Not referenced from README.md, docs/, npm/, or any Cargo metadata.
README uses assets/screenshot.png. Reduces repo size by 226 KB.
Also cleaned up working-directory cruft (untracked, no commit needed):
apps/ (empty), python/ (empty after egg-info removed),
counterpoint.copilot.db, firebase-debug.log, excalidraw.log, .DS_Store.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Auto-format pass after the tool-call rendering work, footer chip,
mention popup, subagent split, and parse-counter de-flake.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Promote `tools/subagent.rs` (4206 lines) to a folder module:
tools/subagent/
mod.rs — runtime types, manager, tool implementations (~3577 lines)
tests.rs — extracted test module (~631 lines)
This is the safe first step. The audit doc proposed a 4-way split
(mod / spec / executor / tests). I tried the 3-way (mod / tools / tests)
and the runtime <-> tool-impl coupling produces unresolved-symbol errors
because shared helpers (`SubAgentTask`, `run_subagent_task`,
`build_allowed_tools`, `normalize_role_alias`, `parse_spawn_request`,
the agent prompt constants) are referenced from both layers. Doing that
split right needs a small API design pass to decide which helpers
graduate to the manager API and which stay tool-private — out of scope
for a structural reorg. Pulled the test module out as the cleanest
no-API-change win and left a path open for the bigger split later.
Public API unchanged — `pub mod subagent;` still exports the same items
because `mod.rs` is a drop-in replacement for `subagent.rs`.
954 → 954 tests, 0 failures.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The audit doc claimed the wiring was "in place" but only the App state
fields existed (`mention_menu_selected`, `mention_menu_hidden`) — no
helpers, no widget rendering, no key handling. Building it out fully so
the popup actually shows when the user types `@` in the composer and
Up/Down/Enter/Tab/Esc behave the way the slash menu does.
What's new:
1. `file_mention::visible_mention_menu_entries(app, limit)` — the entries
source. Returns `Vec<String>` from the workspace walk, gated on the
`mention_menu_hidden` flag and on the cursor being inside an `@token`.
2. `file_mention::apply_mention_menu_selection(app, entries)` — splices
the selected entry into the input via the existing `replace_file_mention`,
resets `mention_menu_hidden`, surfaces a status confirmation.
3. `ComposerWidget::new(app, max_height, slash_entries, mention_entries)`
— second menu slot. The widget renders whichever slice is non-empty,
addressed by the matching selected index. Mention entries get an `@`
prefix so the popup row reads like the actual mention being composed.
Mention takes precedence (positional check is stricter than slash's
"starts-with-/").
4. ui.rs key handler:
- Up/Down navigate `mention_menu_selected` when the popup is open.
- Enter applies `apply_mention_menu_selection` instead of submitting.
- Tab applies the selection (then falls through to the existing slash /
command-completion / file-mention chain).
- Esc hides the popup until the next input edit (`insert_str` already
resets `mention_menu_hidden`, so typing re-opens it).
6 new tests in `ui/tests.rs`:
- mention_popup_is_empty_when_cursor_is_not_in_a_mention
- mention_popup_lists_workspace_matches_for_cursor_partial
- mention_popup_respects_hidden_flag
- apply_mention_menu_selection_splices_selected_entry
- apply_mention_menu_selection_is_noop_outside_a_mention
- apply_mention_menu_selection_with_no_entries_is_noop
Also fixes a stray duplicate `#[cfg(...)]` and an unused-doc-comment
warning that landed when the parse-counter went thread-local — back to
baseline 7 clippy warnings.
948 → 954 tests, 0 failures.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Audit pass found the auto-scroll paths are already gated correctly:
- `mark_history_updated` only bumps history_version + needs_redraw — does
NOT scroll.
- All tool-cell handlers (`handle_tool_call_started`,
`handle_tool_call_complete`, `push_active_tool_cell`,
`register_tool_cell`) call `mark_history_updated` only — none of them
call `scroll_to_bottom`.
- `add_message` and `flush_active_cell` gate their auto-scroll on
`user_scrolled_during_stream`.
- The per-stream lock clears at TurnComplete (ui.rs ~557) and when the
user scrolls back to the live tail (widgets/mod.rs ~126).
