feat(persona): Lane D — Rust persona/turn-execute chains drain -> prompt -> inference (#1409)

[joelteply]

Lane D atomic slice — closes alpha card #1409

Adds the persona/turn-execute command in CognitionModule that executes a full persona turn in one Rust hop:

drain inbox
  -> wrap in PersonaTurnFrame
  -> derive ResponsePrompt (lazy output)
  -> build InferenceRequest (prompt_text path)
  -> dispatch \`inference/llm/request\` via command_executor
  -> bundle replayRecord + inferenceResponse
  -> persist replay record (v2 schema with response_prompt captured from #1412)

What changed

• src/persona/turn_frame.rs: new ResponsePrompt::to_prompt_text helper that flattens system_prompt + chat messages into a single deterministic plain-text prompt for adapter-based engines. Format: \"<system>\\n\\nrole: content\\nrole: content\\n...\". Empty system_prompt produces no leading paragraph; lowercase role matches PromptRole serde format.
• src/modules/cognition.rs: new persona/turn-execute command. Inputs: persona_id (required), window_ms (default 80), max_items (default 16), composition_artifact_id (default Uuid::nil()), max_tokens (default 512), max_duration_ms (default 10_000). Returns { replayRecord, inferenceResponse }. Empty drain returns the null pair (no-op, not Err). Missing persona returns typed Err per never-swallow rule.

Tests

+9 new tests, all green:

persona::turn_frame (6 new, 18 total):

• to_prompt_text_renders_each_message_as_role_colon_content
• to_prompt_text_prepends_system_prompt_when_present
• to_prompt_text_skips_empty_system_prompt
• to_prompt_text_handles_mixed_roles_in_order
• to_prompt_text_handles_no_messages
• to_prompt_text_empty_prompt_returns_empty_string

modules::cognition::turn_execute_tests (3 new):

• turn_execute_persona_not_found_returns_typed_error
• turn_execute_empty_drain_returns_null_bundle
• turn_execute_bad_max_items_returns_typed_error

The dispatch-success path runs through command_executor::executor() which is only initialized at runtime startup (ipc/mod.rs). End-to-end dispatch tests live in the integration suite; unit-tests here cover param-parse + short-circuit + persona-not-found paths.

Why this matters

The TS persona loop previously executed each stage with its own IPC round-trip (drain → build prompt → call inference), 3 round-trips per turn, with prompt-building in TS. Lane D pulls all three into the substrate so:

1. Prompt built in Rust where the turn-frame lives
2. Production replay record carries the exact prompt that fed inference (v2 schema from feat(persona): Lane D — bump turn-frame replay record to v2 with response_prompt #1412)
3. Persona turn becomes one observable unit on the bus

Validation command

cd src/workers/continuum-core && cargo test --features metal,accelerate --lib persona::turn_frame turn_execute

Result locally: 21/21 pass (18 turn_frame + 3 turn_execute).

Stack

Builds atop:

• feat(persona): Lane D — bump turn-frame replay record to v2 with response_prompt #1412 (v2 replay schema with response_prompt) — merged
• feat(inference): inference-llm PR-5 — Runtime registration #1404 (InferenceLlmModule runtime registration) — merged

Follow-up work (not in scope):

• Integration test booting Runtime with InferenceLlmModule registered + asserting end-to-end dispatch
• TS callsite migration to use `persona/turn-execute` instead of the 3-call chain

🤖 Generated with Claude Code