@code-yeongyu/senpi 2026.8.18 → 2026.8.20-2

package/dist/modes/interactive/split-trailing-assistant-text.d.ts

@@ -0,0 +1,7 @@
+import type { AssistantMessage, TextContent, ThinkingContent } from "@earendil-works/pi-ai/compat";
+export type TrailingAssistantSplit = {
+    head: AssistantMessage["content"];
+    tail: Array<TextContent | ThinkingContent>;
+};
+export declare function splitTrailingAssistantContent(content: AssistantMessage["content"]): TrailingAssistantSplit;
+//# sourceMappingURL=split-trailing-assistant-text.d.ts.map

package/dist/modes/interactive/split-trailing-assistant-text.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"split-trailing-assistant-text.d.ts","sourceRoot":"","sources":["../../../src/modes/interactive/split-trailing-assistant-text.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,WAAW,EAAE,eAAe,EAAE,MAAM,8BAA8B,CAAC;AAEnG,MAAM,MAAM,sBAAsB,GAAG;IACpC,IAAI,EAAE,gBAAgB,CAAC,SAAS,CAAC,CAAC;IAClC,IAAI,EAAE,KAAK,CAAC,WAAW,GAAG,eAAe,CAAC,CAAC;CAC3C,CAAC;AAEF,wBAAgB,6BAA6B,CAAC,OAAO,EAAE,gBAAgB,CAAC,SAAS,CAAC,GAAG,sBAAsB,CAiB1G"}

package/dist/modes/interactive/split-trailing-assistant-text.js

@@ -0,0 +1,19 @@
+export function splitTrailingAssistantContent(content) {
+    let lastTool = -1;
+    for (let i = 0; i < content.length; i++) {
+        if (content[i]?.type === "toolCall") {
+            lastTool = i;
+        }
+    }
+    if (lastTool < 0 || lastTool === content.length - 1) {
+        return { head: content, tail: [] };
+    }
+    const tail = content
+        .slice(lastTool + 1)
+        .filter((block) => block.type === "text" || block.type === "thinking");
+    if (tail.length === 0) {
+        return { head: content, tail: [] };
+    }
+    return { head: content.slice(0, lastTool + 1), tail };
+}
+//# sourceMappingURL=split-trailing-assistant-text.js.map

package/dist/modes/interactive/split-trailing-assistant-text.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"split-trailing-assistant-text.js","sourceRoot":"","sources":["../../../src/modes/interactive/split-trailing-assistant-text.ts"],"names":[],"mappings":"AAOA,MAAM,UAAU,6BAA6B,CAAC,OAAoC;IACjF,IAAI,QAAQ,GAAG,CAAC,CAAC,CAAC;IAClB,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QACzC,IAAI,OAAO,CAAC,CAAC,CAAC,EAAE,IAAI,KAAK,UAAU,EAAE,CAAC;YACrC,QAAQ,GAAG,CAAC,CAAC;QACd,CAAC;IACF,CAAC;IACD,IAAI,QAAQ,GAAG,CAAC,IAAI,QAAQ,KAAK,OAAO,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACrD,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC;IACpC,CAAC;IACD,MAAM,IAAI,GAAG,OAAO;SAClB,KAAK,CAAC,QAAQ,GAAG,CAAC,CAAC;SACnB,MAAM,CAAC,CAAC,KAAK,EAA0C,EAAE,CAAC,KAAK,CAAC,IAAI,KAAK,MAAM,IAAI,KAAK,CAAC,IAAI,KAAK,UAAU,CAAC,CAAC;IAChH,IAAI,IAAI,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACvB,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC;IACpC,CAAC;IACD,OAAO,EAAE,IAAI,EAAE,OAAO,CAAC,KAAK,CAAC,CAAC,EAAE,QAAQ,GAAG,CAAC,CAAC,EAAE,IAAI,EAAE,CAAC;AACvD,CAAC","sourcesContent":["import type { AssistantMessage, TextContent, ThinkingContent } from \"@earendil-works/pi-ai/compat\";\n\nexport type TrailingAssistantSplit = {\n\thead: AssistantMessage[\"content\"];\n\ttail: Array<TextContent | ThinkingContent>;\n};\n\nexport function splitTrailingAssistantContent(content: AssistantMessage[\"content\"]): TrailingAssistantSplit {\n\tlet lastTool = -1;\n\tfor (let i = 0; i < content.length; i++) {\n\t\tif (content[i]?.type === \"toolCall\") {\n\t\t\tlastTool = i;\n\t\t}\n\t}\n\tif (lastTool < 0 || lastTool === content.length - 1) {\n\t\treturn { head: content, tail: [] };\n\t}\n\tconst tail = content\n\t\t.slice(lastTool + 1)\n\t\t.filter((block): block is TextContent | ThinkingContent => block.type === \"text\" || block.type === \"thinking\");\n\tif (tail.length === 0) {\n\t\treturn { head: content, tail: [] };\n\t}\n\treturn { head: content.slice(0, lastTool + 1), tail };\n}\n"]}

package/dist/modes/interactive/tips/catalog/dag-tips.d.ts

@@ -0,0 +1,7 @@
+export declare const DAG_TIPS: {
+    id: string;
+    bindings: never[];
+    requiresCommand: string;
+    render: () => string;
+}[];
+//# sourceMappingURL=dag-tips.d.ts.map

package/dist/modes/interactive/tips/catalog/dag-tips.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"dag-tips.d.ts","sourceRoot":"","sources":["../../../../../src/modes/interactive/tips/catalog/dag-tips.ts"],"names":[],"mappings":"AAEA,eAAO,MAAM,QAAQ;;;;;GA+De,CAAC"}

package/dist/modes/interactive/tips/catalog/dag-tips.js

@@ -0,0 +1,57 @@
+export const DAG_TIPS = [
+    {
+        id: "dag.one-keyword-mastery",
+        bindings: [],
+        requiresCommand: "dag",
+        render: () => 'One keyword makes you a master of graph engineering: say "mass-ulw" and your work becomes a dependency graph that schedules itself.',
+    },
+    {
+        id: "dag.what-it-is",
+        bindings: [],
+        requiresCommand: "dag",
+        render: () => 'Trigger "mass-ulw" when some tasks must wait for others - describe the work, get a graph that runs in the right order.',
+    },
+    {
+        id: "dag.parallel-waves",
+        bindings: [],
+        requiresCommand: "dag",
+        render: () => "A graph run executes in waves: everything with no pending dependency starts at once, the rest starts the moment it is unblocked.",
+    },
+    {
+        id: "dag.depends-on",
+        bindings: [],
+        requiresCommand: "dag",
+        render: () => "Say what needs what - review after implementation, docs after both - and the ordering is enforced for you.",
+    },
+    {
+        id: "dag.categories",
+        bindings: [],
+        requiresCommand: "dag",
+        render: () => "Each node in the graph picks its own worker category, so cheap steps stay cheap and hard steps get the strong model.",
+    },
+    {
+        id: "dag.status-view",
+        bindings: [],
+        requiresCommand: "dag",
+        render: () => "Use /dag to watch a run: which nodes finished, which are running, and which one failed.",
+    },
+    {
+        id: "dag.resume",
+        bindings: [],
+        requiresCommand: "dag",
+        render: () => "Graph runs are journaled - if the session dies mid-run it resumes later and never redoes the nodes that already finished.",
+    },
+    {
+        id: "dag.vs-parallel",
+        bindings: [],
+        requiresCommand: "dag",
+        render: () => "Fully independent jobs just need parallel subagents; reach for a graph when the ordering between them is the whole point.",
+    },
+    {
+        id: "dag.scale",
+        bindings: [],
+        requiresCommand: "dag",
+        render: () => "Stop hand-running steps in sequence. Describe the dependencies once and let a dozen agents fan out and rejoin on their own.",
+    },
+];
+//# sourceMappingURL=dag-tips.js.map