- Explicit user actions (vim G, End, session resume, message submit) call
`scroll_to_bottom` directly — that's correct.
5 new regression tests in ui/tests.rs lock the contract so a future
contributor adding `app.scroll_to_bottom()` to a tool-cell handler hits a
red CI immediately:
- add_message_does_not_scroll_when_user_scrolled_away
- add_message_pins_to_tail_when_user_was_following
- tool_call_started_does_not_scroll_when_user_scrolled_away
- tool_call_complete_does_not_scroll_when_user_scrolled_away
- mark_history_updated_does_not_call_scroll_to_bottom
948 → 948 (no changes; tests were already passing — they just weren't
written yet).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`parse_invocations_increment` and `render_parsed_does_not_call_parse` both
read the global PARSE_INVOCATIONS atomic. They were racing whenever any
other test in the suite called `parse()` in parallel — the global counter
would tick once for each unrelated call and the assertion (== 2 / == 0)
would mismatch.
Switching to `thread_local!<Cell<u64>>` gives each test thread its own
counter, so concurrent callers from other tests can't pollute the result.
Tested across 8 sequential full-suite runs: 8/8 green (was ~40% green).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
FooterProps gains an `agents` chip slot, populated by `footer_agents_chip`
which mirrors the rest of the footer chips: empty `Vec<Span>` when
`running_agent_count == 0` (chip hides), "1 agent" / "N agents" otherwise,
DeepSeek-sky color matching the model badge.
The widget's `auxiliary_spans` includes it in the same drop-from-end
fit-to-width chain as the existing chips, so on narrow terminals the cost
chip drops first as before.
The "0 running" wording the audit doc called out wasn't actually in
FooterProps — that wording is in the agent sidebar (ui.rs ~2960) and was
already fixed there to swap to "N done" once nothing is in flight. So the
P2.5 work here is the additive footer surface, not a wording fix.
4 new tests in widgets/footer.rs:
- footer_agents_chip_is_empty_when_no_agents_running
- footer_agents_chip_uses_singular_for_one
- footer_agents_chip_uses_plural_for_many
- footer_agents_chip_renders_into_widget
939 → 943 tests, 0 failures.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The engine used to fire `Event::ToolCallStarted` from `ContentBlockStart::ToolUse`
with `input: json!({})` — before any `Delta::InputJsonDelta` had streamed in.
The UI's `handle_tool_call_started` baked the placeholder into the cell at
creation time and never refreshed, so users saw `<command>` and `<file>`
literals while the args finished streaming.
Fix relocates the emission to `ContentBlockStop` (where the input is finalized
already) and routes it through a new `final_tool_input(state)` helper that
prefers the parsed buffer over a stale empty initial input. Three regression
tests in `engine/tests.rs` pin the contract.
Also bundled (same theme — make in-flight tool cells read right):
- Progressive labels via `exploring_label`: "Read foo.rs" → "Reading foo.rs",
"List X" → "Listing X", "Search pattern" → "Searching for `pattern`",
"List files" → "Listing files". 5 tests in `ui/tests.rs`.
- `running_status_label_with_elapsed` in `history.rs`: from 3 s onward the
status segment becomes `running (Ns)` and ticks every second, driven by
the existing CX#3 status-animation tick. Below 3 s no badge — quick
reads/greps stay quiet. Wired through `render_tool_header`. 2 tests.
- Spinner cadence sped up: `TOOL_STATUS_SYMBOL_MS` 1800 → 720 ms per glyph,
so the 4-glyph "heartbeat" is ~2.88 s instead of ~7.2 s.
929 → 939 tests, 0 failures.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces "tool start pushes new cell" with a single ActiveCell that
collects parallel/serial tool entries at the transcript tail and
flushes as a contiguous block on first assistant text or turn complete.
Stops the bounce when many tools fire concurrently.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Codex pattern — instead of appending a new ToolCall history cell for each
parallel tool invocation, keep one Exploring/Searching/Reading active cell at
the tail of the transcript and mutate its contents in place as new tool calls
fire. Drops cell churn and keeps the visual anchor stable while multiple tools
stream concurrently.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- HistoryCell::Thinking — live shows first ~4 lines + Ctrl+O affordance;
transcript_lines() returns full content with all paragraphs.
