feat(runtime-api): daemon API quartet for whalescale (#561 #562 #563 #564) by Hmbown · Pull Request #567 · Hmbown/CodeWhale

Hmbown · 2026-05-04T06:59:08Z

Summary

Bridge work to unblock whalescale-desktop's Composer/Settings/Archived-chats flows. Each endpoint solves a current daemon-side limitation that whalescale was either working around (per-turn override) or shipping as cut (Settings → Usage / Archived chats). All four endpoints land together because they share the same runtime_api.rs route table and runtime_threads.rs data model.

runtime API: CORS allowed-origins should be configurable (currently hardcoded) #561 / whalescale#255 — CORS allow-list configurable. New [runtime_api] cors_origins config field, --cors-origin URL (repeatable) flag on deepseek serve --http, and DEEPSEEK_CORS_ORIGINS env var. User entries stack on top of built-in defaults (localhost:3000, localhost:1420, tauri://localhost).
runtime API: PATCH /v1/threads/{id} should accept more fields than just archived #562 / whalescale#256 — PATCH /v1/threads/{id} extended from archived-only to the full editable field set: allow_shell, trust_mode, auto_approve, model, mode, title, system_prompt. Empty-string clears title/system_prompt. New title field on ThreadRecord is additive — no schema_version bump (old readers ignore it without misinterpretation).
runtime API: list-archived-only filter for /v1/threads #563 / whalescale#260 — archived_only=true query param on GET /v1/threads and /v1/threads/summary. Backed by a new ThreadListFilter enum.
runtime API: GET /v1/usage aggregation endpoint #564 / whalescale#261 — GET /v1/usage?since=&until=&group_by=<day|model|provider|thread> aggregates token totals + cost (via pricing.rs) across all threads/turns. Empty time ranges yield empty buckets (never 404).

docs/RUNTIME_API.md and config.example.toml updated. After this merges I will comment the new endpoint shapes on whalescale#255/256/260/261 so the desktop side can wire them.

Test plan

🤖 Generated with Claude Code

…564) Bridge work to unblock whalescale-desktop's Settings/Composer/Archived-chats flows without requiring a daemon recompile per dev-port or client-side aggregation. #561 / whalescale#255 — CORS allow-list configurable * Add `[runtime_api] cors_origins` config field, `--cors-origin URL` (repeatable) flag on `deepseek serve --http`, and `DEEPSEEK_CORS_ORIGINS` env var. User entries stack on top of the built-in defaults (localhost:3000, localhost:1420, tauri://localhost). Resolution preserves first-seen order and drops empty/duplicate values; invalid HeaderValues log a warning and are skipped. * Refactor `cors_layer()` to read merged origins from `RuntimeApiState`. #562 / whalescale#256 — `PATCH /v1/threads/{id}` accepts the full editable field set * Extend `UpdateThreadRequest` with `allow_shell`, `trust_mode`, `auto_approve`, `model`, `mode`, `title`, `system_prompt`. Each is optional; missing means no change. Empty-string clears `title`/ `system_prompt`. Empty `model`/`mode` rejected with 400. * Add `title: Option<String>` to `ThreadRecord` (additive, no schema bump per documented criteria — old readers ignore the field without misinterpretation). `list_threads_summary` now returns the user-set title when present, falling back to the derived input-summary title. * `thread.updated` event payload now carries a `changes` map with only the fields that actually changed. #563 / whalescale#260 — list-archived-only filter * New `archived_only=true` query param on `GET /v1/threads` and `GET /v1/threads/summary`. Backed by a new `ThreadListFilter` enum (`ActiveOnly` | `IncludeArchived` | `ArchivedOnly`). `archived_only` takes precedence over `include_archived`. Default behavior unchanged. #564 / whalescale#261 — `GET /v1/usage` aggregation * New `RuntimeThreadManager::aggregate_usage` walks all threads/turns, filters by inclusive `since`/`until` RFC 3339 bounds, accumulates token totals + cost (via `pricing::calculate_turn_cost_from_usage`), and groups by `day` (default), `model`, `provider`, or `thread`. * New `GET /v1/usage` route. `since`/`until`/`group_by` query params, `since > until` and unknown `group_by` rejected with 400. Empty time ranges yield empty `buckets` (never 404). 5 new tests cover preflight Allow-Origin echoing for both default and extra origins, the extended PATCH field set + clear-by-empty + 400 paths, the archived_only filter on list + summary endpoints, and the /v1/usage envelope + validation errors. Existing 13 runtime_api tests continue to pass; the parity gates and full workspace test suite are clean. `docs/RUNTIME_API.md` and `config.example.toml` updated to document the new params, body shape, endpoint, and CORS knob. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Copilot

Pull request overview

Adds four runtime-daemon HTTP API enhancements needed by whalescale-desktop (CORS configuration, richer thread PATCH, archived-only listing, and usage aggregation), updating both the runtime data model and the route table/documentation.

