HALO Integration

When a HALO profile is linked to an AI Cloud project, the two platforms operate as a unified system. AI Cloud handles structured content, recognition, and routing; HALO handles agentic reasoning, tool execution, and dynamic responses. This page maps every point where the two platforms exchange data or share behaviour.

NLU Pipeline

When HALO is enabled, it participates as a processing block in the NLU pipeline alongside AI Cloud's Standard Recognition engines. You control the order — HALO can be placed before or after Standard Recognition — and you can disable HALO from the pipeline entirely without removing the integration.

• HALO enabled — interactions can be routed to HALO agents and tools for agentic handling.

• HALO disabled — AI Cloud processes interactions using Intent Recognition and Rule-based Recognition only. Useful for performance-sensitive bots or clients restricted from using generative AI.

→ Hybrid NLU

Variable Sync

AI Cloud passes two types of variables into HALO on every interaction, and HALO can write values back.

Context Variables

Static key-value pairs with a fixed set of allowed values. AI Cloud sends these to HALO on each request. If HALO has a context variable with the same name, it assumes that value. Context variables are defined in AI Cloud — the HALO counterpart should mirror the same allowed values for consistent behaviour.

→ CAIC Context Variables

Conversation Variables

Dynamic values that persist across turns. AI Cloud sends these to HALO on each interaction. When HALO changes a context variable via an agent or tool, the updated value is written back to the matching conversation variable in AI Cloud — creating a live two-way sync for the duration of the conversation.

→ Conversation Variables

Metadata

Metadata fields are key-value pairs returned to the client alongside every AI Cloud output. When a HALO tool sets a metadata value during execution, AI Cloud maps it back into the response payload. This is the primary mechanism for HALO tools to signal structured actions to the client — for example, triggering a live chat handover or a page redirect.

→ Metadata

Content References

HALO agents and tools can be referenced directly within AI Cloud content — in Articles, Events, and Dialogs. There are two mechanisms:

• Output-level — an agent or tool is added as an output on a content item. When that item matches, AI Cloud hands off to the agent (which holds the conversation until complete) or invokes the tool (single-shot, then NLP flow resumes).

• Dialog node — the HALO Tool Call node invokes a tool mid-flow within a running dialog. The dialog continues executing after the tool completes. This is analogous to the API Callout node and supports the same success/error handler pattern.

If any referenced agent or tool is not published to HALO's live environment, a Production publication from AI Cloud will fail.

→ Using HALO Agents and Tools in Content · HALO Tool Call Node

Analytics

AI Cloud projects with a linked HALO profile get access to HALO Analytics directly from the CMS Analytics screen. Clicking Open HALO Studio opens HALO's analytics dashboards without requiring a separate login. HALO Analytics provides agentic-layer insights — conversation resolution rates, customer feedback, and topic analysis — that complement AI Cloud's content-level dashboards.

→ CAIC Analytics · HALO Analytics

Publication

AI Cloud and HALO have separate publication flows and different environment models:

Publishing AI Cloud to Staging has no effect on HALO. Publishing to Production requires that any HALO agent or tool referenced in your CMS content is already published to HALO's live environment — or that the Publish HALO toggle is enabled to auto-publish the linked HALO profile alongside AI Cloud.

AI Cloud publishes the entire CMS in one operation. HALO supports selective publication — individual agents, tools, and other resources can be published independently.

→ CAIC Publication · HALO Publication