package/dist/modes/interactive/tips/catalog/dag-tips.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"dag-tips.js","sourceRoot":"","sources":["../../../../../src/modes/interactive/tips/catalog/dag-tips.ts"],"names":[],"mappings":"AAEA,MAAM,CAAC,MAAM,QAAQ,GAAG;IACvB;QACC,EAAE,EAAE,yBAAyB;QAC7B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,KAAK;QACtB,MAAM,EAAE,GAAG,EAAE,CACZ,qIAAqI;KACtI;IACD;QACC,EAAE,EAAE,gBAAgB;QACpB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,KAAK;QACtB,MAAM,EAAE,GAAG,EAAE,CACZ,wHAAwH;KACzH;IACD;QACC,EAAE,EAAE,oBAAoB;QACxB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,KAAK;QACtB,MAAM,EAAE,GAAG,EAAE,CACZ,kIAAkI;KACnI;IACD;QACC,EAAE,EAAE,gBAAgB;QACpB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,KAAK;QACtB,MAAM,EAAE,GAAG,EAAE,CACZ,4GAA4G;KAC7G;IACD;QACC,EAAE,EAAE,gBAAgB;QACpB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,KAAK;QACtB,MAAM,EAAE,GAAG,EAAE,CACZ,sHAAsH;KACvH;IACD;QACC,EAAE,EAAE,iBAAiB;QACrB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,KAAK;QACtB,MAAM,EAAE,GAAG,EAAE,CAAC,yFAAyF;KACvG;IACD;QACC,EAAE,EAAE,YAAY;QAChB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,KAAK;QACtB,MAAM,EAAE,GAAG,EAAE,CACZ,2HAA2H;KAC5H;IACD;QACC,EAAE,EAAE,iBAAiB;QACrB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,KAAK;QACtB,MAAM,EAAE,GAAG,EAAE,CACZ,2HAA2H;KAC5H;IACD;QACC,EAAE,EAAE,WAAW;QACf,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,KAAK;QACtB,MAAM,EAAE,GAAG,EAAE,CACZ,6HAA6H;KAC9H;CACkC,CAAC","sourcesContent":["import type { TipDefinition } from \"./types.ts\";\n\nexport const DAG_TIPS = [\n\t{\n\t\tid: \"dag.one-keyword-mastery\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dag\",\n\t\trender: () =>\n\t\t\t'One keyword makes you a master of graph engineering: say \"mass-ulw\" and your work becomes a dependency graph that schedules itself.',\n\t},\n\t{\n\t\tid: \"dag.what-it-is\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dag\",\n\t\trender: () =>\n\t\t\t'Trigger \"mass-ulw\" when some tasks must wait for others - describe the work, get a graph that runs in the right order.',\n\t},\n\t{\n\t\tid: \"dag.parallel-waves\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dag\",\n\t\trender: () =>\n\t\t\t\"A graph run executes in waves: everything with no pending dependency starts at once, the rest starts the moment it is unblocked.\",\n\t},\n\t{\n\t\tid: \"dag.depends-on\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dag\",\n\t\trender: () =>\n\t\t\t\"Say what needs what - review after implementation, docs after both - and the ordering is enforced for you.\",\n\t},\n\t{\n\t\tid: \"dag.categories\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dag\",\n\t\trender: () =>\n\t\t\t\"Each node in the graph picks its own worker category, so cheap steps stay cheap and hard steps get the strong model.\",\n\t},\n\t{\n\t\tid: \"dag.status-view\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dag\",\n\t\trender: () => \"Use /dag to watch a run: which nodes finished, which are running, and which one failed.\",\n\t},\n\t{\n\t\tid: \"dag.resume\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dag\",\n\t\trender: () =>\n\t\t\t\"Graph runs are journaled - if the session dies mid-run it resumes later and never redoes the nodes that already finished.\",\n\t},\n\t{\n\t\tid: \"dag.vs-parallel\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dag\",\n\t\trender: () =>\n\t\t\t\"Fully independent jobs just need parallel subagents; reach for a graph when the ordering between them is the whole point.\",\n\t},\n\t{\n\t\tid: \"dag.scale\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dag\",\n\t\trender: () =>\n\t\t\t\"Stop hand-running steps in sequence. Describe the dependencies once and let a dozen agents fan out and rejoin on their own.\",\n\t},\n] satisfies readonly TipDefinition[];\n"]}

package/dist/modes/interactive/tips/catalog/memory-tips.d.ts

@@ -0,0 +1,7 @@
+export declare const MEMORY_TIPS: {
+    id: string;
+    bindings: never[];
+    requiresCommand: string;
+    render: () => string;
+}[];
+//# sourceMappingURL=memory-tips.d.ts.map

package/dist/modes/interactive/tips/catalog/memory-tips.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"memory-tips.d.ts","sourceRoot":"","sources":["../../../../../src/modes/interactive/tips/catalog/memory-tips.ts"],"names":[],"mappings":"AAGA,eAAO,MAAM,WAAW;;;;;GAqGY,CAAC"}

package/dist/modes/interactive/tips/catalog/memory-tips.js

@@ -0,0 +1,100 @@
+import { APP_NAME } from "../../../../config.js";
+export const MEMORY_TIPS = [
+    {
+        id: "memory.persistent",
+        bindings: [],
+        requiresCommand: "memory",
+        render: () => `${APP_NAME} remembers across sessions. Close the terminal, come back tomorrow, and the context is still there.`,
+    },
+    {
+        id: "memory.remember-command",
+        bindings: [],
+        requiresCommand: "remember",
+        render: () => "Use /remember to save something you want kept - a preference, a decision, a fact worth reusing.",
+    },
+    {
+        id: "memory.just-tell-it",
+        bindings: [],
+        requiresCommand: "remember",
+        render: () => 'You can also just say it: "remember that we deploy on Fridays" is enough. No syntax, no file to open.',
+    },
+    {
+        id: "memory.search-command",
+        bindings: [],
+        requiresCommand: "search",
+        render: () => "Use /search to look through everything remembered so far, in plain words.",
+    },
+    {
+        id: "memory.memory-command",
+        bindings: [],
+        requiresCommand: "memory",
+        render: () => "Use /memory to see what is currently remembered about you and this project.",
+    },
+    {
+        id: "memory.init-command",
+        bindings: [],
+        requiresCommand: "init",
+        render: () => "Use /init to start memory for a new project, so it learns this codebase from the first session.",
+    },
+    {
+        id: "memory.people-command",
+        bindings: [],
+        requiresCommand: "people",
+        render: () => "Use /people to keep notes on teammates - who owns what, who to ask, what they prefer.",
+    },
+    {
+        id: "memory.reflect-command",
+        bindings: [],
+        requiresCommand: "reflect",
+        render: () => "Use /reflect to have the session reviewed now and the parts worth keeping written down.",
+    },
+    {
+        id: "memory.dream-command",
+        bindings: [],
+        requiresCommand: "dream",
+        render: () => "Memory tidies itself in the background between turns - /dream shows that work and lets you start it yourself.",
+    },
+    {
+        id: "memory.sleeptime-command",
+        bindings: [],
+        requiresCommand: "sleeptime",
+        render: () => "Use /sleeptime to decide how often memory reorganizes itself while you are not looking.",
+    },
+    {
+        id: "memory.memfs-command",
+        bindings: [],
+        requiresCommand: "memfs",
+        render: () => "Memory is plain files you own - browse them with /memfs and edit anything that looks wrong.",
+    },
+    {
+        id: "memory.repository-command",
+        bindings: [],
+        requiresCommand: "memory-repository",
+        render: () => "Every memory change is a git commit - use /memory-repository to see the history and undo a bad one.",
+    },
+    {
+        id: "memory.doctor-command",
+        bindings: [],
+        requiresCommand: "doctor",
+        render: () => "Use /doctor when memory feels off; it checks the setup and tells you what to fix.",
+    },
+    {
+        id: "memory.facts-command",
+        bindings: [],
+        requiresCommand: "facts",
+        render: () => "Use /facts to see the individual things learned about you, and drop the ones that no longer hold.",
+    },
+    {
+        id: "memory.recompile-command",
+        bindings: [],
+        requiresCommand: "recompile",
+        render: () => "Edited memory files by hand? Run /recompile so this session picks the changes up right away.",
+    },
+    {
+        id: "memory.stop-repeating",
+        bindings: [],
+        requiresCommand: "memory",
+        render: () => "Explain your setup once. Memory means you stop re-explaining it at the start of every session.",
+    },
+];
+//# sourceMappingURL=memory-tips.js.map

