Skip to content

armgit / macro

Public Imported from GitHub
0

Macro is a unified workspace for teams: email, chat, docs, tasks, agents, calls, and CRM — @-linked together with shared AI memory.

fix(search): make email address domains and local parts searchable (#4898)

gbirman0eb783868f6b
diff --git a/crates/opensearch_client/src/search/emails.rs b/crates/opensearch_client/src/search/emails.rs
index c8ec2ad070..32c885d36a 100644
--- a/crates/opensearch_client/src/search/emails.rs
+++ b/crates/opensearch_client/src/search/emails.rs
@@ -18,10 +18,26 @@ impl SearchQueryConfig for EmailSearchConfig {
const ENTITY_INDEX: OpenSearchEntityType = OpenSearchEntityType::Emails;
}
-/// Keyword-indexed email-address fields. Matched explicitly: a bare word
-/// only matches an address with that exact local-part (`alice` → `alice@*`),
-/// not addresses that merely start with the characters.
-const EMAIL_KEYWORD_FIELDS: &[&str] = &["sender", "reply_to", "recipients", "cc", "bcc"];
+/// Email-address fields matched explicitly, never as a character prefix.
+/// The raw keyword-indexed addresses plus their `.parts` sub-fields, which
+/// the `email_parts` analyzer splits into positioned segment tokens
+/// (`[email protected]` indexes `jane`, `doe`, `mail`, `foo`, `com`).
+/// A bare word matches only a whole segment, so `ali` won't match
+/// `[email protected]` and `immipart` won't match `immipartner.com`.
+/// Dotted terms like `foo.com` are quoted into phrases and match any
+/// in-order adjacent run of segments.
+const EMAIL_KEYWORD_FIELDS: &[&str] = &[
+ "sender",
+ "reply_to",
+ "recipients",
+ "cc",
+ "bcc",
+ "sender.parts",
+ "reply_to.parts",
+ "recipients.parts",
+ "cc.parts",
+ "bcc.parts",
+];
/// Text-analyzed fields. Matched as a prefix so `scri` matches `script`.
const EMAIL_TEXT_FIELDS: &[&str] = &[
@@ -33,25 +49,29 @@ const EMAIL_TEXT_FIELDS: &[&str] = &[
"bcc_names",
];
-/// Builds the simple_query_string over the keyword email-address fields.
-/// Single-word terms become `(term | term@*)` — an exact token match OR a
-/// local-part match on the address. This deliberately does NOT use a trailing
-/// `*` on the bare term, so partial local-parts don't leak across addresses
-/// (e.g. `ali` won't match `[email protected]` via this group).
-/// Email-like terms (containing `@`) and multi-word terms (quoted phrases
-/// from `split_search_terms`) are wrapped in quotes to force phrase matching.
-/// Without the quotes, `default_operator: "AND"` would turn `(reply test)`
-/// into `reply AND test` and match each token independently anywhere in the
-/// field instead of as an adjacent phrase. The standard analyzer also
-/// tokenizes `[email protected]` into `[alice, example, com]`, so phrase
-/// matching is what keeps the address from matching across unrelated tokens.
+/// Builds the simple_query_string over the email-address fields.
+/// Single-word terms become `(term | term@*)` — an exact segment-token match
+/// on the `.parts` sub-fields OR an exact-local-part match on the raw
+/// address. The `term@*` pattern is redundant with `.parts` once the index
+/// carries the sub-fields but keeps the query compatible with an index that
+/// predates them. This deliberately does NOT use a trailing `*` on the bare
+/// term, so partial segments don't leak across addresses (e.g. `ali` won't
+/// match `[email protected]` via this group).
+/// Terms containing `@`, `.`, or spaces (quoted phrases from
+/// `split_search_terms`) are wrapped in quotes to force phrase matching. On
+/// the `.parts` sub-fields the `email_parts` analyzer splits such terms into
+/// segment tokens, so the phrase requires an in-order adjacent run —
+/// `mail.foo.com`, `foo.com`, and full addresses all match `[email protected]`,
+/// but reordered or gapped segments don't. Without the quotes,
+/// `default_operator: "AND"` would match each segment independently anywhere
+/// in the field.
/// The email pattern is lowercased because email addresses are case-insensitive.
/// Terms are ANDed together with `+`.
fn build_keyword_query_string(terms: &[String]) -> String {
terms
.iter()
.map(|term| {
- if term.contains('@') || term.contains(' ') {
+ if term.contains('@') || term.contains('.') || term.contains(' ') {
format!("\"{}\"", term)
} else {
format!("({} | {}@*)", term, term.to_lowercase())
diff --git a/crates/opensearch_client/src/search/emails/test.rs b/crates/opensearch_client/src/search/emails/test.rs
index 8b7ac3a274..91649d0698 100644
--- a/crates/opensearch_client/src/search/emails/test.rs
+++ b/crates/opensearch_client/src/search/emails/test.rs
@@ -34,6 +34,12 @@ fn test_build_keyword_query_string_mixed_single_and_multi_word() {
assert_eq!(result, "(Teo | teo@*) + \"hello world\"");
}
+#[test]
+fn test_build_keyword_query_string_dotted_term_uses_phrase() {
+ let result = build_keyword_query_string(&["example.com".to_string()]);
