(Se is_active_agent: true for retornado, um humano assumiu o controle do bate-papo e message será null.)*
Modo 2: streaming de eventos enviados pelo servidor (?stream=true)
Ao anexar ?stream=true ao URL, você recebe um fluxo de eventos enviados pelo servidor (SSE) em tempo real. Seu código deve escutar estes tipos de eventos específicos:
- `event: token`: Contém
{"text": "chunk"}. Disparado para cada parte do texto de resposta da IA. - `event: status`: Contém
{"text": "Searching knowledge base..."}. Representa estados de execução da ferramenta de back-end. - `event: done`: Contém
{"messageId": "uuid", "conversationId": "uuid", "message": "full text"}. Disparado quando a geração está completamente concluída. - `event: error`: Contém
{"error": "Error message"}. Disparado se ocorrerem limites de taxa ou erros.
Terminais de aquisição ao vivo
GET /api/chat/sessions
Recuperar sessões históricas.
POST /api/chat/toggle-agent
Pause a IA para controle humano. Carga útil: { "sessionId": "...", "isActive": true }
POST /api/chat/agent-message do agente
Envie uma mensagem humana. Carga útil: { "sessionId": "...", "message": "..." }
Complete chat response contract
{
"message": "I've opened the booking calendar for you.",
"messageId": "uuid",
"conversationId": "uuid",
"artifacts": [
{
"type": "booking",
"title": "Book a consultation"
}
],
"sources": []
}A human takeover returns { message: null, conversationId, is_active_agent: true }. A streaming request can therefore return JSON before an SSE stream begins; check the response Content-Type.
Complete SSE event reference
| Event | Payload and behavior |
|---|---|
token | { text } — committed assistant text chunk. |
status | { text } — tool or retrieval progress for user-facing status UI. |
artifacts | { artifacts: [...] } — structured tool-generated UI payloads. |
sources | { sources, tag } — structured citations when enhanced streaming is enabled. |
provisional | { text } — temporary pre-tool text with X-Chirps-Stream-Version: 2. |
reset | {} — discard accumulated provisional text. |
commit | {} — commit accumulated provisional text. |
done | { messageId, conversationId, message } — authoritative final message. |
error | { error } — terminal generation error. |
X-Chirps-Stream-Version: 2 to receive provisional/reset/commit and structured source events. Clients that do not implement those events should omit the header and rely on token, artifacts, done, and error.9. Configurações avançadas do assistente
PATCH /api/assistants/:id
Atualizar configuração principal
Modifique as configurações principais, a personalidade e os limites de raspagem do assistente.
{
"business_name": "Acme Corp",
"business_url": "acme.com",
"business_description": "We sell anvils.",
"system_prompt": "You are a helpful assistant.",
"language": "en",
"tone_personality": "Friendly",
"support_email": "[email protected]",
"additional_instructions": "Never mention the Roadrunner.",
"is_multi_language": true,
"is_dev_mode": false,
"is_active": true,
"max_pages": 15,
"max_depth": 2,
"widget_config_merge": {
"rateLimitEnabled": true
}
}*Nota: Todos os campos são opcionais. Envie apenas os campos que deseja atualizar.*
Excluir assistente
DELETE /api/assistants/:id
Exclui permanentemente o assistente e distribui a exclusão em cascata para todos os nós de conhecimento, conversas e mensagens. Se o assistente nunca tiver sido totalmente configurado, isso reembolsará a cota mensal de criação.
10. Telefonia e Voz
GET /api/phone-numbers - Lista os números disponíveis.
POST /api/phone-numbers/:subscriptionId/assign assigns an already-purchased Chirps phone subscription to an assistant. Send { "assistantId": "uuid" }; send { "assistantId": null } to unassign it. This endpoint does not purchase or provision a new number.
POST /api/assistants/:id/outbound - Aciona chamada de saída. Carga útil: { "phoneNumber": "+123...", "callReason": "Confirm the requested appointment", "conversationId": "optional-uuid" }
11. Extração de leads
GET /api/leads?assistantId=uuid&page=1&limit=50&status=new - Busque leads estruturados e paginados que a IA extraiu durante as conversas.
GET /api/leads?id=leadId - Obtenha detalhes de um único lead, incluindo a transcrição completa da conversa.
Public lead response
Lead objects contain id, assistant_id, conversation_id, session_id, name, email, phone, status, notes, artifact_id, ai_score, ai_summary, form_data, created_at, and the assistant business name. Internal voice-confirmation/idempotency identifiers are never returned.
12. Reservas e calendário
GET /api/bookings?assistantId=uuid&start=ISO-8601&end=ISO-8601 - Recupera compromissos do calendário agendados pelo assistente, opcionalmente filtrados por data.