package/dist/modes/interactive/tips/catalog/memory-tips.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"memory-tips.js","sourceRoot":"","sources":["../../../../../src/modes/interactive/tips/catalog/memory-tips.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,MAAM,uBAAuB,CAAC;AAGjD,MAAM,CAAC,MAAM,WAAW,GAAG;IAC1B;QACC,EAAE,EAAE,mBAAmB;QACvB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,QAAQ;QACzB,MAAM,EAAE,GAAG,EAAE,CACZ,GAAG,QAAQ,qGAAqG;KACjH;IACD;QACC,EAAE,EAAE,yBAAyB;QAC7B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,UAAU;QAC3B,MAAM,EAAE,GAAG,EAAE,CAAC,iGAAiG;KAC/G;IACD;QACC,EAAE,EAAE,qBAAqB;QACzB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,UAAU;QAC3B,MAAM,EAAE,GAAG,EAAE,CACZ,uGAAuG;KACxG;IACD;QACC,EAAE,EAAE,uBAAuB;QAC3B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,QAAQ;QACzB,MAAM,EAAE,GAAG,EAAE,CAAC,2EAA2E;KACzF;IACD;QACC,EAAE,EAAE,uBAAuB;QAC3B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,QAAQ;QACzB,MAAM,EAAE,GAAG,EAAE,CAAC,6EAA6E;KAC3F;IACD;QACC,EAAE,EAAE,qBAAqB;QACzB,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,MAAM;QACvB,MAAM,EAAE,GAAG,EAAE,CAAC,iGAAiG;KAC/G;IACD;QACC,EAAE,EAAE,uBAAuB;QAC3B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,QAAQ;QACzB,MAAM,EAAE,GAAG,EAAE,CAAC,uFAAuF;KACrG;IACD;QACC,EAAE,EAAE,wBAAwB;QAC5B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,SAAS;QAC1B,MAAM,EAAE,GAAG,EAAE,CAAC,yFAAyF;KACvG;IACD;QACC,EAAE,EAAE,sBAAsB;QAC1B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,OAAO;QACxB,MAAM,EAAE,GAAG,EAAE,CACZ,+GAA+G;KAChH;IACD;QACC,EAAE,EAAE,0BAA0B;QAC9B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,WAAW;QAC5B,MAAM,EAAE,GAAG,EAAE,CAAC,yFAAyF;KACvG;IACD;QACC,EAAE,EAAE,sBAAsB;QAC1B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,OAAO;QACxB,MAAM,EAAE,GAAG,EAAE,CAAC,6FAA6F;KAC3G;IACD;QACC,EAAE,EAAE,2BAA2B;QAC/B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,mBAAmB;QACpC,MAAM,EAAE,GAAG,EAAE,CACZ,qGAAqG;KACtG;IACD;QACC,EAAE,EAAE,uBAAuB;QAC3B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,QAAQ;QACzB,MAAM,EAAE,GAAG,EAAE,CAAC,mFAAmF;KACjG;IACD;QACC,EAAE,EAAE,sBAAsB;QAC1B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,OAAO;QACxB,MAAM,EAAE,GAAG,EAAE,CAAC,mGAAmG;KACjH;IACD;QACC,EAAE,EAAE,0BAA0B;QAC9B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,WAAW;QAC5B,MAAM,EAAE,GAAG,EAAE,CAAC,8FAA8F;KAC5G;IACD;QACC,EAAE,EAAE,uBAAuB;QAC3B,QAAQ,EAAE,EAAE;QACZ,eAAe,EAAE,QAAQ;QACzB,MAAM,EAAE,GAAG,EAAE,CAAC,gGAAgG;KAC9G;CACkC,CAAC","sourcesContent":["import { APP_NAME } from \"../../../../config.ts\";\nimport type { TipDefinition } from \"./types.ts\";\n\nexport const MEMORY_TIPS = [\n\t{\n\t\tid: \"memory.persistent\",\n\t\tbindings: [],\n\t\trequiresCommand: \"memory\",\n\t\trender: () =>\n\t\t\t`${APP_NAME} remembers across sessions. Close the terminal, come back tomorrow, and the context is still there.`,\n\t},\n\t{\n\t\tid: \"memory.remember-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"remember\",\n\t\trender: () => \"Use /remember to save something you want kept - a preference, a decision, a fact worth reusing.\",\n\t},\n\t{\n\t\tid: \"memory.just-tell-it\",\n\t\tbindings: [],\n\t\trequiresCommand: \"remember\",\n\t\trender: () =>\n\t\t\t'You can also just say it: \"remember that we deploy on Fridays\" is enough. No syntax, no file to open.',\n\t},\n\t{\n\t\tid: \"memory.search-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"search\",\n\t\trender: () => \"Use /search to look through everything remembered so far, in plain words.\",\n\t},\n\t{\n\t\tid: \"memory.memory-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"memory\",\n\t\trender: () => \"Use /memory to see what is currently remembered about you and this project.\",\n\t},\n\t{\n\t\tid: \"memory.init-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"init\",\n\t\trender: () => \"Use /init to start memory for a new project, so it learns this codebase from the first session.\",\n\t},\n\t{\n\t\tid: \"memory.people-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"people\",\n\t\trender: () => \"Use /people to keep notes on teammates - who owns what, who to ask, what they prefer.\",\n\t},\n\t{\n\t\tid: \"memory.reflect-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"reflect\",\n\t\trender: () => \"Use /reflect to have the session reviewed now and the parts worth keeping written down.\",\n\t},\n\t{\n\t\tid: \"memory.dream-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"dream\",\n\t\trender: () =>\n\t\t\t\"Memory tidies itself in the background between turns - /dream shows that work and lets you start it yourself.\",\n\t},\n\t{\n\t\tid: \"memory.sleeptime-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"sleeptime\",\n\t\trender: () => \"Use /sleeptime to decide how often memory reorganizes itself while you are not looking.\",\n\t},\n\t{\n\t\tid: \"memory.memfs-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"memfs\",\n\t\trender: () => \"Memory is plain files you own - browse them with /memfs and edit anything that looks wrong.\",\n\t},\n\t{\n\t\tid: \"memory.repository-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"memory-repository\",\n\t\trender: () =>\n\t\t\t\"Every memory change is a git commit - use /memory-repository to see the history and undo a bad one.\",\n\t},\n\t{\n\t\tid: \"memory.doctor-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"doctor\",\n\t\trender: () => \"Use /doctor when memory feels off; it checks the setup and tells you what to fix.\",\n\t},\n\t{\n\t\tid: \"memory.facts-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"facts\",\n\t\trender: () => \"Use /facts to see the individual things learned about you, and drop the ones that no longer hold.\",\n\t},\n\t{\n\t\tid: \"memory.recompile-command\",\n\t\tbindings: [],\n\t\trequiresCommand: \"recompile\",\n\t\trender: () => \"Edited memory files by hand? Run /recompile so this session picks the changes up right away.\",\n\t},\n\t{\n\t\tid: \"memory.stop-repeating\",\n\t\tbindings: [],\n\t\trequiresCommand: \"memory\",\n\t\trender: () => \"Explain your setup once. Memory means you stop re-explaining it at the start of every session.\",\n\t},\n] satisfies readonly TipDefinition[];\n"]}