- ExecCell — live caps with head/tail + omission marker; transcript
emits all wrapped lines without truncation.
- Tool/Patch/Mcp/Review cells — live caps + affordance; transcript
uncapped.
- User/Assistant/System/Plan/Diff/etc — display == transcript.
- Pager (Ctrl+O / Ctrl+T) flows through transcript_lines via
history_cell_to_text — opening the pager on a thinking or capped tool
cell shows the full body.
Updated affordance assertion to match the post-CX#9 wording
(press Ctrl+O for full text).
911/911 tests pass; clippy -D warnings clean; fmt clean.
CX#5 (newline-boundary streaming gate):
- New crates/tui/src/tui/streaming/line_buffer.rs — LineBuffer holds
text after the last \n until the next \n arrives, so partial code
fences never become visible state.
- Wired into BlockState in streaming/mod.rs. Assistant text gates;
thinking deltas bypass (reasoning stays live).
- 9 unit tests including the partial-fence regression case.
CX#11 (pure-render footer):
- New crates/tui/src/tui/widgets/footer.rs — FooterProps / FooterToast
/ FooterWidget. Pure render of pre-computed props.
- ui.rs::render_footer rewritten to build props once and delegate to
FooterWidget. Visual output identical; existing 10 footer tests
pass unchanged. 5 new from_app tests for the props builder.
908/908 tests pass; cargo clippy --workspace -D warnings clean;
cargo fmt clean.
Adds a structured rlm_query tool for parallel/batched LLM fan-out.
The model calls it with one prompt or up to 16 concurrent prompts;
children dispatch via tokio::join_all against the existing DeepSeek
client. Default child model is deepseek-v4-flash; override per-call
via the model field. Available in Plan / Agent / YOLO. Cost folds
into the session's running total automatically.
Fixes scroll-stuck regression (#56): TranscriptScroll::resolve_top
and scrolled_by now use a three-level fallback chain (same line →
same cell line 0 → nearest cell at-or-before) instead of teleporting
to ToBottom when an anchor cell vanishes.
Loosens command-safety chains (#57): cargo build && cargo test and
similar chains of known-safe commands now escalate to RequiresApproval
instead of being hard-blocked as Dangerous. Chains containing unknown
commands still block.
Suppresses the GettingCrowded footer chip — context-percent header
already covers conversation pressure.
Refactors:
- Extracts file_mention parsing/completion/expansion (~450 LOC) from
the 5,500-line ui.rs into crates/tui/src/tui/file_mention.rs.
- Deletes truly unused helpers (write_bytes, timestamped_filename,
extension_from_url, output_path, has_project_doc, primary_doc_path).
Tests: 853 pass. cargo clippy --workspace -D warnings clean.
cargo fmt --all -- --check clean.
Closes#46#47#48#49#50#53#54#55#56#57#58.
Captures the full RLM-fundamental story across the design doc, MODES.md,
and the Hetun prompt. Tracking issues are now #46–#55 (helper layer
filed as #53, Hetun as #54, vendoring as #55).
What this nails down:
- **Hetun mode** is added at the END of the Tab cycle (Plan → Agent →
YOLO → Hetun → Plan), not as a Plan replacement. Default landing mode
is unchanged so people don't accidentally start there. Plan stays as
it is.
- **Mission-level approval, not block-level.** Hetun runs a research
phase, presents one mission card, and only executes after explicit
user approval. Inside the execution turn the repl block runs straight
through with no per-block prompts — that's the whole point of the
mode.
- **The user's configured model is left alone on enter/exit.** Pro/max
users stay on Pro/max. The flash-as-coordinator behaviour is internal
to the runtime (ZIGRLM_RLM_CMD always points to flash regardless of
mode). No global model swap.
- **No /hetun slash command.** Tab cycles into the mode; /plan keeps
switching to Plan as today.
- **The helper layer (#53) is fundamental, not aleph-derived.** A
curated ~20-function ctx-helper module + AST-validated Python sandbox
baked into the repl runtime so a single block can load → slice → fan
out flash queries → aggregate without crossing tool boundaries.
