feat(phase9): ContentTags schema, system prompt, and GBNF grammar

ContentTags serde-serialisable. CONTENT_TAGS_SYSTEM is the system
message rendered at extraction time; INTENT_CLOSED_SET is the single
source of truth for the enum values the grammar restricts. Grammar is
strict: lowercase hyphen-joined topic 3+ chars (max enforced by
max_tokens at call site), intent from the closed set, JSON-only
output. Recursive topic-rest matches the existing GBNF style in this
file.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

crates/llm/src/grammars.rs

@@ -1,3 +1,18 @@
+// Phase 9 content-tag extraction. Restricts the model output to a
+// strict {topic, intent} JSON object where topic is a lowercase
+// hyphen-joined slug of at least 3 chars (no upper bound is encoded
+// in the grammar — max_tokens caps it in practice) and intent is one
+// of the six closed-set values. Recursive `topic-rest` keeps the
+// shape compatible with the existing GBNF style in this file.
+pub const CONTENT_TAGS_GRAMMAR: &str = r##"
+root        ::= "{" ws "\"topic\":" ws topic-str ws "," ws "\"intent\":" ws intent ws "}" ws
+topic-str   ::= "\"" topic-char topic-char topic-char topic-rest "\""
+topic-rest  ::= "" | topic-char topic-rest
+topic-char  ::= [a-z0-9-]
+intent      ::= "\"planning\"" | "\"reflection\"" | "\"venting\"" | "\"capture\"" | "\"decision\"" | "\"question\""
+ws          ::= ([ \t\n] ws)?
+"##;
+
 pub const TASK_ARRAY_GRAMMAR: &str = r#"
 root   ::= "[" ws string ws "," ws string ws "," ws string rest3 ws "]"
 rest3  ::= "" | "," ws string rest4

crates/llm/src/lib.rs

@@ -15,7 +15,9 @@ pub mod grammars;
 pub mod model_manager;
 pub mod prompts;
 
+pub use grammars::CONTENT_TAGS_GRAMMAR;
 pub use model_manager::{recommend_tier, LlmModelId, LlmModelInfo};
+pub use prompts::{is_valid_intent, ContentTags, CONTENT_TAGS_SYSTEM, INTENT_CLOSED_SET};
 
 const DEFAULT_CONTEXT_TOKENS: u32 = 4096;
 const MAX_CONTEXT_TOKENS: u32 = 8192;

crates/llm/src/prompts.rs

@@ -4,6 +4,39 @@ between 3 and 7 concrete, physical micro-steps. Each step must be a short \
 imperative sentence, actionable today, with no commentary. Output ONLY a \
 JSON array of strings.";
 
+// Phase 9 content-tag extraction. The model emits a {topic, intent}
+// JSON pair under a strict GBNF (see grammars::CONTENT_TAGS_GRAMMAR).
+// CONTENT_TAGS_SYSTEM is the system message; the user message wraps
+// the transcript text.
+pub const CONTENT_TAGS_SYSTEM: &str = "\
+You tag a transcript with ONE topic and ONE intent. \
+TOPIC is a 1 to 3 token lowercase hyphen-joined noun phrase naming the \
+dominant subject. Examples: interview-prep, grant-application, \
+daily-standup. \
+INTENT is exactly one of: planning, reflection, venting, capture, \
+decision, question. \
+Return JSON only, with this exact shape: \
+{\"topic\":\"...\",\"intent\":\"...\"}";
+
+#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
+pub struct ContentTags {
+    pub topic: String,
+    pub intent: String,
+}
+
+pub const INTENT_CLOSED_SET: &[&str] = &[
+    "planning",
+    "reflection",
+    "venting",
+    "capture",
+    "decision",
+    "question",
+];
+
+pub fn is_valid_intent(s: &str) -> bool {
+    INTENT_CLOSED_SET.contains(&s)
+}
+
 pub const EXTRACT_TASKS_SYSTEM: &str = "\
 You are a task-extraction assistant. Given a transcript of spoken notes, \
 output a JSON array of action items the speaker committed to. Each item must \