@oh-my-pi/pi-coding-agent 15.5.13 → 15.5.15

package/CHANGELOG.md

@@ -2,6 +2,37 @@
 
 ## [Unreleased]
 
+## [15.5.15] - 2026-05-30
+### Changed
+
+- Enabled the agent loop's tool-call batch cap for Anthropic Claude sessions, cutting oversized streamed tool-use bursts into runnable batches before continuing the conversation.
+
+### Removed
+
+- Removed the `calc` tool (deterministic arithmetic evaluator) and its `calc.enabled` setting. The model can compute via `eval` instead.
+
+### Fixed
+
+- Fixed Anthropic Claude tool-call batching to clear and reapply the Claude-specific batch cap whenever the session model changes
+
+## [15.5.14] - 2026-05-29
+### Added
+
+- Added progress status output for `llm()` calls in `eval`, including the resolved model, tier, and returned character count
+- Added an `llm(prompt, opts)` helper to both `eval` runtimes (JavaScript and Python) for oneshot, stateless LLM calls. `opts.model` selects a tier — `"smol"` (`pi/smol`), `"default"` (the session's active model, falling back to `pi/default`), or `"slow"` (`pi/slow`, with high reasoning effort on reasoning-capable models). Pass `system` for a system prompt and a plain JSON-Schema `schema` to force a structured response (the helper returns the parsed object instead of the completion string). Calls carry no conversation history and expose no agent-visible tools; they route host-side through the existing tool bridge under the reserved name `__llm__` (`packages/coding-agent/src/eval/llm-bridge.ts`).
+
+### Fixed
+
+- Fixed a rewind/restore loop (and a follow-on handoff failure) caused by assistant turns whose tool results are off the resolved conversation path — e.g. selecting such a turn in `/tree`, restoring a session whose head is a mid-batch turn, or branching a new message in right after a turn whose tool calls hadn't resolved on that branch. `buildSessionContext` walks the leaf→root path, so any turn whose `tool_result` children live on a sibling branch (or below the leaf) ends up with **dangling** `tool_use` blocks. `transformMessages` then fabricated one synthetic `"aborted"`/`"No result provided"` result per dangling call plus a `<turn-aborted>` developer note, which both rendered as phantom failed calls on a turn that "hadn't run anything yet" and re-injected the failed batch into the model's context, prompting it to re-issue the batch (the spiral). `buildSessionContext` now rewrites **every** assistant turn on the resolved path that has dangling `tool_use`: it drops the unpaired `tool_use` blocks, drops `redacted_thinking` blocks, and clears `thinking` signatures (the provider encoder then emits them as plain text), dropping a turn entirely if no content remains. Turns whose tool calls *are* paired on the path are left untouched. Stripping the calls alone was insufficient — a *modified* assistant turn that still carried signed `thinking`/`redacted_thinking` was rejected by Anthropic with `messages.N.content.M: 'thinking' or 'redacted_thinking' blocks in the latest assistant message cannot be modified`, which surfaced as `Handoff generation failed: 400` on navigation. Live turns are unaffected — their results persist on the same path before any context rebuild.
+
+- Fixed external extension loading on Windows compiled binaries: bare `@oh-my-pi/pi-*` value imports (e.g. `import { AssistantMessageEventStream } from "@oh-my-pi/pi-ai"`) failed with `Cannot find package '\$bunfs\root\packages\…'` because `legacy-pi-compat.ts` built shim override paths from a hardcoded POSIX `/$bunfs/root/packages` literal. Win32 normalised the leading slash to a backslash and the resulting path never resolved against the real bunfs mount (`<drive>:\~BUN\root\…`). The bunfs package root is now derived from `import.meta.dir`, so override paths stay platform-native on Windows, Linux, and macOS ([#1514](https://github.com/can1357/oh-my-pi/issues/1514)).
+- Fixed the interactive prompt showing no cursor in Ghostty. A prior change wired the editor's cursor mode to a new `getUseTerminalCursorMarker()` (which always reported the *requested* preference) instead of the resolved hardware-cursor visibility, so when Ghostty force-hid the hardware cursor the editor stayed in terminal-cursor (marker-only) mode and drew no glyph — leaving no visible caret with either `showHardwareCursor`/`PI_HARDWARE_CURSOR` value. The editor now follows `ui.getShowHardwareCursor()`: a hidden hardware cursor falls back to the steady software-cursor glyph (which still emits `CURSOR_MARKER` for IME positioning).
+
+### Changed
+
+- Changed the `eval` tool's `display()` JSON tree in the transcript to use the shared `renderJsonTreeLines` renderer (the same one behind tool args, MCP results, and subagent output) instead of its own format. This drops the redundant `Object(N)` / `Array(N)` type labels and the per-output `JSON output N` header in favor of type icons plus bare keys; the `display[N]` header is now shown only when a cell emits more than one `display()` value.
+- Reverted the sticky `Todos` panel task glyphs to the pre-15.5.12 checkbox icons: completed tasks render `theme.checkbox.checked` (not `theme.status.success`) and in-progress tasks render `theme.checkbox.unchecked` (not the running glyph). Removed the animated spinner entirely — in-progress tasks and pending tasks with a matching in-flight subagent still highlight via the `accent` colour, but the panel now paints once per state change instead of on an 80 ms timer. Subagent auto-checkmarking, the advancing window (`selectStickyTodoWindow`), `todoMatchesAnyDescription` highlighting, and the all-done close animation are unchanged.
+
 ## [15.5.13] - 2026-05-29
 ### Breaking Changes
 