Inspired by aleph's pattern but our own native primitive — not a port.
- **Hetun research methodology adopts Sakana's Fugu patterns.** The
research phase is recursive novelty sampling + hierarchical narrative
tree synthesis + multi-detector cross-verification (flash for
breadth, Pro for depth) + hypothesis-verification loop. Not "fan out
8 fixed queries". This is what makes "Plan + Recursive Agents"
meaningful versus a flash-coordinator wrapper.
- **No version-number framing anywhere.** The plan ships as one cohesive
RLM landing across #46/#48/#49/#50/#53/#54/#55 — order is dependency,
not release schedule. We keep shipping.
- **Auto-compaction stays automatic.** Removed a manual /compact nag
from the Hetun prompt; the existing coherence + capacity system
already handles this.
Files:
docs/rlm-design.md new — full design doc with Hetun details
docs/research-react-vs-rlm.md new — supporting research treatment
docs/MODES.md 4-mode cycle, Hetun added at end, Plan kept
crates/tui/src/prompts/hetun.txt prompt teaching the recursive-novelty
+ hierarchical-synthesis + verification-loop
rhythm, mission-card structure, two-step gate
.gitignore ignore .claude/scheduled_tasks.lock runtime
Closes nothing yet — implementation lands across the tracking issues.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The Release / parity job runs with `--locked` and clippy `-D warnings`.
Main CI ran without either flag, so commits could pass main CI but fail
the release pipeline at the parity stage — which has been silently
blocking every npm publish since v0.4.6 (latest npm = 0.4.8 even though
git tags reach v0.5.2). Most failures were either fmt drift caused by
new stable rustc / rustfmt revisions or lockfile drift the workspace CI
never noticed.
Aligns the Lint job's clippy step with `--locked -- -D warnings` and
the Test job with `--locked` + an explicit `git diff --exit-code --
Cargo.lock` lockfile drift guard. From here on, anything that would
fail Release / parity also fails main CI on the same push, so we
never push a tag we know will fail the publish.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Removed the `[`Settings`]` intra-doc link that referenced a type not
in scope at the module-doc level — RUSTDOCFLAGS=-Dwarnings rejects it.
Replaced with backtick code formatting; the rest of the doc is
unchanged.
Caught by the Documentation CI job on v0.5.2.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Single-feature release that lands #39 (the /model two-pane picker)
on top of the v0.5.1 quality-of-life batch. Bumps workspace +
npm wrapper + Cargo.lock in lockstep; check-versions.sh verifies.
Closes#39.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`/model` with no argument now opens a two-pane modal: model on the left
(deepseek-v4-pro flagship vs deepseek-v4-flash fast-and-cheap, with the
current id appearing as a "current (custom)" row when it isn't one of
the listed defaults), and thinking effort on the right (Off, High,
Max). Tab/←/→ swaps panes, ↑/↓ moves within the focused pane, Enter
applies both, Esc cancels.
Effort exposes only the three rows DeepSeek behaviorally distinguishes
per the Thinking Mode docs — `low`/`medium` are mapped server-side to
`high`, and `xhigh` to `max`, so listing them as separate choices would
mislead. The legacy variants stay valid in `~/.deepseek/settings.toml`
for back-compat (the existing `cycle_next` already only visits
Off→High→Max), the picker just doesn't surface them.
Apply path:
* mutates app.model and app.reasoning_effort
* resets last_*_tokens / cache / replay-token gauges so the next-turn
footer numbers reflect the new model rather than stale ones
* persists `default_model` and `reasoning_effort` to settings via the
existing Settings::set/save flow so the choice survives restart
* forwards Op::SetModel + Op::SetCompaction to the engine so the
running session picks up the new compaction budget
* surfaces a one-line summary describing what changed
* if persistence fails, the in-memory change still applies and a
"(not persisted: ...)" suffix is appended to the status line
`/model <id>` keeps working unchanged for power users; only the
no-argument branch was redirected to the new modal.
