Data injection attributes

Data Injection lets an AI Employee periodically fetch live external data — such as inventory levels, pricing, or availability — from a webhook endpoint and inject it into the agent's runtime context. This keeps the agent's responses grounded in current information without requiring a full redeployment. The four attributes below are the user-configurable controls for this feature; system-managed internals such as the payload cache and timer state are handled automatically.

::: ⚠️ CAUTION

These are Level C (advanced) attributes hidden by default on the Attributes page. Enable Show Hidden to see them. Only configure these attributes after your webhook endpoint and payload are set up and validated. :::

Quick reference

---

project_attributes_setting_data_injection_is_enabled

Type: Boolean
Default: False

This attribute controls whether the AI Employee's data injection fetch loop is active. When set to True, the agent periodically calls your configured webhook endpoint and injects the retrieved data into the additional business context available during conversations. When set to False, the fetch loop stops and any previously injected external context is cleared.

::: 🗒️ NOTE

Enable this attribute only after you have configured a webhook URL and validated the payload and update period. If the fetch returns an empty response, no external context is injected until data becomes available. :::

Behavior by value:

• True: The agent fetches external data on the schedule defined by project_attributes_setting_data_injection_update_period and makes it available in context.
• False: The fetch loop is stopped and injected external context is removed. Any static additional business context you have configured continues to work independently.

Setup checklist before enabling:

1. Go to Integrations and open the API webhook connector settings.
2. Configure the webhook URL and required request handling.
3. Confirm the payload and update period are set correctly.
4. Set this attribute to True.

---

project_attributes_setting_data_injection_update_period

Type: Number (seconds)
Default: 14400

This attribute defines the interval, in seconds, between automatic external data fetch cycles. The default value of 14400 corresponds to a four-hour refresh window.

Behavior by value:

• Lower interval: More frequent refreshes. Useful when your external data changes often, but increases webhook/API traffic and operational cost.
• Higher interval: Less frequent refreshes. Appropriate when your data is relatively stable and low-frequency updates are sufficient.

::: ⚠️ CAUTION

Very short intervals can significantly increase webhook and API traffic. Start with a moderate interval and reduce it only if your external data requires near-real-time updates. :::

::: 🗒️ NOTE

This attribute has no effect unless project_attributes_setting_data_injection_is_enabled is set to True. Interval changes take effect at the start of the next scheduled fetch cycle. :::

---

project_attributes_setting_data_injection_max_summary_tokens

Type: Number
Default: 1000

This attribute sets the token limit used when summarizing fetched external data. It applies only when project_attributes_setting_data_injection_needs_summary is enabled. The summarization process preserves the original content when it is short enough and compresses it when it exceeds this threshold.

Behavior by value:

• Lower value: Stronger compression and lower token usage, but with a higher risk of losing important detail from the source data.
• Higher value: Richer context retention at the cost of higher token usage and reduced summarization benefit.

::: 🗒️ NOTE

This attribute has no effect when project_attributes_setting_data_injection_needs_summary is False — raw fetched data is injected as-is in that case. :::

::: ⚠️ CAUTION

Setting this value too low can degrade the AI Employee's decision quality by over-compressing critical facts from the external data source. Use the smallest value that still preserves the key information your agent needs. :::

---

project_attributes_setting_data_injection_needs_summary

Type: Boolean
Default: False

This attribute controls whether the fetched external data is summarized before being injected into the agent's context. When disabled, the raw payload from the webhook is stored and injected as received. When enabled, the runtime runs a summarization step — capped by project_attributes_setting_data_injection_max_summary_tokens — before saving the result to context.

Behavior by value:

• True: Retrieved data is summarized before injection. Useful for large or verbose external payloads where conciseness improves agent reasoning.
• False: Retrieved data is injected as received. Appropriate when the external payload is already concise or when preserving exact formatting matters.

::: 🗒️ NOTE

Summarization only runs when the fetched payload is non-empty. This attribute has no effect unless project_attributes_setting_data_injection_is_enabled is True. :::