AI API reference

Access all Quant AI features programmatically via the v3 REST API. This page provides an overview of available endpoints organised by category.

Base URL

All AI endpoints share the following base URL, on the dashboard host of your platform.

https://dashboard.quantcdn.io/api/v3/organizations/{org}/ai     # QuantCDN
https://dash.quantgov.cloud/api/v3/organizations/{org}/ai       # QuantGov

Replace {org} with your organisation machine name. The path accepts both organizations and organisations. The AI API is not served from the content API host (api.quantcdn.io).

Authentication

Send a dashboard API token as a Bearer token in the Authorization header. Create the token in the dashboard under your profile, API Tokens. A project token from the Integrations page (the Quant-Token used by the content API) does not work here and returns 401.

curl -H "Authorization: Bearer YOUR_API_TOKEN" \
  https://dashboard.quantcdn.io/api/v3/organizations/{org}/ai/models

Note

The token acts with the permissions of its user. See Overview for the AI permissions.

Models

Method	Endpoint	Description
GET	/models	List available AI models
GET	/models/{modelId}	Get model details and capabilities

Filter models by capability with the feature query parameter: chat, embeddings, vision, streaming, or all (for example /models?feature=chat). Use a model ID from this list in your requests. On QuantGov the list contains Australian-resident models only.

Chat inference

Method	Endpoint	Description
POST	/chat	Send a chat message (buffered response)
POST	/chat/stream	Send a chat message (streaming SSE response)
GET	/chat/executions/{id}	Poll for async execution result
POST	/chat/callback	Submit client tool results to a durable execution in waiting_callback

Example chat request:

curl -X POST https://dashboard.quantcdn.io/api/v3/organizations/{org}/ai/chat \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "modelId": "MODEL_ID_FROM_MODELS_LIST",
    "messages": [{"role": "user", "content": "Hello"}],
    "maxTokens": 1024
  }'

There is no public WebSocket endpoint. Use /chat/stream for streaming.

Field	Description
modelId	Required. A model ID from GET /models
messages	Required. Each message has a role (user, assistant, system) and content (a string, or text, image, video, and document blocks with base64 data; 25 MB per request)
systemPrompt, maxTokens, temperature, topP, stopSequences	Optional generation settings
responseFormat	{"type": "json", "jsonSchema": {...}} for structured JSON output on models that support it
toolConfig	Tool calling: tools[].toolSpec with name, description, and inputSchema.json (a JSON Schema). When the model asks for a tool, the response contains response.toolUse with toolUseId, name, and input. Set autoExecute: true with an allowedTools list to let the platform run built-in tools
sessionId	Continue a session; omit for a stateless call
async	true for durable execution

Embeddings

Method	Endpoint	Description
POST	/embeddings	Generate text embeddings. input is a string or an array of up to 100 strings

Image generation

Image generation is not available.

Sessions

Method	Endpoint	Description
GET	/sessions	List sessions
POST	/sessions	Create a session
GET	/sessions/{id}	Get session details
PUT	/sessions/{id}	Update a session
DELETE	/sessions/{id}	Delete a session
PUT	/sessions/{id}/extend	Extend session TTL

A session keeps the conversation on the server. Create it with a required userId and optional sessionGroup, metadata, and expirationMinutes (default 60, 5 to 1440), then pass its sessionId in chat requests. Extend it with additionalMinutes (up to 1440). List sessions by userId or sessionGroup.

Agents

Method	Endpoint	Description
GET	/agents	List agents
POST	/agents	Create an agent
GET	/agents/{id}	Get agent details
PUT	/agents/{id}	Update an agent
DELETE	/agents/{id}	Delete an agent
POST	/agents/{id}/chat	Chat with an agent
GET / PUT / DELETE	/agents/{id}/overlay	Organisation overlay on a pre-built agent. PUT replaces the whole overlay

Agent chat takes message and optional sessionId, userId, stream (SSE), and async. Without a sessionId a new session is created. The agent runs its own tools.