Files:
* tui/model_picker.rs — new ModelPickerView struct + ModalView impl,
plus eight unit tests (initial state, low/medium normalisation,
custom model preservation, arrow navigation, focus toggle, Enter
emits ModelPickerApplied with the right values, Esc closes silently,
and a guard that the picker exposes exactly off/high/max).
* tui/views/mod.rs — adds ModalKind::ModelPicker and
ViewEvent::ModelPickerApplied carrying both new and previous
model+effort so the handler can describe the diff.
* tui/app.rs — adds AppAction::OpenModelPicker.
* commands/core.rs — `/model` no-arg branch now returns
AppAction::OpenModelPicker; `/model <id>` shortcut is unchanged.
* tui/ui.rs — pushes ModelPickerView on the action and adds
apply_model_picker_choice() that handles persistence + engine sync
when ViewEvent::ModelPickerApplied fires.
* tui/mod.rs — registers the new submodule.
Closes#39 (against v0.5.2).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Bumps workspace version to 0.5.1 and finalises the changelog with the
issues that landed since v0.5.0:
* #25, #27, #31, #33, #34 (already on main)
* #28 @file Tab-completion
* #29 per-workspace trust list with /trust slash command
* #30 reasoning-replay token chip in the footer
* #36 regression tests for sidebar gutter bleed
scripts/release/check-versions.sh is green: workspace=0.5.1, npm=0.5.1,
Cargo.lock in sync.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two related TUI affordances bundled because they share ui.rs and the
ui/tests.rs file.
#30 — Reasoning-content replay telemetry, end-to-end:
* models.rs — Usage gains reasoning_replay_tokens: Option<u32>.
* client.rs — sanitize_thinking_mode_messages now returns the
approximate replay-token count (~4 chars/token); the streaming
pipeline overlays it onto the parsed MessageDelta usage so the
server-reported and client-estimated numbers reach the engine
together.
* app.rs — App stores last_reasoning_replay_tokens.
* ui.rs — TurnComplete handler copies the value into the App; new
footer_reasoning_replay_spans renders an `rsn N.Nk` chip in the
footer next to the cache hit-rate, warning-coloured when replay
tokens exceed 50% of the input budget.
* ui/tests.rs — covers chip-on, chip-hidden-when-zero, and the
sanitizer's None-on-non-thinking-model path.
#28 — Tab-complete @file mentions against the workspace:
* ui.rs — adds partial_file_mention_at_cursor (with a guard against
`user@example.com`-style false positives) and
try_autocomplete_file_mention. Walks the workspace via the
existing ignore::WalkBuilder, ranks prefix matches above
substring matches, applies the unique match outright, extends to
the longest common prefix when multiple match, and surfaces
ambiguous candidates via the status line. Wired into the existing
Tab handler after the slash-command branch.
* ui/tests.rs — covers cursor-inside-mention extraction, email
guard, prefix vs substring ranking, single-match application,
common-prefix extension, no-match status, and the
no-mention-no-op path.
The mention-expansion path that ships file contents to the model is
unchanged — this is purely a discovery aid for typing the path.
Inline-contents and a fuzzy popup picker are queued for v0.5.2.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a persistent allowlist of external paths the agent may read/write
from outside the current workspace, scoped to the workspace it was
granted in. The list lives in ~/.deepseek/workspace-trust.json with
schema {"workspaces": {"<ws>": ["<trusted>", ...]}}; canonical paths on
both sides keep symlink-aliased macOS tempdirs sane.
Surface area:
* crates/tui/src/workspace_trust.rs — new module: load_for / add /
remove plus *_at variants for tests that need an explicit file path
rather than HOME mutation.
* tools/spec.rs — ToolContext gains trusted_external_paths and
resolve_path consults it before returning PathEscape, both for the
existing-path branch and the to-be-created (parent-canonical) branch.
* core/engine.rs — build_tool_context loads the trust snapshot on every
tool dispatch so /trust mutations apply on the next call.
* commands/config.rs — /trust now takes subcommands (add, remove,
list, on, off, status) instead of being a single all-or-nothing
toggle. Tilde expansion handled in-line.
* commands/mod.rs — registry entry updated with the new usage string
and a dispatcher that forwards args.
* tools/diagnostics.rs — adds trusted_external_paths to the JSON
output so the agent and the user can see the list at a glance.