Changes:

Make runtime API CORS allow-list extensible via CLI/env/config while preserving built-in dev defaults.
Extend threads API: archived-only filtering for list endpoints; broaden PATCH /v1/threads/{id} to multiple editable fields including title.
Add GET /v1/usage to aggregate token/cost totals and buckets across threads/turns.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
docs/RUNTIME_API.md	Documents new query params, PATCH body shape, usage endpoint, and CORS configuration sources/defaults.
crates/tui/src/runtime_threads.rs	Adds `title`, thread list filter enum, update request fields, and implements usage aggregation over stored turns.
crates/tui/src/runtime_api.rs	Wires new query params, CORS layer configuration, and `/v1/usage` route/handler into the router.
crates/tui/src/main.rs	Adds `--cors-origin` flag and merges CLI/env/config into runtime API options.
crates/tui/src/config.rs	Introduces `[runtime_api]` config table for additional CORS origins and merges it during config overlay.
config.example.toml	Adds documented example section for `[runtime_api] cors_origins`.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

+        let mut buckets: BTreeMap<String, UsageBucket> = BTreeMap::new();
+        let mut totals = UsageTotals::default();
+
+        for thread in self.store.list_threads()? {
+            let turns = self.store.list_turns_for_thread(&thread.id)?;
+            for turn in turns {
+                if let Some(s) = since
+                    && turn.created_at < s
+                {
+                    continue;
+                }
+                if let Some(u) = until
+                    && turn.created_at > u
+                {
+                    continue;
+                }
+                let Some(usage) = turn.usage.as_ref() else {
+                    continue;
+                };
+                let cached = usage.prompt_cache_hit_tokens.unwrap_or(0) as u64;
+                let reasoning = usage.reasoning_tokens.unwrap_or(0) as u64;
+                let input = usage.input_tokens as u64;
+                let output = usage.output_tokens as u64;
+                let cost = crate::pricing::calculate_turn_cost_from_usage(&thread.model, usage)
+                    .unwrap_or(0.0);
+
+                totals.input_tokens += input;
+                totals.output_tokens += output;
+                totals.cached_tokens += cached;
+                totals.reasoning_tokens += reasoning;
+                totals.cost_usd += cost;
+                totals.turns += 1;
+
+                let key = match group_by {
+                    UsageGroupBy::Day => turn.created_at.format("%Y-%m-%d").to_string(),
+                    UsageGroupBy::Model => thread.model.clone(),
+                    UsageGroupBy::Provider => provider_label_for_model(&thread.model).to_string(),
+                    UsageGroupBy::Thread => thread.id.clone(),
+                };
+                let bucket = buckets.entry(key.clone()).or_insert_with(|| UsageBucket {
+                    key,
+                    ..UsageBucket::default()
+                });
+                bucket.input_tokens += input;
+                bucket.output_tokens += output;
+                bucket.cached_tokens += cached;
+                bucket.reasoning_tokens += reasoning;
+                bucket.cost_usd += cost;
+                bucket.turns += 1;
+            }
+        }
+
+        let group_by_str = match group_by {
+            UsageGroupBy::Day => "day",
+            UsageGroupBy::Model => "model",
+            UsageGroupBy::Provider => "provider",
+            UsageGroupBy::Thread => "thread",
+        }
+        .to_string();
+
+        Ok(UsageAggregation {
+            since,
+            until,
+            group_by: group_by_str,
+            totals,
+            buckets: buckets.into_values().collect(),


+    let filter = resolve_thread_filter(query.include_archived, query.archived_only);
    let threads = state
        .runtime_threads
-        .list_threads(query.include_archived.unwrap_or(false), query.limit)
+        .list_threads(filter, query.limit)
        .await



+/// Resolve the user-supplied CORS origins for `deepseek serve --http`.
+///
+/// Sources, in priority order (later sources extend earlier ones):


gemini-code-assist

Code Review

This pull request introduces several enhancements to the runtime API, including configurable CORS origins, extended thread management capabilities (such as title support and filtering), and a new usage aggregation endpoint. My review identified a significant performance bottleneck in the aggregate_usage function due to inefficient file system operations, a need for more robust provider labeling in the usage tracking logic, and a potential inaccuracy in historical cost reporting when thread models are updated. I have provided a code suggestion to improve the provider labeling logic.