+ assert_eq!(result, "\"example.com\"");
+}
+
#[test]
fn test_build_keyword_query_string_email_term_uses_phrase() {
let result = build_keyword_query_string(&["[email protected]".to_string()]);
@@ -187,6 +193,11 @@ fn test_email_search_args_build_injects_simple_query_string() -> anyhow::Result<
assert!(keyword_fields.contains(&serde_json::json!("recipients")));
assert!(keyword_fields.contains(&serde_json::json!("cc")));
assert!(keyword_fields.contains(&serde_json::json!("bcc")));
+ assert!(keyword_fields.contains(&serde_json::json!("sender.parts")));
+ assert!(keyword_fields.contains(&serde_json::json!("reply_to.parts")));
+ assert!(keyword_fields.contains(&serde_json::json!("recipients.parts")));
+ assert!(keyword_fields.contains(&serde_json::json!("cc.parts")));
+ assert!(keyword_fields.contains(&serde_json::json!("bcc.parts")));
assert!(!keyword_fields.contains(&serde_json::json!("subject")));
let text_sqs = should
@@ -206,6 +217,7 @@ fn test_email_search_args_build_injects_simple_query_string() -> anyhow::Result<
assert!(text_fields.contains(&serde_json::json!("sender_name")));
assert!(text_fields.contains(&serde_json::json!("recipient_names")));
assert!(!text_fields.contains(&serde_json::json!("sender")));
+ assert!(!text_fields.contains(&serde_json::json!("sender.parts")));
Ok(())
}
@@ -282,7 +294,7 @@ fn test_build_bool_query() -> anyhow::Result<()> {
{
"simple_query_string": {
"default_operator": "AND",
- "fields": ["sender", "reply_to", "recipients", "cc", "bcc"],
+ "fields": ["sender", "reply_to", "recipients", "cc", "bcc", "sender.parts", "reply_to.parts", "recipients.parts", "cc.parts", "bcc.parts"],
"query": "(test | test@*)"
}
},
diff --git a/crates/opensearch_client/src/search/unified/test.rs b/crates/opensearch_client/src/search/unified/test.rs
index 67bc2e865b..671e650689 100644
--- a/crates/opensearch_client/src/search/unified/test.rs
+++ b/crates/opensearch_client/src/search/unified/test.rs
@@ -674,7 +674,7 @@ fn test_build_unified_search_request_content() -> anyhow::Result<()> {
{
"simple_query_string": {
"default_operator": "AND",
- "fields": ["sender", "reply_to", "recipients", "cc", "bcc"],
+ "fields": ["sender", "reply_to", "recipients", "cc", "bcc", "sender.parts", "reply_to.parts", "recipients.parts", "cc.parts", "bcc.parts"],
"query": "(test | test@*)"
}
},
diff --git a/infra/stacks/opensearch/helpers/constants.ts b/infra/stacks/opensearch/helpers/constants.ts
index d3ca7c9cd1..deb8d9490b 100644
--- a/infra/stacks/opensearch/helpers/constants.ts
+++ b/infra/stacks/opensearch/helpers/constants.ts
@@ -13,7 +13,7 @@ export const PROJECTS_ALIAS = 'projects';
export const CHANNELS_INDEX = 'channels_v2';
export const CHATS_INDEX = 'chats_v2';
export const DOCUMENTS_INDEX = 'documents_v2';
-export const EMAILS_INDEX = 'emails_v1';
+export const EMAILS_INDEX = 'emails_v2';
export const CALL_RECORDS_INDEX = 'call_records_v2';
export const PROJECTS_INDEX = 'projects_v1';
diff --git a/infra/stacks/opensearch/helpers/scripts/create_indices.ts b/infra/stacks/opensearch/helpers/scripts/create_indices.ts
index 8e3e1e6a60..30bb304f70 100644
--- a/infra/stacks/opensearch/helpers/scripts/create_indices.ts
+++ b/infra/stacks/opensearch/helpers/scripts/create_indices.ts
@@ -424,10 +424,44 @@ const PROJECTS_BODY = {
},
};
+// Splits addresses into positioned segment tokens for the `.parts`
+// sub-fields: `[email protected]` -> [jane, doe, mail, foo, com].
+// Same-width space replacements keep highlight offsets aligned, and the
+// positions make dotted query phrases require in-order adjacent segments.
+const EMAIL_PARTS_ANALYSIS = {
+ char_filter: {
+ email_separators: {
+ type: 'mapping',
+ mappings: ['@ => \\u0020', '. => \\u0020', '+ => \\u0020'],
+ },
+ },
+ analyzer: {
+ email_parts: {
+ type: 'custom',
+ char_filter: ['email_separators'],
+ tokenizer: 'standard',
+ filter: ['lowercase'],
+ },
+ },
+};
+
+const EMAIL_ADDRESS_FIELD = {
+ type: 'keyword',
+ index: true,
+ doc_values: true,
+ fields: {
+ parts: {
+ type: 'text',
+ analyzer: 'email_parts',
+ },
+ },
+};
+
const EMAIL_BODY = {
settings: {
...SHARD_SETTINGS,
refresh_interval: '2s',
+ analysis: EMAIL_PARTS_ANALYSIS,
},
mappings: {
dynamic: 'false',
@@ -440,31 +474,11 @@ const EMAIL_BODY = {
index: true,
doc_values: true,
},
- sender: {
- type: 'keyword',
- index: true,
- doc_values: true,
- },
- reply_to: {
- type: 'keyword',
- index: true,
- doc_values: true,
- },
- recipients: {
- type: 'keyword',
- index: true,
- doc_values: true,
- },
- cc: {
- type: 'keyword',
- index: true,
- doc_values: true,
- },
- bcc: {
- type: 'keyword',
- index: true,
- doc_values: true,
- },
+ sender: EMAIL_ADDRESS_FIELD,
+ reply_to: EMAIL_ADDRESS_FIELD,
+ recipients: EMAIL_ADDRESS_FIELD,
+ cc: EMAIL_ADDRESS_FIELD,
+ bcc: EMAIL_ADDRESS_FIELD,
sender_name: {
type: 'text',
analyzer: 'standard',