The interactive "Allow once / Always allow / Deny" prompt that the
issue describes is deferred — for v0.5.1 the workflow is "grant
ahead with /trust add". A future change will add a hook in
ToolContext::resolve_path that surfaces an ApprovalRequest when an
escape path is hit, so the slash-command remains the durable
mechanism while the prompt becomes the discovery one.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds two snapshot tests against ChatWidget rendering to lock in that long
single-line tool results never write any cells outside chat_area at the
widths reported in the bug (80, 120, 165, 200 cols), and that the
scrollbar coexists with content along the right edge instead of
overdrawing the penultimate column. The acceptance criterion in the
issue specifically requires this regression coverage; the tests pass
against current code, so existing rendering is the baseline being
guarded.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
A user couldn't find an `NVIDIA_API_KEY` block in `~/.deepseek/config.toml`
because the example file only mentioned NIM as commented-out alternates
to the top-level keys. Two fixes:
- `config.example.toml` now has explicit `[providers.deepseek]` and
`[providers.nvidia_nim]` sections (placed after all top-level keys so
the TOML still parses cleanly), each documenting `api_key` /
`base_url` / `model` plus the env vars that override them. Both
providers can be stored at once and toggled via `/provider` or
`--provider` without re-entering keys.
- `setup --status` "missing api_key" message is now provider-aware: on
`nvidia-nim` it points at `NVIDIA_API_KEY` + `[providers.nvidia_nim]`
+ `deepseek auth set --provider nvidia-nim`, instead of the
DeepSeek-only hint.
Audit verified: the v0.5.0 multi-turn replay fix path
(`should_replay_reasoning_content` → `requires_reasoning_content` in
`crates/tui/src/client.rs:1796`) keys off the model name (matches
`deepseek-v4`), not the provider, so NIM-hosted V4 models get the
replay automatically. No NIM-specific 400-class regression there.
Closes#37 (docs/UX); the live multi-turn-against-NIM verification
remains a manual smoke step listed in the issue (no NIM creds in CI).
Minimum-viable version of the handoff artifact described in #32:
- New `HANDOFF_RELATIVE_PATH = ".deepseek/handoff.md"` convention.
- `system_prompt_for_mode_with_context` now reads that path on every
prompt rebuild and prepends a `## Previous Session Handoff` block to
the system prompt when the file is non-empty. A fresh agent gets the
prior session's blockers/decisions/files-touched in turn-1 context
with zero discovery cost.
- Agent prompt updated to make the convention explicit: "if the block
appears, read it first; before exit/`/compact`, write or update it
via `write_file`."
- `.deepseek/` is already gitignored, so the handoff travels with the
workspace but doesn't pollute commits unless the user opts in.
Tests cover: present-and-non-empty (block injected with file content),
missing file (no block), empty/whitespace-only file (no block). A
unique marker in the injected block (`"left a handoff at .deepseek/..."`)
discriminates the actual block from the agent prompt's own description
of the convention.
Out of scope for v0.5.1: a `/handoff` slash command, a startup banner
toast, automatic write on exit, and the diff-against-HEAD-on-resume
mechanism. The agent can already write the file via `write_file` when
the user types `write a session handoff`.
Closes#32.
Tool-surface audit pass:
- FILE OPERATIONS rewritten so each line states the niche, not just the
verb. read_file mentions PDF auto-extraction + `pages` slicing.
- New SEARCH section consolidates grep_files / file_search / web_search /
fetch_url so the model sees them next to each other and picks the
right one. fetch_url (#33) added; previously absent from the prompt.
- request_user_input pulled out of FILE OPERATIONS into its own USER
section — it never belonged there.
- SUB-AGENTS list shrinks by 3: drops `spawn_agent` (use `agent_spawn`),
`close_agent` (use `agent_cancel`), and the `agent_assign /
assign_agent` dual-name. The underlying dispatchers still resolve those
names, so existing sessions don't break — they just no longer
pollute the model's tool list.
Adds `docs/TOOL_SURFACE.md` with the rationale, the v0.5.1 final
surface, and the dropped aliases. Calls out that grep_files is pure-Rust
(no rg/grep shell-out, so the "fall back to grep" AC from #35 is
vacuously satisfied — the tool has no shell dependency to fall back from).
