Summary
The telemetry reference (resources/help/telemetry.mdx) lists 14 experimental_* fields under
infos, but the engine's Infos struct — the thing actually serialized into the infos telemetry
payload — currently emits 30. So 16 experimental_* fields are collected but not listed on the
telemetry page. Since that page is what a privacy-conscious operator reads to decide whether to run
with telemetry on, it seemed worth reconciling.
I know the page already says "This list is liable to change with every new version" — this is just
to help it catch up, not a gotcha.
How I checked (reproducible)
- Emitted (source of truth): the
Infos struct in
crates/meilisearch/src/analytics/segment_analytics.rs (#[derive(Serialize)]; its own doc-comment says "the infos field we send in the analytics … sent as-is in amplitude"). It goes into compute_traits → the identify traits.
- Documented:
experimental_ names in resources/help/telemetry.mdx @ 3060253e7.
- Diffing the two sets: 30 emitted − 14 documented = 16 emitted but not documented (and 0 documented-but-not-emitted).
The 16 experimental_* fields sent but not on the telemetry page
experimental_allowed_ip_networks
experimental_chat_completions
experimental_drop_search_after
experimental_dynamic_search_rules
experimental_foreign_keys
experimental_get_task_documents_route
experimental_legacy_search
experimental_limit_batched_tasks_total_size
experimental_max_number_of_batched_tasks
experimental_nb_searches_per_core
experimental_no_edition_2024_for_dumps
experimental_no_snapshot_compaction
experimental_personalization
experimental_queue_documents_fetch
experimental_render_route
experimental_task_queue_compaction_route
(All under infos. in the payload, same as the documented ones.)
Suggested fix
Add these rows to the infos section of the telemetry page. I'd be glad to open the PR — just wanted
to confirm the list with you first, since a couple (e.g. experimental_chat_completions,
experimental_personalization, experimental_render_route) look like recently-added features.
Disclosure (per your CONTRIBUTING): this was prepared with the help of a generative-AI coding
tool (Claude Code). I verified every claim myself — the 16-field delta is a mechanical diff of the two
cited, SHA-pinned sources above — and I own this contribution.
Summary
The telemetry reference (
resources/help/telemetry.mdx) lists 14experimental_*fields underinfos, but the engine'sInfosstruct — the thing actually serialized into theinfostelemetrypayload — currently emits 30. So 16
experimental_*fields are collected but not listed on thetelemetry page. Since that page is what a privacy-conscious operator reads to decide whether to run
with telemetry on, it seemed worth reconciling.
I know the page already says "This list is liable to change with every new version" — this is just
to help it catch up, not a gotcha.
How I checked (reproducible)
Infosstruct incrates/meilisearch/src/analytics/segment_analytics.rs(#[derive(Serialize)]; its own doc-comment says "theinfosfield we send in the analytics … sent as-is in amplitude"). It goes intocompute_traits→ the identifytraits.experimental_names inresources/help/telemetry.mdx@3060253e7.The 16
experimental_*fields sent but not on the telemetry pageexperimental_allowed_ip_networksexperimental_chat_completionsexperimental_drop_search_afterexperimental_dynamic_search_rulesexperimental_foreign_keysexperimental_get_task_documents_routeexperimental_legacy_searchexperimental_limit_batched_tasks_total_sizeexperimental_max_number_of_batched_tasksexperimental_nb_searches_per_coreexperimental_no_edition_2024_for_dumpsexperimental_no_snapshot_compactionexperimental_personalizationexperimental_queue_documents_fetchexperimental_render_routeexperimental_task_queue_compaction_route(All under
infos.in the payload, same as the documented ones.)Suggested fix
Add these rows to the
infossection of the telemetry page. I'd be glad to open the PR — just wantedto confirm the list with you first, since a couple (e.g.
experimental_chat_completions,experimental_personalization,experimental_render_route) look like recently-added features.Disclosure (per your CONTRIBUTING): this was prepared with the help of a generative-AI coding
tool (Claude Code). I verified every claim myself — the 16-field delta is a mechanical diff of the two
cited, SHA-pinned sources above — and I own this contribution.