gemini-code-assist · 2026-05-04T07:05:08Z

+        for thread in self.store.list_threads()? {
+            let turns = self.store.list_turns_for_thread(&thread.id)?;


The current implementation of aggregate_usage has a significant performance bottleneck. It iterates over all threads and, for each thread, calls list_turns_for_thread, which performs a full directory scan of the turns/ directory. This results in $O(N_{threads} \times M_{total_turns})$ file system operations and JSON parses. For a user with a large number of threads and turns, this will be extremely slow and inefficient.

gemini-code-assist · 2026-05-04T07:05:08Z

+fn provider_label_for_model(model: &str) -> &'static str {
+    if model.starts_with("deepseek-ai/") {
+        "nvidia-nim"
+    } else if model.starts_with("deepseek-") {
+        "deepseek"
+    } else if model.starts_with("openai/") || model.starts_with("anthropic/") {
+        "openrouter"
+    } else {
+        "unknown"
+    }
+}


The provider_label_for_model logic is incomplete and may mislabel usage buckets. For example, OpenRouter and Novita models often use the deepseek/ prefix (e.g., deepseek/deepseek-v4-pro), which would currently be labeled as "unknown". Additionally, the deepseek-ai/ prefix is ambiguous as it is used by both NVIDIA NIM and self-hosted SGLang instances.

Suggested change

fn provider_label_for_model(model: &str) -> &'static str {

if model.starts_with("deepseek-ai/") {

"nvidia-nim"

} else if model.starts_with("deepseek-") {

"deepseek"

} else if model.starts_with("openai/") || model.starts_with("anthropic/") {

"openrouter"

} else {

"unknown"

}

}

fn provider_label_for_model(model: &str) -> &'static str {

if model.starts_with("deepseek-ai/") {

"nvidia-nim"

} else if model.starts_with("deepseek-") {

"deepseek"

} else if model.starts_with("openai/") || model.starts_with("anthropic/") || model.starts_with("deepseek/") {

"openrouter"

} else {

"unknown"

}

}

gemini-code-assist · 2026-05-04T07:05:08Z

+                let cost = crate::pricing::calculate_turn_cost_from_usage(&thread.model, usage)
+                    .unwrap_or(0.0);


Usage cost calculation in aggregate_usage uses the current model of the thread for all historical turns. Since a thread's model can now be changed via PATCH /v1/threads/{id}, this will lead to inaccurate historical cost reporting if a thread was switched between models with different pricing (e.g., switching from pro to flash). To fix this accurately, the model used for each turn should be persisted in the TurnRecord.