Closes#35.
The reasoning-effort tier (`max` chip + whale icon) and the live/context
indicators were the only signals on the right of the header. Switching
to nvidia-nim left the right-hand side identical to a DeepSeek session,
so it wasn't obvious at a glance that requests were going to a different
backend.
Now: when `app.api_provider != Deepseek`, the header surfaces a bold
`NIM` chip on the right, leftmost in the chip cluster (so it survives
the narrow-width fallback variants in `right_spans`). Default-DeepSeek
sessions are unchanged — `provider_label = None` short-circuits the
chip.
Closes#38.
`read_file` now detects PDFs by extension or `%PDF-` magic bytes and
shells out to `pdftotext -layout` (poppler) to return plain text
directly to the model. New optional `pages` arg accepts `N` or `N-M`
slices so big papers can be read in pieces without burning context.
When `pdftotext` isn't on `$PATH`, the tool returns a structured
`{type: "binary_unavailable", kind: "pdf", reason, hint}` payload with
install hints (`brew install poppler` / `apt install poppler-utils`)
instead of crashing or returning UTF-8 garbage from a binary file.
Tests cover extension detection (case-insensitive), magic-byte sniffing
on extension-less files, the negative case for plain text, the pages
arg parser (single, range, whitespace, invalid forms), and the
binary_unavailable branch when `pdftotext` is absent.
.docx / .epub / .html stripping deferred — same dispatch can take more
extractors later.
Closes#34.
Complements `web_search` for cases where the URL is already known —
GitHub repos, blog posts, spec pages — and a search-engine round trip is
overkill or actively unhelpful (which #25 had been making worse).
Surface:
- `fetch_url(url, format?, max_bytes?, timeout_ms?)`
- `format`: `markdown` (default), `text`, `raw`
- HTTPS preferred, http:// allowed; non-http schemes rejected up front
- Follows up to 5 redirects; 1 MB default cap (10 MB hard ceiling); 15 s
default timeout (60 s ceiling)
- HTML responses are stripped to readable text via the same regex
pattern used by `web_search` (script/style strip → tag strip → entity
decode → whitespace collapse)
- 4xx / 5xx responses still return the body (with `success: false`) so
the caller can read JSON error envelopes
Capabilities: `ReadOnly + Network`. Approval: `Auto` (matches
`web_search`). Registered in `with_web_tools` so it's available wherever
`web_search` is.
Tests cover: format parsing aliases, scheme rejection, missing/empty
url validation, html-to-text stripping. The over-the-wire cases
(redirect chains, oversized truncation) are exercised by integration
tests once the test suite is wired to a local mock HTTP server —
deferring that since the unit tests already lock in the input
validation and HTML processing.
Closes#33.
The thinking-mode sanitizer now sums the byte size of every replayed
`reasoning_content` field in the outgoing chat-completions body and
emits an `info`-level log line:
Reasoning-content replay: 7 assistant message(s), ~3.2K input tokens (12,884 chars) being re-sent in this request
This is visible under `RUST_LOG=deepseek_tui=info` (or higher). It's the
first step toward the footer/status-line indicator described in #30 —
the model's input-side reasoning replay is now observable per turn,
even before it gets a dedicated UI surface.
Tests cover both branches: bodies that already have reasoning_content
(count is summed across all assistant turns) and bodies where the
sanitizer had to inject the `(reasoning omitted)` placeholder (the
placeholder bytes are included in the count since they ship over the
wire).
Footer integration deferred — that needs a new event from client → engine
→ TUI to surface the count alongside `cache N%` / `$X.XX`. Part of #30
remains open.
Adds scripts/release/check-versions.sh and a `versions` CI job that runs
on every push/PR. Verifies:
- no per-crate Cargo.toml carries a literal version (must inherit the
workspace version)
- npm/deepseek-tui/package.json matches the workspace version
- Cargo.lock is in sync with the manifests
Closes#31.
