Private metadata allowlist

This document lists all telemetry events that are allowed to export specific privateMetadata fields.

By default, privateMetadata is never exported from customers because it can contain arbitrarily-shaped data that may accidentally include sensitive or private contents. The events listed below have been individually reviewed and approved for export of specific private metadata fields only.

To learn more, refer to the telemetry documentation.

Feature	Action	Allowed `privateMetadata` fields
`cody.completion`	(all)	`languageId` — Provides customers with valuable language-specific insights for code completion usage patterns. `otherCompletionProviders` — Provides internal teams insights on competitor completion providers customers are using. `providerModel` — Provides internal teams insights on the specific model used for completions. `resolvedModel` — The resolved model used for the completion request.
`cody.hoverCommands`	visible	`languageId` — Provides customers with valuable language-specific insights for hover command usage patterns.
`blob.codeintel`	(all)	`languageId` — Provides customers with valuable language-specific insights for code intelligence usage patterns.
`cody.chat-question`	submitted	`chatModel` — Provides customers with chat-model specific insights to help determine which models best suit their use cases. `chatAgent` — Provides customers with chat-agent (e.g. deep-cody) specific insights to help determine which agents best suit their use cases.
`cody.chatResponse`	(all)	`chatModel` — Provides customers with chat-model specific insights to help determine which models best suit their use cases.
`cody.chatResponseNew`	(all)	`chatModel` — Provides customers with chat-model specific insights to help determine which models best suit their use cases.
`cody.command.doc`	executed	`model` — Provides customers with model-specific insights to help determine which models best suit their use cases.
`cody.command.edit`	executed	`model` — Provides customers with model-specific insights to help determine which models best suit their use cases.
`cody.command.fix`	executed	`model` — Provides customers with model-specific insights to help determine which models best suit their use cases.
`cody.command.test`	executed	`model` — Provides customers with model-specific insights to help determine which models best suit their use cases.
`search`	execute	`lang` — High-cardinality language identifier; not sensitive and helpful for determining how users are using search.
`codeintel.explorePanel.usages`	(all)	`lang` — High-cardinality language identifier; not sensitive and helpful for determining how users are using search.
`codeintel.explorePanel`	(all)	`lang` — High-cardinality language identifier; not sensitive and helpful for determining how users are using search.
`codeintel`	(all)	`lang` — High-cardinality language identifier; not sensitive and helpful for determining how users are using search.
`cody.autoedit`	(all)	`model` — High-cardinality model identifier; not sensitive and helpful for determining what model is being used. `languageId` — Identifies the language of the code; not sensitive and high cardinality.
`cody.fixup.user`	(all)	`model` — High-cardinality model identifier; not sensitive and helpful for determining what model is being used.
`cody.fixup.user`	(all)	`model` — High-cardinality model identifier; not sensitive and helpful for determining what model is being used.
`cody.fixup.response`	(all)	`model` — High-cardinality model identifier; not sensitive and helpful for determining what model is being used.
`cody.fixup.apply`	(all)	`model` — High-cardinality model identifier; not sensitive and helpful for determining what model is being used.
`cody.fixup.persistence`	(all)	`model` — High-cardinality model identifier; not sensitive and helpful for determining what model is being used.
`cody.modelSelector`	(all)	`modelId` — High-cardinality model identifier; helpful for determining the model selected in the model selector. `modelProvider` — High-cardinality model provider; helpful for determining the model selected in the model selector.
`cody.smart-apply.context`	applied	`model` — High-cardinality model identifier; helpful for determining the model that was selected.
`deepsearch`	search.toolcall	`toolName` — High-cardinality tool name; helpful for determining which tools are being used during deep search. `toolId` — High-cardinality tool identifier; helpful for determining which tools are being used during deep search. `model` — High-cardinality model identifier; helpful for determining which models are being used during deep search.
`externalApi`	request	`procedure` — ConnectRPC procedure path (e.g. '/sourcegraph.users.v1.UsersService/GetUser'); not sensitive and needed to distinguish which external API RPCs are being used.