GET /api/bookings?id=bookingId - Obtenha detalhes de uma única reserva, incluindo a transcrição da conversa.
DELETE /api/bookings/cancel?id=bookingId - Cancela uma reserva ativa. Adicione &silent=true somente quando a integração precisar suprimir intencionalmente o e-mail de cancelamento de contingência do Chirps.
POST /api/bookings/reschedule - Ajuste uma reserva. Carga útil: { "bookingId": "uuid", "newStartTime": "ISO-8601", "newEndTime": "ISO-8601" }
Public booking response and time zones
Booking objects contain the public appointment, guest, status, notes, and timestamp fields plus business_timezone. Times are ISO-8601 instants; format them in the viewer's time zone while preserving the returned business zone for schedule context. Internal confirmation IDs, provider calendar-event IDs, and database range columns are omitted.
13. Uso de conta e faturamento
GET /api/billing/usage
Recupere o uso do ciclo de cobrança do workspace, os limites do plano de assinatura e a contagem total de assistentes. Observação: Ao usar uma chave de API, a solicitação tem automaticamente como escopo o espaço de trabalho da chave de API; você não precisa fornecer um teamId.
Resposta (`200 OK`):
{
"planName": "Pro",
"periodStr": "2023-10-01",
"assistantCount": 3,
"usage": {
"message_count": 1450,
"harvest_count": 12,
"recrawl_count": 5,
"train_count": 2,
"assistants_created_count": 1,
"voice_minutes_used": 0,
"sms_count": 0,
"email_count": 0,
"whatsapp_count": 0
},
"limits": {
"messages": 5000,
"assistants": 5,
"training": 50,
"recrawls": 20
},
"workspaceOwnerId": "uuid"
}14. Catálogo completo de endpoints
The reviewed, API-key-compatible surface. Paths and JSON property names are never translated.
What belongs in the public developer API
These routes accept a scoped Chirps API key and enforce its workspace boundary. Dashboard-only billing/OAuth routes, browser widget callbacks, provider webhooks, cron jobs, and internal MCP routes are intentionally excluded.
id for duplicate protection. Do not automatically retry other write operations unless the response is clearly safe to retry. 402 means a plan quota was reached; 429 means throttled; 5xx may be transient.Assistants
| Method | Path | Scope | Request | Successful result |
|---|---|---|---|---|
GET | /api/assistants | assistants:read | — | Assistant array; credential fields omitted |
POST | /api/assistants | assistants:write | Assistant creation JSON | Created assistant and setup identifiers |
GET | /api/assistants/:id | assistants:read | — | Assistant configuration; secrets omitted |
PATCH | /api/assistants/:id | assistants:write | Supported fields only | Updated assistant |
DELETE | /api/assistants/:id | assistants:write | — | Deletion confirmation |
GET | /api/assistants/:id/setup | assistants:read | — | Setup and crawl status |
POST | /api/assistants/:id/setup | assistants:write | Empty body | SSE setup progress stream |
GET | /api/analytics/overview?assistantId=:id&range=30 | analytics:read | Query: range 7–365 | Engagement, lead, booking and agent metrics |
Knowledge and training
| Method | Path | Scope | Request | Successful result |
|---|---|---|---|---|
GET | /api/assistants/:id/nodes | knowledge:read | — | Knowledge node array |
POST | /api/assistants/:id/nodes | knowledge:write | { title, refined_content, type?, metadata? } | Stored and vectorized node |
PATCH | /api/assistants/:id/nodes/:nodeId | knowledge:write | { title?, refined_content?, is_active? } | Updated and re-vectorized node |
DELETE | /api/assistants/:id/nodes/:nodeId | knowledge:write | — | Node deletion confirmation |
POST | /api/assistants/:id/nodes/:nodeId/re-crawl | knowledge:write | Empty body | Refreshed URL node |
POST | /api/assistants/:id/nodes/upload | knowledge:write | multipart/form-data: file, title? | Extracted knowledge node |
POST | /api/assistants/:id/nodes/deep-crawl | knowledge:write | { maxPages? } | SSE crawl progress stream |
POST | /api/assistants/:id/nodes/sync | knowledge:write | Empty body | Vector store synchronization result |
POST | /api/assistants/:id/train | knowledge:write | Empty body | Compiled assistant prompt |
POST | /api/assistants/:id/harvest | knowledge:write | Empty body | Conversation insights added to knowledge |
POST | /api/knowledge/urls | knowledge:write | Deprecated | Always returns 410 with the assistant-scoped deep-crawl replacement path |
Widget, artifacts and connectors
| Method | Path | Scope | Request | Successful result |
|---|---|---|---|---|