package/dist/modes/interactive/tips/registry.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"registry.d.ts","sourceRoot":"","sources":["../../../../src/modes/interactive/tips/registry.ts"],"names":[],"mappings":"AAOA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,oBAAoB,CAAC;AAGxD,YAAY,EAAE,aAAa,EAAE,MAAM,oBAAoB,CAAC;AAExD,eAAO,MAAM,eAAe,EAAE,SAAS,aAAa,EASnD,CAAC"}
+{"version":3,"file":"registry.d.ts","sourceRoot":"","sources":["../../../../src/modes/interactive/tips/registry.ts"],"names":[],"mappings":"AASA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,oBAAoB,CAAC;AAGxD,YAAY,EAAE,aAAa,EAAE,MAAM,oBAAoB,CAAC;AAExD,eAAO,MAAM,eAAe,EAAE,SAAS,aAAa,EAWnD,CAAC"}

package/dist/modes/interactive/tips/registry.js

@@ -1,6 +1,8 @@
 import { CLI_TIPS } from "./catalog/cli-tips.js";
+import { DAG_TIPS } from "./catalog/dag-tips.js";
 import { ETHOS_TIPS } from "./catalog/ethos-tips.js";
 import { INPUT_TIPS } from "./catalog/input-tips.js";
+import { MEMORY_TIPS } from "./catalog/memory-tips.js";
 import { MODEL_TIPS } from "./catalog/model-tips.js";
 import { SESSION_TIPS } from "./catalog/session-tips.js";
 import { SETTINGS_TIPS } from "./catalog/settings-tips.js";
@@ -14,6 +16,8 @@ export const TIP_DEFINITIONS = [
     ...SETTINGS_TIPS,
     ...CLI_TIPS,
     ...SUBAGENT_TIPS,
+    ...MEMORY_TIPS,
+    ...DAG_TIPS,
     ...ETHOS_TIPS,
 ];
 //# sourceMappingURL=registry.js.map

package/dist/modes/interactive/tips/registry.js.map

@@ -1 +1 @@
-{"version":3,"file":"registry.js","sourceRoot":"","sources":["../../../../src/modes/interactive/tips/registry.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,MAAM,uBAAuB,CAAC;AACjD,OAAO,EAAE,UAAU,EAAE,MAAM,yBAAyB,CAAC;AACrD,OAAO,EAAE,UAAU,EAAE,MAAM,yBAAyB,CAAC;AACrD,OAAO,EAAE,UAAU,EAAE,MAAM,yBAAyB,CAAC;AACrD,OAAO,EAAE,YAAY,EAAE,MAAM,2BAA2B,CAAC;AACzD,OAAO,EAAE,aAAa,EAAE,MAAM,4BAA4B,CAAC;AAC3D,OAAO,EAAE,aAAa,EAAE,MAAM,4BAA4B,CAAC;AAE3D,OAAO,EAAE,cAAc,EAAE,MAAM,6BAA6B,CAAC;AAI7D,MAAM,CAAC,MAAM,eAAe,GAA6B;IACxD,GAAG,UAAU;IACb,GAAG,UAAU;IACb,GAAG,YAAY;IACf,GAAG,cAAc;IACjB,GAAG,aAAa;IAChB,GAAG,QAAQ;IACX,GAAG,aAAa;IAChB,GAAG,UAAU;CACb,CAAC","sourcesContent":["import { CLI_TIPS } from \"./catalog/cli-tips.ts\";\nimport { ETHOS_TIPS } from \"./catalog/ethos-tips.ts\";\nimport { INPUT_TIPS } from \"./catalog/input-tips.ts\";\nimport { MODEL_TIPS } from \"./catalog/model-tips.ts\";\nimport { SESSION_TIPS } from \"./catalog/session-tips.ts\";\nimport { SETTINGS_TIPS } from \"./catalog/settings-tips.ts\";\nimport { SUBAGENT_TIPS } from \"./catalog/subagent-tips.ts\";\nimport type { TipDefinition } from \"./catalog/types.ts\";\nimport { WORKSPACE_TIPS } from \"./catalog/workspace-tips.ts\";\n\nexport type { TipDefinition } from \"./catalog/types.ts\";\n\nexport const TIP_DEFINITIONS: readonly TipDefinition[] = [\n\t...MODEL_TIPS,\n\t...INPUT_TIPS,\n\t...SESSION_TIPS,\n\t...WORKSPACE_TIPS,\n\t...SETTINGS_TIPS,\n\t...CLI_TIPS,\n\t...SUBAGENT_TIPS,\n\t...ETHOS_TIPS,\n];\n"]}
+{"version":3,"file":"registry.js","sourceRoot":"","sources":["../../../../src/modes/interactive/tips/registry.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,MAAM,uBAAuB,CAAC;AACjD,OAAO,EAAE,QAAQ,EAAE,MAAM,uBAAuB,CAAC;AACjD,OAAO,EAAE,UAAU,EAAE,MAAM,yBAAyB,CAAC;AACrD,OAAO,EAAE,UAAU,EAAE,MAAM,yBAAyB,CAAC;AACrD,OAAO,EAAE,WAAW,EAAE,MAAM,0BAA0B,CAAC;AACvD,OAAO,EAAE,UAAU,EAAE,MAAM,yBAAyB,CAAC;AACrD,OAAO,EAAE,YAAY,EAAE,MAAM,2BAA2B,CAAC;AACzD,OAAO,EAAE,aAAa,EAAE,MAAM,4BAA4B,CAAC;AAC3D,OAAO,EAAE,aAAa,EAAE,MAAM,4BAA4B,CAAC;AAE3D,OAAO,EAAE,cAAc,EAAE,MAAM,6BAA6B,CAAC;AAI7D,MAAM,CAAC,MAAM,eAAe,GAA6B;IACxD,GAAG,UAAU;IACb,GAAG,UAAU;IACb,GAAG,YAAY;IACf,GAAG,cAAc;IACjB,GAAG,aAAa;IAChB,GAAG,QAAQ;IACX,GAAG,aAAa;IAChB,GAAG,WAAW;IACd,GAAG,QAAQ;IACX,GAAG,UAAU;CACb,CAAC","sourcesContent":["import { CLI_TIPS } from \"./catalog/cli-tips.ts\";\nimport { DAG_TIPS } from \"./catalog/dag-tips.ts\";\nimport { ETHOS_TIPS } from \"./catalog/ethos-tips.ts\";\nimport { INPUT_TIPS } from \"./catalog/input-tips.ts\";\nimport { MEMORY_TIPS } from \"./catalog/memory-tips.ts\";\nimport { MODEL_TIPS } from \"./catalog/model-tips.ts\";\nimport { SESSION_TIPS } from \"./catalog/session-tips.ts\";\nimport { SETTINGS_TIPS } from \"./catalog/settings-tips.ts\";\nimport { SUBAGENT_TIPS } from \"./catalog/subagent-tips.ts\";\nimport type { TipDefinition } from \"./catalog/types.ts\";\nimport { WORKSPACE_TIPS } from \"./catalog/workspace-tips.ts\";\n\nexport type { TipDefinition } from \"./catalog/types.ts\";\n\nexport const TIP_DEFINITIONS: readonly TipDefinition[] = [\n\t...MODEL_TIPS,\n\t...INPUT_TIPS,\n\t...SESSION_TIPS,\n\t...WORKSPACE_TIPS,\n\t...SETTINGS_TIPS,\n\t...CLI_TIPS,\n\t...SUBAGENT_TIPS,\n\t...MEMORY_TIPS,\n\t...DAG_TIPS,\n\t...ETHOS_TIPS,\n];\n"]}