Skills

Method	Endpoint	Description
GET / POST	/skills	List or create skills
GET / PUT / DELETE	/skills/{id}	Read, update, or delete a skill
POST	/skills/import	Import a skill from GitHub
POST	/skills/{id}/sync	Sync a skill from its source
GET	/skills/collections	List skill collections
POST	/skills/import-collection	Import a directory of skills as a namespaced collection
POST	/skills/collections/{namespace}/sync	Sync a collection
DELETE	/skills/collections/{namespace}	Delete a collection and its skills

Governance

Method	Endpoint	Description
GET / PUT	/governance	Read or replace the governance configuration
GET	/governance/spend	Current spend
GET / POST	/filter-policies	List or create filter policies
GET / PUT / DELETE	/filter-policies/{id}	Read, update, or delete a filter policy

Custom tools

Method	Endpoint	Description
GET	/custom-tools	List custom tools
POST	/custom-tools	Register a custom tool
DELETE	/custom-tools/{id}	Delete a custom tool
GET	/tools/executions/{id}	Get tool execution status

Vector database

Method	Endpoint	Description
GET	/vector-db/collections	List collections
POST	/vector-db/collections	Create a collection
DELETE	/vector-db/collections/{id}	Delete a collection
POST	/vector-db/collections/{id}/documents	Upload documents
DELETE	/vector-db/collections/{id}/documents	Delete documents
POST	/vector-db/collections/{id}/query	Search a collection

Tasks

Method	Endpoint	Description
GET	/tasks	List tasks
POST	/tasks	Create a task
GET	/tasks/{id}	Get task details
PUT	/tasks/{id}	Update a task
DELETE	/tasks/{id}	Delete a task
GET	/tasks/{id}/dependency-graph	Get task dependency graph

Orchestrations

Method	Endpoint	Description
GET	/orchestrations	List orchestrations
POST	/orchestrations	Create an orchestration
DELETE	/orchestrations/{id}	Delete an orchestration
POST	/orchestrations/{id}/start	Start an orchestration
POST	/orchestrations/{id}/pause	Pause an orchestration
POST	/orchestrations/{id}/resume	Resume an orchestration
POST	/orchestrations/{id}/cancel	Cancel an orchestration
GET	/orchestrations/{id}/batches	List orchestration batches

Files

Method	Endpoint	Description
GET	/files	List files
POST	/files	Upload a file
DELETE	/files/{id}	Delete a file

Usage

Method	Endpoint	Description
GET	/usage	Organisation usage breakdown (needs the subscription permission)
GET	/usage/me	Your own usage

Streaming responses

The /chat/stream endpoint returns Server-Sent Events (SSE). A start event opens the stream, each content event carries a piece of text in delta, and a done event closes it with token usage.

id: chunk-0
event: start
data: {"requestId":"abc123","model":"...","streaming":true}

id: chunk-1
event: content
data: {"delta":"Hello","complete":false}

id: chunk-2
event: content
data: {"delta":" there!","complete":false}

id: chunk-3
event: done
data: {"complete":true,"usage":{"inputTokens":8,"outputTokens":15,"totalTokens":23}}

An error that happens after the stream has started (for example a spend limit) arrives as an event: error with a JSON data object that has an error message. The HTTP status of the response stays 200, so read the event type of every event. Treat a stream that ends without a done event as failed.

Tip

Streaming is recommended for interactive chat interfaces where you want to display responses as they are generated, rather than waiting for the full response.

Durable execution

For long-running operations, set async: true in your chat request. The response is 202 with a requestId and a pollUrl. Poll GET /chat/executions/{requestId} until the status is complete (other statuses: pending, running, waiting_callback, failed), then read the result. Start polling at 1 second and back off to 30 seconds.

# Start an async chat request
curl -X POST https://dashboard.quantcdn.io/api/v3/organizations/{org}/ai/chat \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "modelId": "MODEL_ID_FROM_MODELS_LIST",
    "messages": [{"role": "user", "content": "Analyse this dataset..."}],
    "async": true
  }'

# Poll for the result
curl https://dashboard.quantcdn.io/api/v3/organizations/{org}/ai/chat/executions/{requestId} \
  -H "Authorization: Bearer YOUR_API_TOKEN"

Note

Durable executions are ideal for operations that may take longer than typical HTTP timeout limits. The execution persists on the server and can be retrieved at any time.

When the status is waiting_callback, the response lists pendingTools and a callbackId. Run those tools in your application and send the results to POST /chat/callback with {"callbackId": "...", "toolResults": [{"toolUseId": "...", "result": {...}}]}. Each callbackId can be used once.

Error responses

Status	Meaning
401	Missing or invalid token, or a project token instead of a dashboard API token
403	The token’s user lacks permission; AI services are disabled for the organisation; the model is not approved by your governance policy; or (QuantGov) the model is not available, with the error model_not_available_in_region
429	A spend limit was reached

Workflows

Workflows have no REST endpoints. Build them in the dashboard and start them with a webhook or schedule trigger.