The #27 per-mode context budget commit (1be18e69) replaced calls to
compaction_threshold_for_model with compaction_threshold_for_model_and_effort
but left the old name in the import list, which fails under -Dwarnings on
Build, Test, and the npm wrapper smoke job. Also re-runs cargo fmt over
the four files the lint job flagged.
- README: drop the `cargo install deepseek-tui` / `deepseek-tui-cli`
block and the crates.io badge — those packages have been lagging the
workspace version for several releases. Source install (`cargo install
--path crates/tui`) remains documented for hackers.
- README: add a "What's new in v0.5.0" section pointing at the
thinking-mode tool-call 400 fix and the #25 web.run cleanup.
- CHANGELOG: add the [0.5.0] - 2026-04-25 entry covering the per-message
reasoning_content rule, the wire-payload sanitizer, and #25.
Add compaction_threshold_for_model_and_effort() with mode-aware soft
caps based on DeepSeek V4 paper Figure 9 recall-quality data:
Plan / off -> 64K (paper eval: 8K-128K)
Agent / high -> 192K (paper eval: 128K)
YOLO / max -> 384K (paper eval: 384K-512K)
Previously, the 80%-of-window rule gave 800K for V4's 1M window,
which is well past the point where MRCR MMR collapses (0.49 at 1M).
Non-V4 models keep the legacy 80% rule. None/unknown effort defaults
to agent-tier (192K).
Replace all web.run mentions with web_search in prompt files (base,
agent, yolo, plan, normal) and update web_search.rs description.
Model was trying web.run which doesn't exist, wasting turns on
validation errors. Also remove [cite:ref_id] citation format which
required web.run's ref_id system.
Partial fix for #25 — web_search reliability improvements (real
search provider) still needed.
### Fixed
- DeepSeek thinking-mode tool-call rounds now always replay reasoning_content
in all subsequent requests (including across new user turns), matching the
documented API contract that assistant tool-call messages must retain their
reasoning content forever. Previously, reasoning_content was cleared after
the current user turn completed, which could cause HTTP 400 errors.
- Missing reasoning_content on a tool-call assistant message now substitutes
a safe placeholder ("(reasoning omitted)") instead of dropping the tool
calls and their matching tool results, preventing orphaned conversation
chains and API 400 rejections.
- Session checkpoint now persists a Thinking-block placeholder for tool-call
turns that produced no streamed reasoning text, keeping on-disk sessions
structurally correct for subsequent requests.
- Token estimation for compaction now counts thinking tokens across ALL
tool-call rounds (not just the current user turn), aligning with the
updated reasoning_content replay rule.
### Changed
- Internal crate dependency pins bumped 0.4.5 → 0.4.9 to match workspace.
- npm wrapper version and deepseekBinaryVersion bumped to 0.4.9.
- README fully rewritten: clearer feature highlights, V4 model focus,
keyboard shortcut table, improved docs index, and more engaging layout.
- CHANGELOG entry for 0.4.9 with comparison URLs.
Bump workspace version 0.4.6 -> 0.4.7 and ship the bug fix flagged by Devin
on PR #18: an uncommented `DEEPSEEK_API_KEY=` line in .env.example caused
`cp .env.example .env` to load an empty key, which `apply_env_overrides`
then propagated into the config and `Config::validate()` rejected on
startup with "api_key cannot be empty string".
- .env.example: comment out the empty `DEEPSEEK_API_KEY=` placeholder.
- crates/tui/src/config.rs: skip empty `DEEPSEEK_API_KEY` env values in
`apply_env_overrides`, matching the facade's empty-string filter.
- Add `apply_env_overrides_ignores_empty_api_key` regression test.
- Bump Cargo.toml workspace version, npm wrapper version + binary version,
and Cargo.lock.
Bundles new work since v0.4.4: NIM provider support, file mention
attachments, provider switch UX, setup --status/--clean/--tools/--plugins,
doctor --json, and protocol-recovery hardening that strips fake
tool-call wrappers from streaming model output.
Local main's unpushed commits had removed DEFAULT_TEXT_MODEL from the
crate::config import in main.rs, but the merged branch's new code at
two call sites still uses it. Textual three-way merge took the local
import line and the branch's call sites, producing a build break.
Re-add the symbol to the import.
Hunter Bown