package/dist/senpi

@@ -4,9 +4,14 @@ import { spawn } from "node:child_process";
 import { existsSync } from "node:fs";
 import { join } from "node:path";
 import { fileURLToPath } from "node:url";
+import { enableStartupCompileCache } from "./compile-cache.js";
 import { APP_NAME, DISPLAY_VERSION, getPackageDir } from "./config.js";
-import { releaseInheritedInspectorForChild } from "./inspector-policy.js";
+import { hasInheritedInspectorOption, releaseInheritedInspectorForChild } from "./inspector-policy.js";
 import { handleBootstrapSelfUpdate } from "./self-update-bootstrap.js";
+// Must run before cli-main is loaded, by either path: it caches the engine graph this process
+// imports on the fast path below, and it publishes NODE_COMPILE_CACHE so a spawned cli-main child
+// inherits this process's cache directory instead of resolving and re-filling its own.
+enableStartupCompileCache();
 process.title = APP_NAME;
 process.env.PI_CODING_AGENT = "true";
 process.env.AI_AGENT = APP_NAME;
@@ -29,7 +34,21 @@ function isMissingBundledWorkspaceDependencies(packageDir) {
         return !existsSync(join(packageDir, "node_modules", "@earendil-works", name, "dist", "index.js"));
     });
 }
-async function runFullCli() {
+/**
+ * Decide whether the agent needs its own process.
+ *
+ * Two things justify the extra Node process, and only two. An inherited Inspector option means a
+ * debugger socket has to be released here and re-opened over there, which a same-process load
+ * cannot do. Custom exec arguments (`--max-old-space-size`, a loader `--import`, ...) were chosen
+ * for the process that runs the agent, and they are only applied at process start, so they must be
+ * replayed onto a fresh one. Brand scrubbing does NOT justify it: `cli-main` calls
+ * `scrubBrandFromEnvironment()` itself, so loading it here scrubs this process's environment before
+ * anything the agent spawns can inherit it.
+ */
+function requiresIsolatedProcess() {
+    return process.execArgv.length > 0 || hasInheritedInspectorOption();
+}
+async function spawnFullCli() {
     const extension = import.meta.url.endsWith(".ts") ? ".ts" : ".js";
     const fullCliPath = fileURLToPath(new URL(`./cli-main${extension}`, import.meta.url));
     releaseInheritedInspectorForChild();
@@ -60,5 +79,14 @@ if (isMissingBundledWorkspaceDependencies(getPackageDir())) {
         process.exit();
     }
 }
-process.exitCode = await runFullCli();
+if (requiresIsolatedProcess()) {
+    process.exitCode = await spawnFullCli();
+}
+else {
+    // Entry-point process-structure seam: `cli-main` runs `main()` at module scope and owns
+    // `process.exitCode` and any `process.exit()` of its own, so importing it here IS the run - there
+    // is no result to forward. It has to be a dynamic import: a static one would evaluate the whole
+    // engine graph before the `--version` and bootstrap-repair paths above, which answer without it.
+    await import("./cli-main.js");
+}
 //# sourceMappingURL=cli.js.map

package/dist/utils/management-http.d.ts

@@ -4,8 +4,10 @@ export interface FetchRetryOptions {
     maxRetries?: number;
     /** Retry transient HTTP responses as well as transport failures. Defaults to true. */
     retryOnStatus?: boolean;
-    /** Per-attempt timeout. A new timeout is created for every attempt. */
+    /** Overall time budget shared by all attempts. */
     timeoutMs?: number;
+    /** Per-attempt timeout. A new timeout is created for every attempt. */
+    attemptTimeoutMs?: number;
 }
 /**
  * Fetch a management HTTP resource with a bounded immediate retry.
@@ -15,8 +17,8 @@ export interface FetchRetryOptions {
  * agent/model operations: those can fail after the HTTP request starts and are
  * retried by their semantic caller instead.
  *
- * Caller cancellation is terminal. When timeoutMs is supplied, it is the
- * overall time budget shared by all attempts.
+ * Caller cancellation and timeoutMs are terminal. attemptTimeoutMs aborts
+ * only the current attempt so a hung connection can be retried.
  */
 export declare function fetchWithRetry(input: FetchInput, init?: RequestInit | undefined, options?: FetchRetryOptions): Promise<Response>;
 export {};

package/dist/utils/management-http.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"management-http.d.ts","sourceRoot":"","sources":["../../src/utils/management-http.ts"],"names":[],"mappings":"AAAA,KAAK,UAAU,GAAG,UAAU,CAAC,OAAO,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;AAI9C,MAAM,WAAW,iBAAiB;IACjC,gFAAgF;IAChF,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,sFAAsF;IACtF,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,uEAAuE;IACvE,SAAS,CAAC,EAAE,MAAM,CAAC;CACnB;AAED;;;;;;;;;;GAUG;AACH,wBAAsB,cAAc,CACnC,KAAK,EAAE,UAAU,EACjB,IAAI,GAAE,WAAW,GAAG,SAAqB,EACzC,OAAO,GAAE,iBAAsB,GAC7B,OAAO,CAAC,QAAQ,CAAC,CAuCnB"}
+{"version":3,"file":"management-http.d.ts","sourceRoot":"","sources":["../../src/utils/management-http.ts"],"names":[],"mappings":"AAAA,KAAK,UAAU,GAAG,UAAU,CAAC,OAAO,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;AAI9C,MAAM,WAAW,iBAAiB;IACjC,gFAAgF;IAChF,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,sFAAsF;IACtF,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,kDAAkD;IAClD,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,uEAAuE;IACvE,gBAAgB,CAAC,EAAE,MAAM,CAAC;CAC1B;AAED;;;;;;;;;;GAUG;AACH,wBAAsB,cAAc,CACnC,KAAK,EAAE,UAAU,EACjB,IAAI,GAAE,WAAW,GAAG,SAAqB,EACzC,OAAO,GAAE,iBAAsB,GAC7B,OAAO,CAAC,QAAQ,CAAC,CA+CnB"}

