vh
3b9c610587
Replaces the stateless _render_event / _render_event_to_log helpers with stateful per-turn presenters (CliPresenterState / TuiPresenterState). Coalesces thinking-event deltas into a single growing display per run; demotes telemetry events with editorial hierarchy; formats duration + usage for human reading. Headline behavior change: a 50-token thinking phase now renders as ONE coalesced growing line in CLI (or one closed RichLog entry + per-delta live Static widget in TUI), not 50 lines of [thinking] spam. Editorial promotion line (issue #12 INV-002): - Load-bearing (no demotion prefix): Text, Done, Error, Cancelled - Demoted telemetry (`. ` ASCII prefix in CLI; dim `· ` in TUI): WorkerPhase, Thinking, TextBoundary, ToolStart, ToolResult Stateful coalescing: - Thinking deltas accumulate into thinking_buffer; first non-thinking event closes the run with a single \n boundary in CLI / one closed dim RichLog entry in TUI. - TUI adds a dedicated Static(id="thinking-current") widget that shows the last ~200 chars of the active run, mirroring per-delta updates. Two-views-of-thinking decoupling per INV-004: chronological RichLog + always-visible widget. - CLI INV-005: when stdout text was streamed mid-line, text_written_since_newline triggers a stdout flush + \n before the next stderr terminal label — guarantees [done] / [error] / [cancelled] land on their own line in a TTY without breaking pipe-to-file scripted consumers. Formatting helpers (issue #12 INV-006 / INV-007): - _format_duration_ms — autoscale `347ms` / `5.5s` / `1.2m` - _format_usage — natural-language `6756 in -> 126 out (6882 total, 0 cached)` with arrow="->" CLI / "→" TUI Cross-frontier design pass (eitri-smithy-dev, althing 01KSBE52YZR5E3SPTKA672JE43) returned 16-of-16 confirmed decisions + 4 material divergences applied: - ASCII `. ` prefix in CLI (`·` is U+00B7, not ASCII) - RichLog one-closed-entry-per-run + Static per-delta updates (not inline-mirror as initially proposed) - presenter-state object instead of pure-function rendering - Framed as "contract semantics amendment", not "polish" Volva paraphrase round (5 prose-precision fixes applied to 12.contract.md): INV-001 "growing display" semantics; single hide mechanism for the Static widget (Textual reactive `display: bool`); [render_error] security clause (type-only, no exception message); text_written_since_newline `\n`-terminated text corner case; [create_session] integration path (bypasses state.render — not an SSE Event variant). Volva code-review round (5 findings applied): - F1 drift: render-exception fallback now writes BOTH a plain-label fallback line for the original event AND the `[render_error] <type>` line (was missing the fallback half). - F2 drift: dim Rich style applied to all demoted-telemetry RichLog writes via `rich.text.Text(..., style="dim")` (was plain str). - F3 drift: belt-and-braces widget clear+hide on EVERY terminal event (Done/Error/Cancelled), even when thinking_open was False. - F4 precision: _format_usage gains PRE-001 assertion on the four expected usage keys. - F5 precision: _run_turn signature amended in issue #3 contract to document the new `state: CliPresenterState | None = None` test- injection kwarg. [create_session] lifecycle line demoted to `. create_session:` (written directly by _amain; bypasses state.render since it's not a wire-level SSE Event variant). Pre-amendment _render_event / _render_event_to_log and their test classes removed under the no-backwards-compat rule. Issues #3 and #4 contracts amended in-place: #3 (CliPresenterState CLASS + FN block + helper FN blocks + _run_turn signature + _amain create_session demotion); #4 (TuiPresenterState CLASS + FN block + compose Static widget + _stream_turn_worker state construction). 209 tests GREEN; ruff clean. Bumps v0.1.0 → v0.2.0 (minor — output shape change breaks pre-amendment grep patterns like `[thinking] '`; no public API surface change beyond the rendering contract). Persistent-memory commit-along: captures the issue #12 decision, forward direction (require end_user_id for every access — declined worldtree-dev's requires_end_user_id offer because we'll send it universally), and the Heimdall scope-model foot-gun note (the "per-Tier-1-agent scope add" diagnosis was a phantom ask resolved by worldtree-dev's correction; agent.call:* baseline covers all Tier 1).