package/dist/types/config/model-registry.d.ts

@@ -78,7 +78,7 @@ export declare const ModelsConfigFile: ConfigFile<{
         authHeader?: boolean | undefined;
         auth?: "apiKey" | "none" | "oauth" | undefined;
         discovery?: {
-            type: "llama.cpp" | "lm-studio" | "ollama" | "openai-models-list";
+            type: "llama.cpp" | "lm-studio" | "ollama" | "openai-models-list" | "proxy";
         } | undefined;
         models?: {
             id: string;

package/dist/types/config/models-config-schema.d.ts

@@ -170,6 +170,7 @@ export declare const ProviderDiscoverySchema: z.ZodObject<{
         "lm-studio": "lm-studio";
         ollama: "ollama";
         "openai-models-list": "openai-models-list";
+        proxy: "proxy";
     }>;
 }, z.core.$strip>;
 export declare const ProviderAuthSchema: z.ZodEnum<{
@@ -259,6 +260,7 @@ export declare const ModelsConfigSchema: z.ZodObject<{
                 "lm-studio": "lm-studio";
                 ollama: "ollama";
                 "openai-models-list": "openai-models-list";
+                proxy: "proxy";
             }>;
         }, z.core.$strip>>;
         models: z.ZodOptional<z.ZodArray<z.ZodObject<{

package/dist/types/config/settings-schema.d.ts

@@ -23,7 +23,7 @@ export declare const TAB_METADATA: Record<SettingTab, {
     icon: `tab.${string}`;
 }>;
 /** Status line segment identifiers */
-export type StatusLineSegmentId = "pi" | "model" | "mode" | "path" | "git" | "pr" | "subagents" | "token_in" | "token_out" | "token_total" | "token_rate" | "cost" | "context_pct" | "context_total" | "time_spent" | "time" | "session" | "hostname" | "cache_read" | "cache_write" | "session_name" | "usage";
+export type StatusLineSegmentId = "pi" | "model" | "mode" | "path" | "git" | "pr" | "subagents" | "token_in" | "token_out" | "token_total" | "token_rate" | "cost" | "context_pct" | "context_total" | "time_spent" | "time" | "session" | "hostname" | "cache_read" | "cache_write" | "cache_hit" | "session_name" | "usage";
 /** Submenu choice metadata. */
 export type SubmenuOption<V extends string = string> = {
     value: V;
@@ -2375,15 +2375,6 @@ export declare const SETTINGS_SCHEMA: {
             readonly description: "Enable the debug tool for DAP-based debugging";
         };
     };
-    readonly "calc.enabled": {
-        readonly type: "boolean";
-        readonly default: false;
-        readonly ui: {
-            readonly tab: "tools";
-            readonly label: "Calculator";
-            readonly description: "Enable the calculator tool for basic calculations";
-        };
-    };
     readonly "tts.enabled": {
         readonly type: "boolean";
         readonly default: false;

package/dist/types/eval/__tests__/llm-bridge.test.d.ts

@@ -0,0 +1 @@
+export {};

package/dist/types/eval/llm-bridge.d.ts

@@ -0,0 +1,25 @@
+import type { ToolSession } from "../tools";
+import type { JsStatusEvent } from "./js/shared/types";
+/** Synthetic bridge name reserved for the `llm()` helper across both runtimes. */
+export declare const EVAL_LLM_BRIDGE_NAME = "__llm__";
+type LlmTier = "smol" | "default" | "slow";
+export interface EvalLlmBridgeOptions {
+    session: ToolSession;
+    signal?: AbortSignal;
+    emitStatus?: (event: JsStatusEvent) => void;
+}
+export interface EvalLlmResult {
+    text: string;
+    details: {
+        model: string;
+        tier: LlmTier;
+        structured: boolean;
+    };
+}
+/**
+ * Run a single stateless completion on behalf of an eval cell's `llm()` call.
+ * Returns a `{ text, details }` value shaped like a {@link callSessionTool}
+ * result so the existing bridge transport carries it to either runtime.
+ */
+export declare function runEvalLlm(args: unknown, options: EvalLlmBridgeOptions): Promise<EvalLlmResult>;
+export {};