package/dist/utils/management-http.js

@@ -7,23 +7,23 @@ const RETRYABLE_STATUS_CODES = new Set([408, 425, 429, 500, 502, 503, 504]);
  * agent/model operations: those can fail after the HTTP request starts and are
  * retried by their semantic caller instead.
  *
- * Caller cancellation is terminal. When timeoutMs is supplied, it is the
- * overall time budget shared by all attempts.
+ * Caller cancellation and timeoutMs are terminal. attemptTimeoutMs aborts
+ * only the current attempt so a hung connection can be retried.
  */
 export async function fetchWithRetry(input, init = undefined, options = {}) {
     const maxRetries = options.maxRetries === undefined || !Number.isFinite(options.maxRetries)
         ? 2
         : Math.max(0, Math.floor(options.maxRetries));
     const retryOnStatus = options.retryOnStatus ?? true;
-    const parentSignal = init?.signal;
+    const parentSignal = init?.signal ?? undefined;
     const timeoutSignal = options.timeoutMs !== undefined && options.timeoutMs > 0 ? AbortSignal.timeout(options.timeoutMs) : undefined;
-    const signal = timeoutSignal
-        ? parentSignal
-            ? AbortSignal.any([parentSignal, timeoutSignal])
-            : timeoutSignal
-        : parentSignal;
+    const attemptTimeoutMs = options.attemptTimeoutMs !== undefined && options.attemptTimeoutMs > 0 ? options.attemptTimeoutMs : undefined;
     for (let attempt = 0;; attempt++) {
-        signal?.throwIfAborted();
+        parentSignal?.throwIfAborted();
+        timeoutSignal?.throwIfAborted();
+        const attemptTimeoutSignal = attemptTimeoutMs ? AbortSignal.timeout(attemptTimeoutMs) : undefined;
+        const signals = [parentSignal, timeoutSignal, attemptTimeoutSignal].filter((signal) => signal !== undefined);
+        const signal = signals.length > 1 ? AbortSignal.any(signals) : signals[0];
         try {
             const response = await fetch(input, signal ? { ...init, signal } : init);
             const shouldRetry = retryOnStatus && RETRYABLE_STATUS_CODES.has(response.status) && attempt < maxRetries;
@@ -38,9 +38,13 @@ export async function fetchWithRetry(input, init = undefined, options = {}) {
             }
         }
         catch (error) {
+            const attemptTimedOut = attemptTimeoutSignal?.aborted === true && !parentSignal?.aborted && !timeoutSignal?.aborted;
             if (parentSignal?.aborted ||
                 timeoutSignal?.aborted ||
-                (error instanceof Error && error.name === "AbortError" && timeoutSignal === undefined) ||
+                (error instanceof Error &&
+                    error.name === "AbortError" &&
+                    !attemptTimedOut &&
+                    timeoutSignal === undefined) ||
                 attempt >= maxRetries) {
                 throw error;
             }

package/dist/utils/management-http.js.map

@@ -1 +1 @@
-{"version":3,"file":"management-http.js","sourceRoot":"","sources":["../../src/utils/management-http.ts"],"names":[],"mappings":"AAEA,MAAM,sBAAsB,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAW5E;;;;;;;;;;GAUG;AACH,MAAM,CAAC,KAAK,UAAU,cAAc,CACnC,KAAiB,EACjB,OAAgC,SAAS,EACzC,UAA6B,EAAE;IAE/B,MAAM,UAAU,GACf,OAAO,CAAC,UAAU,KAAK,SAAS,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,OAAO,CAAC,UAAU,CAAC;QACvE,CAAC,CAAC,CAAC;QACH,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,UAAU,CAAC,CAAC,CAAC;IAChD,MAAM,aAAa,GAAG,OAAO,CAAC,aAAa,IAAI,IAAI,CAAC;IACpD,MAAM,YAAY,GAAG,IAAI,EAAE,MAAM,CAAC;IAClC,MAAM,aAAa,GAClB,OAAO,CAAC,SAAS,KAAK,SAAS,IAAI,OAAO,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,OAAO,CAAC,OAAO,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IAC/G,MAAM,MAAM,GAAG,aAAa;QAC3B,CAAC,CAAC,YAAY;YACb,CAAC,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC,YAAY,EAAE,aAAa,CAAC,CAAC;YAChD,CAAC,CAAC,aAAa;QAChB,CAAC,CAAC,YAAY,CAAC;IAEhB,KAAK,IAAI,OAAO,GAAG,CAAC,GAAI,OAAO,EAAE,EAAE,CAAC;QACnC,MAAM,EAAE,cAAc,EAAE,CAAC;QAEzB,IAAI,CAAC;YACJ,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC,CAAC,EAAE,GAAG,IAAI,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;YACzE,MAAM,WAAW,GAAG,aAAa,IAAI,sBAAsB,CAAC,GAAG,CAAC,QAAQ,CAAC,MAAM,CAAC,IAAI,OAAO,GAAG,UAAU,CAAC;YACzG,IAAI,CAAC,WAAW;gBAAE,OAAO,QAAQ,CAAC;YAClC,IAAI,CAAC;gBACJ,MAAM,QAAQ,CAAC,IAAI,EAAE,MAAM,EAAE,CAAC;YAC/B,CAAC;YAAC,MAAM,CAAC;gBACR,6EAA6E;gBAC7E,wCAAwC;YACzC,CAAC;QACF,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YAChB,IACC,YAAY,EAAE,OAAO;gBACrB,aAAa,EAAE,OAAO;gBACtB,CAAC,KAAK,YAAY,KAAK,IAAI,KAAK,CAAC,IAAI,KAAK,YAAY,IAAI,aAAa,KAAK,SAAS,CAAC;gBACtF,OAAO,IAAI,UAAU,EACpB,CAAC;gBACF,MAAM,KAAK,CAAC;YACb,CAAC;QACF,CAAC;IACF,CAAC;AACF,CAAC","sourcesContent":["type FetchInput = Parameters<typeof fetch>[0];\n\nconst RETRYABLE_STATUS_CODES = new Set([408, 425, 429, 500, 502, 503, 504]);\n\nexport interface FetchRetryOptions {\n\t/** Number of additional attempts after the initial request. Defaults to two. */\n\tmaxRetries?: number;\n\t/** Retry transient HTTP responses as well as transport failures. Defaults to true. */\n\tretryOnStatus?: boolean;\n\t/** Per-attempt timeout. A new timeout is created for every attempt. */\n\ttimeoutMs?: number;\n}\n\n/**\n * Fetch a management HTTP resource with a bounded immediate retry.\n *\n * This is intentionally a transport-level helper for idempotent management\n * requests (version checks, catalogs, and downloads). It must not be used for\n * agent/model operations: those can fail after the HTTP request starts and are\n * retried by their semantic caller instead.\n *\n * Caller cancellation is terminal. When timeoutMs is supplied, it is the\n * overall time budget shared by all attempts.\n */\nexport async function fetchWithRetry(\n\tinput: FetchInput,\n\tinit: RequestInit | undefined = undefined,\n\toptions: FetchRetryOptions = {},\n): Promise<Response> {\n\tconst maxRetries =\n\t\toptions.maxRetries === undefined || !Number.isFinite(options.maxRetries)\n\t\t\t? 2\n\t\t\t: Math.max(0, Math.floor(options.maxRetries));\n\tconst retryOnStatus = options.retryOnStatus ?? true;\n\tconst parentSignal = init?.signal;\n\tconst timeoutSignal =\n\t\toptions.timeoutMs !== undefined && options.timeoutMs > 0 ? AbortSignal.timeout(options.timeoutMs) : undefined;\n\tconst signal = timeoutSignal\n\t\t? parentSignal\n\t\t\t? AbortSignal.any([parentSignal, timeoutSignal])\n\t\t\t: timeoutSignal\n\t\t: parentSignal;\n\n\tfor (let attempt = 0; ; attempt++) {\n\t\tsignal?.throwIfAborted();\n\n\t\ttry {\n\t\t\tconst response = await fetch(input, signal ? { ...init, signal } : init);\n\t\t\tconst shouldRetry = retryOnStatus && RETRYABLE_STATUS_CODES.has(response.status) && attempt < maxRetries;\n\t\t\tif (!shouldRetry) return response;\n\t\t\ttry {\n\t\t\t\tawait response.body?.cancel();\n\t\t\t} catch {\n\t\t\t\t// The response is being discarded before a retry. There is nothing useful to\n\t\t\t\t// do if cancelling its body also fails.\n\t\t\t}\n\t\t} catch (error) {\n\t\t\tif (\n\t\t\t\tparentSignal?.aborted ||\n\t\t\t\ttimeoutSignal?.aborted ||\n\t\t\t\t(error instanceof Error && error.name === \"AbortError\" && timeoutSignal === undefined) ||\n\t\t\t\tattempt >= maxRetries\n\t\t\t) {\n\t\t\t\tthrow error;\n\t\t\t}\n\t\t}\n\t}\n}\n"]}
+{"version":3,"file":"management-http.js","sourceRoot":"","sources":["../../src/utils/management-http.ts"],"names":[],"mappings":"AAEA,MAAM,sBAAsB,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAa5E;;;;;;;;;;GAUG;AACH,MAAM,CAAC,KAAK,UAAU,cAAc,CACnC,KAAiB,EACjB,OAAgC,SAAS,EACzC,UAA6B,EAAE;IAE/B,MAAM,UAAU,GACf,OAAO,CAAC,UAAU,KAAK,SAAS,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,OAAO,CAAC,UAAU,CAAC;QACvE,CAAC,CAAC,CAAC;QACH,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,UAAU,CAAC,CAAC,CAAC;IAChD,MAAM,aAAa,GAAG,OAAO,CAAC,aAAa,IAAI,IAAI,CAAC;IACpD,MAAM,YAAY,GAAG,IAAI,EAAE,MAAM,IAAI,SAAS,CAAC;IAC/C,MAAM,aAAa,GAClB,OAAO,CAAC,SAAS,KAAK,SAAS,IAAI,OAAO,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,OAAO,CAAC,OAAO,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IAC/G,MAAM,gBAAgB,GACrB,OAAO,CAAC,gBAAgB,KAAK,SAAS,IAAI,OAAO,CAAC,gBAAgB,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,gBAAgB,CAAC,CAAC,CAAC,SAAS,CAAC;IAE/G,KAAK,IAAI,OAAO,GAAG,CAAC,GAAI,OAAO,EAAE,EAAE,CAAC;QACnC,YAAY,EAAE,cAAc,EAAE,CAAC;QAC/B,aAAa,EAAE,cAAc,EAAE,CAAC;QAChC,MAAM,oBAAoB,GAAG,gBAAgB,CAAC,CAAC,CAAC,WAAW,CAAC,OAAO,CAAC,gBAAgB,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;QAClG,MAAM,OAAO,GAAG,CAAC,YAAY,EAAE,aAAa,EAAE,oBAAoB,CAAC,CAAC,MAAM,CACzE,CAAC,MAAM,EAAyB,EAAE,CAAC,MAAM,KAAK,SAAS,CACvD,CAAC;QACF,MAAM,MAAM,GAAG,OAAO,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;QAE1E,IAAI,CAAC;YACJ,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC,CAAC,EAAE,GAAG,IAAI,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;YACzE,MAAM,WAAW,GAAG,aAAa,IAAI,sBAAsB,CAAC,GAAG,CAAC,QAAQ,CAAC,MAAM,CAAC,IAAI,OAAO,GAAG,UAAU,CAAC;YACzG,IAAI,CAAC,WAAW;gBAAE,OAAO,QAAQ,CAAC;YAClC,IAAI,CAAC;gBACJ,MAAM,QAAQ,CAAC,IAAI,EAAE,MAAM,EAAE,CAAC;YAC/B,CAAC;YAAC,MAAM,CAAC;gBACR,6EAA6E;gBAC7E,wCAAwC;YACzC,CAAC;QACF,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YAChB,MAAM,eAAe,GACpB,oBAAoB,EAAE,OAAO,KAAK,IAAI,IAAI,CAAC,YAAY,EAAE,OAAO,IAAI,CAAC,aAAa,EAAE,OAAO,CAAC;YAC7F,IACC,YAAY,EAAE,OAAO;gBACrB,aAAa,EAAE,OAAO;gBACtB,CAAC,KAAK,YAAY,KAAK;oBACtB,KAAK,CAAC,IAAI,KAAK,YAAY;oBAC3B,CAAC,eAAe;oBAChB,aAAa,KAAK,SAAS,CAAC;gBAC7B,OAAO,IAAI,UAAU,EACpB,CAAC;gBACF,MAAM,KAAK,CAAC;YACb,CAAC;QACF,CAAC;IACF,CAAC;AACF,CAAC","sourcesContent":["type FetchInput = Parameters<typeof fetch>[0];\n\nconst RETRYABLE_STATUS_CODES = new Set([408, 425, 429, 500, 502, 503, 504]);\n\nexport interface FetchRetryOptions {\n\t/** Number of additional attempts after the initial request. Defaults to two. */\n\tmaxRetries?: number;\n\t/** Retry transient HTTP responses as well as transport failures. Defaults to true. */\n\tretryOnStatus?: boolean;\n\t/** Overall time budget shared by all attempts. */\n\ttimeoutMs?: number;\n\t/** Per-attempt timeout. A new timeout is created for every attempt. */\n\tattemptTimeoutMs?: number;\n}\n\n/**\n * Fetch a management HTTP resource with a bounded immediate retry.\n *\n * This is intentionally a transport-level helper for idempotent management\n * requests (version checks, catalogs, and downloads). It must not be used for\n * agent/model operations: those can fail after the HTTP request starts and are\n * retried by their semantic caller instead.\n *\n * Caller cancellation and timeoutMs are terminal. attemptTimeoutMs aborts\n * only the current attempt so a hung connection can be retried.\n */\nexport async function fetchWithRetry(\n\tinput: FetchInput,\n\tinit: RequestInit | undefined = undefined,\n\toptions: FetchRetryOptions = {},\n): Promise<Response> {\n\tconst maxRetries =\n\t\toptions.maxRetries === undefined || !Number.isFinite(options.maxRetries)\n\t\t\t? 2\n\t\t\t: Math.max(0, Math.floor(options.maxRetries));\n\tconst retryOnStatus = options.retryOnStatus ?? true;\n\tconst parentSignal = init?.signal ?? undefined;\n\tconst timeoutSignal =\n\t\toptions.timeoutMs !== undefined && options.timeoutMs > 0 ? AbortSignal.timeout(options.timeoutMs) : undefined;\n\tconst attemptTimeoutMs =\n\t\toptions.attemptTimeoutMs !== undefined && options.attemptTimeoutMs > 0 ? options.attemptTimeoutMs : undefined;\n\n\tfor (let attempt = 0; ; attempt++) {\n\t\tparentSignal?.throwIfAborted();\n\t\ttimeoutSignal?.throwIfAborted();\n\t\tconst attemptTimeoutSignal = attemptTimeoutMs ? AbortSignal.timeout(attemptTimeoutMs) : undefined;\n\t\tconst signals = [parentSignal, timeoutSignal, attemptTimeoutSignal].filter(\n\t\t\t(signal): signal is AbortSignal => signal !== undefined,\n\t\t);\n\t\tconst signal = signals.length > 1 ? AbortSignal.any(signals) : signals[0];\n\n\t\ttry {\n\t\t\tconst response = await fetch(input, signal ? { ...init, signal } : init);\n\t\t\tconst shouldRetry = retryOnStatus && RETRYABLE_STATUS_CODES.has(response.status) && attempt < maxRetries;\n\t\t\tif (!shouldRetry) return response;\n\t\t\ttry {\n\t\t\t\tawait response.body?.cancel();\n\t\t\t} catch {\n\t\t\t\t// The response is being discarded before a retry. There is nothing useful to\n\t\t\t\t// do if cancelling its body also fails.\n\t\t\t}\n\t\t} catch (error) {\n\t\t\tconst attemptTimedOut =\n\t\t\t\tattemptTimeoutSignal?.aborted === true && !parentSignal?.aborted && !timeoutSignal?.aborted;\n\t\t\tif (\n\t\t\t\tparentSignal?.aborted ||\n\t\t\t\ttimeoutSignal?.aborted ||\n\t\t\t\t(error instanceof Error &&\n\t\t\t\t\terror.name === \"AbortError\" &&\n\t\t\t\t\t!attemptTimedOut &&\n\t\t\t\t\ttimeoutSignal === undefined) ||\n\t\t\t\tattempt >= maxRetries\n\t\t\t) {\n\t\t\t\tthrow error;\n\t\t\t}\n\t\t}\n\t}\n}\n"]}