@staryxchen

Picks up the v0.8.10 patch release contents: * Daemon API quartet for whalescale-desktop integration (#561-#564, PR #567). * Bug cluster: macOS seatbelt cargo registry (#558), MCP SIGTERM shutdown (#420), Linux PR_SET_PDEATHSIG (#421). * npm install on older glibc fix (#555/#560 via #556 + #565). * Shell cwd workspace-boundary validation (#524). * Memory help/docs polish (#497 via #569). * Onboarding language picker (#566). * Whale nicknames interleaved with Simplified Chinese. First-time contributors credited in CHANGELOG: @staryxchen, @shentoumengxin, @Vishnu1837, @20bytes. Workspace `Cargo.toml`, all 9 internal path-dep version pins, and `npm/deepseek-tui/package.json` all bumped to 0.8.10. `Cargo.lock` regenerated and committed alongside. Verified locally: * cargo fmt --all -- --check * cargo clippy --workspace --all-targets --all-features --locked -- -D warnings * cargo test --workspace --all-features --locked * bash scripts/release/check-versions.sh Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…wn#562 Hmbown#563 Hmbown#564) (Hmbown#567) Bridge work to unblock whalescale-desktop's Settings/Composer/Archived-chats flows without requiring a daemon recompile per dev-port or client-side aggregation. Hmbown#561 / whalescale#255 — CORS allow-list configurable * Add `[runtime_api] cors_origins` config field, `--cors-origin URL` (repeatable) flag on `deepseek serve --http`, and `DEEPSEEK_CORS_ORIGINS` env var. User entries stack on top of the built-in defaults (localhost:3000, localhost:1420, tauri://localhost). Resolution preserves first-seen order and drops empty/duplicate values; invalid HeaderValues log a warning and are skipped. * Refactor `cors_layer()` to read merged origins from `RuntimeApiState`. Hmbown#562 / whalescale#256 — `PATCH /v1/threads/{id}` accepts the full editable field set * Extend `UpdateThreadRequest` with `allow_shell`, `trust_mode`, `auto_approve`, `model`, `mode`, `title`, `system_prompt`. Each is optional; missing means no change. Empty-string clears `title`/ `system_prompt`. Empty `model`/`mode` rejected with 400. * Add `title: Option<String>` to `ThreadRecord` (additive, no schema bump per documented criteria — old readers ignore the field without misinterpretation). `list_threads_summary` now returns the user-set title when present, falling back to the derived input-summary title. * `thread.updated` event payload now carries a `changes` map with only the fields that actually changed. Hmbown#563 / whalescale#260 — list-archived-only filter * New `archived_only=true` query param on `GET /v1/threads` and `GET /v1/threads/summary`. Backed by a new `ThreadListFilter` enum (`ActiveOnly` | `IncludeArchived` | `ArchivedOnly`). `archived_only` takes precedence over `include_archived`. Default behavior unchanged. Hmbown#564 / whalescale#261 — `GET /v1/usage` aggregation * New `RuntimeThreadManager::aggregate_usage` walks all threads/turns, filters by inclusive `since`/`until` RFC 3339 bounds, accumulates token totals + cost (via `pricing::calculate_turn_cost_from_usage`), and groups by `day` (default), `model`, `provider`, or `thread`. * New `GET /v1/usage` route. `since`/`until`/`group_by` query params, `since > until` and unknown `group_by` rejected with 400. Empty time ranges yield empty `buckets` (never 404). 5 new tests cover preflight Allow-Origin echoing for both default and extra origins, the extended PATCH field set + clear-by-empty + 400 paths, the archived_only filter on list + summary endpoints, and the /v1/usage envelope + validation errors. Existing 13 runtime_api tests continue to pass; the parity gates and full workspace test suite are clean. `docs/RUNTIME_API.md` and `config.example.toml` updated to document the new params, body shape, endpoint, and CORS knob. Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

@staryxchen

Picks up the v0.8.10 patch release contents: * Daemon API quartet for whalescale-desktop integration (Hmbown#561-Hmbown#564, PR Hmbown#567). * Bug cluster: macOS seatbelt cargo registry (Hmbown#558), MCP SIGTERM shutdown (Hmbown#420), Linux PR_SET_PDEATHSIG (Hmbown#421). * npm install on older glibc fix (Hmbown#555/Hmbown#560 via Hmbown#556 + Hmbown#565). * Shell cwd workspace-boundary validation (Hmbown#524). * Memory help/docs polish (Hmbown#497 via Hmbown#569). * Onboarding language picker (Hmbown#566). * Whale nicknames interleaved with Simplified Chinese. First-time contributors credited in CHANGELOG: @staryxchen, @shentoumengxin, @Vishnu1837, @20bytes. Workspace `Cargo.toml`, all 9 internal path-dep version pins, and `npm/deepseek-tui/package.json` all bumped to 0.8.10. `Cargo.lock` regenerated and committed alongside. Verified locally: * cargo fmt --all -- --check * cargo clippy --workspace --all-targets --all-features --locked -- -D warnings * cargo test --workspace --all-features --locked * bash scripts/release/check-versions.sh Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Copilot AI review requested due to automatic review settings May 4, 2026 06:59

Copilot started reviewing on behalf of Hmbown May 4, 2026 06:59 View session

Copilot AI reviewed May 4, 2026

View reviewed changes

gemini-code-assist Bot reviewed May 4, 2026

View reviewed changes

Hmbown merged commit 0047b32 into main May 4, 2026
12 checks passed

Hmbown deleted the feat/v0.8.10-daemon-api branch May 4, 2026 07:18

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(runtime-api): daemon API quartet for whalescale (#561 #562 #563 #564)#567

feat(runtime-api): daemon API quartet for whalescale (#561 #562 #563 #564)#567
Hmbown merged 1 commit into
mainfrom
feat/v0.8.10-daemon-api

Hmbown commented May 4, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

gemini-code-assist Bot May 4, 2026

Uh oh!

gemini-code-assist Bot May 4, 2026

Uh oh!

gemini-code-assist Bot May 4, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		for thread in self.store.list_threads()? {
		let turns = self.store.list_turns_for_thread(&thread.id)?;

		let cost = crate::pricing::calculate_turn_cost_from_usage(&thread.model, usage)
		.unwrap_or(0.0);

Conversation

Hmbown commented May 4, 2026

Summary

Test plan

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot May 4, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot May 4, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot May 4, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants