rotate chat TUI conversation id per launch / clear-chat

CHANGELOG.md

@@ -5,6 +5,7 @@
 
 - Bump `haiku.skills>=0.17.1` and `pydantic-ai-slim>=1.100.0` (the last pre-2.0 release). Migrate off two APIs slated for removal in pydantic-ai 2.0: `Agent(tool_retries=, output_retries=)` → `Agent(retries={"tools": …, "output": …})` in the LLM-as-judge evaluator, and `Evaluator.evaluation_name` class attribute → overriding `get_default_evaluation_name()` on `CitationMRREvaluator` / `CitationMAPEvaluator`.
 - Drop the `item.annotations` fallback in `_picture_description_text`. Docling's `PictureItem` runs a `@model_validator(mode="after")` on load that migrates the deprecated `annotations` field into `meta.description`, so reading `meta.description.text` covers both legacy and current blobs. Tests in `test_converters.py` switched to `meta.description.text` for the same reason.
+- Chat TUI now generates a stable per-launch `thread_id` (rotated on "Clear chat") instead of hardcoding `"tui"`. AGUIAdapter forwards it as the `gen_ai.conversation.id` OTel attribute, so multi-turn TUI sessions group into one Logfire conversation instead of collapsing every launch into a single bucket.
 - Documentation generator swapped from `mkdocs-material` to `zensical`. Drops `mkdocs` / `mkdocs-material` dev deps, replaces `mkdocs.yml` with `zensical.toml`, adds `overrides/main.html` (OG/Twitter share meta) and `docs/stylesheets/extra.css`. `build-docs` workflow now runs `uv run zensical build` and publishes via the GitHub Pages artifact actions instead of `mkdocs gh-deploy`.
 
 ### Fixed

haiku_rag_slim/haiku/rag/chat/app.py

@@ -108,6 +108,10 @@ class ChatApp(App):
         self._is_processing = False
         self._current_worker: Worker[None] | None = None
         self._document_filter: list[str] = []
+        # Stable per-launch id so multi-turn chats land in one Logfire
+        # conversation. AGUIAdapter reads run_input.thread_id and exports it
+        # as the `gen_ai.conversation.id` OTel attribute on every agent run.
+        self._conversation_id = str(uuid.uuid4())
 
     def compose(self) -> "ComposeResult":
         """Compose the UI layout."""
@@ -207,7 +211,7 @@ class ChatApp(App):
         await chat_history.show_thinking()
 
         run_input = RunAgentInput(
-            thread_id="tui",
+            thread_id=self._conversation_id,
             run_id=str(uuid.uuid4()),
             messages=self._messages,
             state=self._state,
@@ -375,6 +379,8 @@ class ChatApp(App):
         # Reset state
         if self._toolset:
             self._state = self._toolset.build_state_snapshot()
+        # Cleared chat starts a fresh Logfire conversation.
+        self._conversation_id = str(uuid.uuid4())
 
     def action_focus_input(self) -> None:
         """Focus the input field, or cancel if processing."""

tests/chat/test_chat_app.py

@@ -219,7 +219,7 @@ async def test_chat_history_thinking_indicator(temp_db_path: Path):
 
 @pytest.mark.asyncio
 async def test_clear_chat_resets_state(temp_db_path: Path):
-    """Test that clearing chat resets state and messages."""
+    """Test that clearing chat resets state, messages, and conversation id."""
     from haiku.rag.chat.widgets.chat_history import ChatHistory
 
     app, mock_client = _make_app(temp_db_path)
@@ -232,10 +232,12 @@ async def test_clear_chat_resets_state(temp_db_path: Path):
             await chat_history.add_message("assistant", "Hi there")
             assert len(chat_history.messages) == 2
 
+            previous_conversation_id = app._conversation_id
             await app.action_clear_chat()
             await pilot.pause()
 
             assert len(chat_history.messages) == 0
+            assert app._conversation_id != previous_conversation_id
 
 
 @pytest.mark.asyncio