PATCH | /api/assistants/:id/widget-config | widget:write | Partial widget configuration | Normalized widget configuration |
POST | /api/assistants/:id/widget-logo | widget:write | multipart/form-data: file (PNG/JPEG/WebP, max 2 MB) | Public logo URL |
GET | /api/assistants/:id/artifacts | artifacts:read | — | Form, booking, card and CSAT configuration |
PUT | /api/assistants/:id/artifacts | artifacts:write | Complete artifact configuration | Normalized configuration |
PATCH | /api/assistants/:id/artifacts | artifacts:write | Partial artifact configuration | Merged normalized configuration |
GET | /api/marketplace/connectors | connectors:read | — | Installable connector definitions; no credentials |
GET | /api/assistants/:id/connectors | connectors:read | — | Workspace connectors and assistant integrations |
POST | /api/assistants/:id/connectors | connectors:write | { provider, name, credentials?: {} } | Installed connector record; no operational access yet |
PUT | /api/assistants/:id/connectors/:connectorId | connectors:write | { name } (named custom APIs only) | Renamed connector; credentials omitted |
DELETE | /api/assistants/:id/connectors | connectors:write | { connectorId } | Connector disabled for this assistant only |
DELETE | /api/assistants/:id/connectors/:connectorId | connectors:write | — | Workspace connector and all assistant integrations deleted |
POST | /api/assistants/:id/connectors/:connectorId/activate | connectors:write | { configValues, enabledTools?, instructions?, name? } | Activated integration and tool list |
Conversations, leads and bookings
| Method | Path | Scope | Request | Successful result |
|---|---|---|---|---|
POST | /api/chat or /api/chat?stream=true | chat:write | { assistantId, sessionId, message, conversationId?, id? } | JSON reply with artifacts/sources or SSE token/artifacts/status/done events |
GET | /api/chat/sessions?assistantId=:id | conversations:read | — | 20 most recent session summaries |
GET | /api/assistants/:id/conversations?limit=20&status=active | conversations:read | Query: limit 1–50, status optional | Conversation summaries |
DELETE | /api/assistants/:id/conversations | conversations:write | { conversationIds: [uuid] } (max 100) | Deletion confirmation |
POST | /api/chat/toggle-agent | conversations:write | { sessionId, isActive } | Takeover state |
POST | /api/chat/agent-message | conversations:write | { sessionId, message } | Stored human-agent message |
POST | /api/chat/typing | conversations:write | { assistantId, sessionId, isTyping } | Typing-state confirmation |
POST | /api/chat/followup-draft | conversations:write | { conversationId|leadId, assistantId?, channel? } | AI-generated draft; consumes AI-message usage |
POST | /api/chat/conversation-followup | conversations:write | { conversationId|leadId, channel, recipient, subject?, message } | Dispatched follow-up result |
GET | /api/leads | leads:read | Query: id, or assistantId with page, limit, status | Lead detail with transcript or paginated leads; internal confirmation identifiers omitted |
POST | /api/leads | artifacts:write | Artifact submission JSON | Captured lead/booking result |
PATCH | /api/leads | leads:write | { leadId, status?, notes? } | Updated lead |
DELETE | /api/leads | leads:write | { leadId } or { leadIds: [uuid] } | Deletion count |
GET | /api/bookings | bookings:read | Query: id, or assistantId with start and end | Booking detail or booking array; provider and idempotency identifiers omitted |
DELETE | /api/bookings/cancel?id=:id | bookings:write | Query: silent=true optional | Cancellation and calendar-sync state |
POST | /api/bookings/reschedule | bookings:write | { bookingId, newStartTime, newEndTime } | Updated booking or available alternatives |
Telephony and account
| Method | Path | Scope | Request | Successful result |
|---|---|---|---|---|
GET | /api/phone-numbers | telephony:read | — | Numbers and assignable assistants in this workspace |
POST | /api/phone-numbers/:subscriptionId/assign | telephony:write | { assistantId } or { assistantId: null } | Routing update confirmation |
POST | /api/assistants/:id/outbound | telephony:write | { phoneNumber, callReason, conversationId? } | Provider call SID and conversation ID |
GET | /api/billing/usage | billing:read | — | Workspace plan, limits and current usage |
GET | /api/metadata?url=:httpsUrl | artifacts:read | Query: public HTTP(S) URL | Bounded page title, description and image metadata |
Inicie seu assistente
em 10 minutos.
Cadastre-se, treine em seu site, personalize alertas e incorpore em seu site.