Skip to main content

AI Assistants

Ⓔ Enterprise Only
This feature is available exclusively in Countly Enterprise.

Feature Metadata​

FieldValue
FeatureAI Assistants
TypeIn-product AI conversation and guidance
Public endpoint count4
Last updated2026-02-15

Overview​

AI Assistants provides conversational help inside Countly features.
It supports thread management, streaming assistant responses, and message feedback.

Conversations are scoped per app and member, with thread history persisted in MongoDB.

EndpointPath
AI Assistants - Load Thread/o/ai-assistants/load-thread
AI Assistants - Create Thread/i/ai-assistants/create-thread
AI Assistants - Rate Message/i/ai-assistants/rate-message
AI Assistants - Send Message/i/ai-assistants/send-message

Returned Data Fields​

Thread Endpoints (load-thread, create-thread)​

FieldTypeDescription
_idStringThread ID
appIdStringApp ID associated with the thread
memberIdStringOwner member ID
createdOnStringThread creation time
messagesArrayMost recent messages (last 20 on load)

Rate Endpoint (rate-message)​

FieldTypeDescription
okNumberSuccess flag (1)

Streaming Endpoint (send-message)​

Event/PayloadDescription
event: startStream started with assistant message metadata (_id, role, createdOn)
token chunks (data: {"type":"token","content":"..."})Incremental generated text
event: doneFinal complete assistant message payload
event: errorStream-time error payload
event: cancelCancellation notification

Database Collections​

CollectionPurpose
countly.ai_assistants_threadsStores thread metadata and message history
countly.appsValidates app existence for thread-bound operations
countly_drill.drill_metaMetadata source used by assistant agents/tools

Configuration & Settings​

AI Assistants feature config (ai-assistants)​

  • apiKey: Provider API key
  • apiProviderBaseURL: Provider base URL
  • drillAgentEnabled: Enable Drill agent
  • cohortAgentEnabled: Enable Cohort agent
  • funnelAgentEnabled: Enable Funnel agent

Security config (security)​

  • Optional proxy settings:
    • proxy_hostname, proxy_port, proxy_username, proxy_password
  • Optional outbound custom headers:
    • api_additional_headers

Workflows​

1. Start or Resume a Conversation​

  1. Call load-thread with app_id to load/create a member thread.
  2. If needed, call create-thread to reset to a fresh thread.
  3. Use send-message for assistant responses over SSE.

2. Stream a Response​

  1. Send prompt using send-message.
  2. Read token chunks in SSE stream.
  3. Consume done event with final assistant message object.

3. Collect Feedback​

  1. Capture assistant messageId from thread/stream.
  2. Call rate-message with rating value.
  3. Use feedback for quality monitoring workflows.

Limitations​

  • send-message requires configured apiKey and apiProviderBaseURL.
  • Thread load returns last 20 messages.
  • Thread storage is capped at 500 messages.
  • Agent availability depends on enabled feature toggles.

Ⓔ Enterprise​

This feature is part of Countly Enterprise.

Get Access:

Already a Customer? Use support portal if you have any questions.

Last Updated: 2026-02-15