Ratatoskr
A Worldtree Conversation API debug TUI. Runs up and down Worldtree's API surface — sessions, turns, persona, tools, admin events, Bifrost state — carrying messages between layers. Like the squirrel.
The product is the observability surface; chat is the input mechanism. Devs run Ratatoskr against a local Worldtree to watch a turn flow through every layer of the system, side-by-side, in one terminal.
Status
v0 scaffold. Design locked; implementation hasn't started. The dev team owns the implementation pass.
Read in this order
docs/design-brief.md— the locked design. Read this first. Every architectural decision is recorded with its rationale, the alternatives considered, and (where relevant) the operator's lock-in moment.docs/SPEC-PIN.md— what Worldtree spec version Ratatoskr is built against, where the vendored snapshot lives, and how to bump the pin.docs/conversation-api-spec.md— the vendored Worldtree spec snapshot. Read this to understand the API surface Ratatoskr consumes. Do not import anything from a Worldtree checkout — the boundary is the spec, not the code. Seedocs/design-brief.md§2.CLAUDE.md— Claude Code conventions for this repo (mostly inherited from the Corviduo template).persistent-memory.md— durable intent across context resets. Update as decisions and state evolve.
Quickstart
# 1. Environment
uv venv && source .venv/bin/activate
uv pip install -e ".[dev]"
# 2. Verify the spec pin
cat docs/SPEC-PIN.md # documented Worldtree SHA + bump procedure
# 3. Tests (none yet; scaffold only)
uv run pytest
# 4. Run against a local Worldtree (once implementation lands)
# Worldtree must be running: python -m core.conversation_api
ratatoskr --agent mimir
What this repo is NOT
- NOT a polished consumer for Worldtree's end-users — that's the web app.
- NOT a Worldtree admin tool — admin CLI is separate (
sessions_cli.pylives in Worldtree). - NOT a featuretracking shadow of the web app — when the Conversation API surface grows, Ratatoskr does not necessarily grow with it.
- NOT a remote-Worldtree debug client — file-tail surfaces (persona log, server log) assume local-dev posture.
The full negative-clause list lives in docs/design-brief.md §6.
Boundary rule
Ratatoskr depends on three things only:
httpx+httpx-sse(network layer)textual(TUI framework)- Worldtree's published Conversation API spec at the pinned SHA
Hard rule: no imports from a Worldtree checkout. No core.* imports,
no from worldtree.*, no shared models, no submodule of Worldtree. The
spec is the entire surface. Boundary smoke test at tests/test_no_worldtree_imports.py
enforces this in CI.
Version-skew strategy
The dev team is decoupled from Worldtree's dev team. To detect drift when Worldtree changes the API:
- Spec-version pin in
pyproject.toml(worldtree-spec-rev). Bump explicitly; bumps are a tracked action. - Recorded-SSE snapshot tests at
tests/snapshots/. Captured against a live Worldtree; replayed in CI. Re-record after every pin bump. - Conformance smoke test — boots Worldtree via Docker compose in CI, runs a one-turn happy path. Catches integration-level drift.
See docs/SPEC-PIN.md for the bump procedure.
Consumer-side discoveries
If you find a spec gap, ambiguity, or missing-but-needed endpoint while working on Ratatoskr: route the discovery back to worldtree-dev via althing rather than via PR on Worldtree directly. The separate-team boundary is intentional and helps catch spec gaps that an in-tree consumer would paper over.
Related repos
- Worldtree — the API Ratatoskr consumes
- brokkr-smithy — authored this design brief
- Skaldsong — another Worldtree API consumer (Python; render-and-aggregate pattern)
- mead-hall — another Worldtree API consumer (TypeScript; server-side broker)