package/docs/compaction.md

@@ -281,7 +281,9 @@ pi.on("session_before_compact", async (event, ctx) => {
   const { preparation, branchEntries, customInstructions, reason, willRetry, signal } = event;
 
   // preparation.messagesToSummarize - messages to summarize
+  // preparation.sourceMessages - optional active-context history prefix, including any previous summary
   // preparation.turnPrefixMessages - split turn prefix (if isSplitTurn)
+  // preparation.turnPrefixSourceMessages - optional active-context prefix through the split turn
   // preparation.previousSummary - previous compaction summary
   // preparation.fileOps - extracted file operations
   // preparation.tokensBefore - context tokens before compaction
@@ -346,6 +348,21 @@ pi.on("session_before_compact", async (event, ctx) => {
 
 See [custom-compaction.ts](../examples/extensions/custom-compaction.ts) for a complete example using a different model.
 
+### session_compact_failed
+
+Fired when manual or automatic compaction fails or is aborted. This is useful for telemetry extensions that need to pair `session_before_compact` attempts with terminal outcomes.
+
+```typescript
+pi.on("session_compact_failed", async (event, ctx) => {
+  const { reason, errorMessage, aborted, willRetry, fromExtension } = event;
+  // reason - "manual" (/compact), "threshold", or "overflow"
+  // errorMessage - present for non-abort failures
+  // aborted - true for cancelled/aborted compactions
+  // willRetry - whether the aborted turn would have retried after compaction
+  // fromExtension - whether extension-provided compaction content was being used
+});
+```
+
 ### session_before_tree
 
 Fired before `/tree` navigation. Always fires regardless of whether user chose to summarize. Can cancel navigation or provide custom summary.

package/docs/custom-provider.md

@@ -765,8 +765,10 @@ interface ProviderModelConfig {
     requiresThinkingAsText?: boolean;
     requiresReasoningContentOnAssistantMessages?: boolean;
     thinkingFormat?: "openai" | "openrouter" | "deepseek" | "together" | "baseten" | "zai" | "qwen" | "chat-template" | "qwen-chat-template" | "string-thinking" | "ant-ling";
-    chatTemplateKwargs?: Record<string, string | number | boolean | null | { "$var": "thinking.enabled" | "thinking.effort"; omitWhenOff?: boolean }>;
-    chatTemplateArgs?: Record<string, string | number | boolean | null | { "$var": "thinking.enabled" | "thinking.effort"; omitWhenOff?: boolean }>;
+    chatTemplateKwargs?: Record<string, string | number | boolean | null | { "$var": "thinking.enabled" | "thinking.effort" | "thinking.budget"; omitWhenOff?: boolean }>;
+    chatTemplateArgs?: Record<string, string | number | boolean | null | { "$var": "thinking.enabled" | "thinking.effort" | "thinking.budget"; omitWhenOff?: boolean }>;
+    thinkingTokenBudgetField?: "thinking_token_budget" | "thinking_budget" | "thinking_budget_tokens";
+    supportsThinkingTokenBudget?: boolean;
     cacheControlFormat?: "anthropic";
     sessionAffinityFormat?: "openai" | "openai-nosession" | "openrouter";
     sendSessionAffinityHeaders?: boolean;
@@ -784,4 +786,5 @@ interface ProviderModelConfig {
 ```
 
 `openrouter` sends `reasoning: { effort }`. `deepseek` sends `thinking: { type: "enabled" | "disabled" }` and `reasoning_effort` when enabled. `together` sends `reasoning: { enabled }` and also `reasoning_effort` when `supportsReasoningEffort` is enabled. `qwen` is for DashScope-style top-level `enable_thinking`. Use `qwen-chat-template` for local Qwen-compatible servers that read `chat_template_kwargs.enable_thinking` and need `preserve_thinking`. Use `chat-template` for configurable `chat_template_kwargs`, for example DeepSeek V3.x behind vLLM with `chatTemplateKwargs: { "thinking": { "$var": "thinking.enabled" } }`. Use `thinkingFormat: "baseten"` with `chatTemplateArgs` when the provider expects toggle values under `chat_template_args` and optionally supports top-level `reasoning_effort`.
+`thinkingTokenBudgetField` sends a clamped per-level thinking budget as a top-level request field (`thinking_token_budget` on vLLM, `thinking_budget` on Qwen/SGLang, `thinking_budget_tokens` on llama.cpp). `supportsThinkingTokenBudget: true` is an alias for the vLLM field name. Do not combine it with `reasoning_effort` on DashScope Qwen models.
 `cacheControlFormat: "anthropic"` applies Anthropic-style `cache_control` markers to the system prompt, last tool definition, and last user, assistant, or tool-result text content.

package/docs/extensions.md

@@ -375,7 +375,8 @@ user sends another prompt ◄─────────────────
 
 /compact or auto-compaction
   ├─► session_before_compact (can cancel or customize)
-  └─► session_compact
+  ├─► session_compact (success)
+  └─► session_compact_failed (failure or abort)
 
 /tree navigation
   ├─► session_before_tree (can cancel or customize)
@@ -510,7 +511,7 @@ pi.on("session_before_reload", () => {
 
 Use this to protect state a reload would destroy — for example running background children owned by the extension runtime. Keep handlers fast and side-effect free; hosts may consult the veto more than once per reload attempt.
 
-#### session_before_compact / session_compact
+#### session_before_compact / session_compact / session_compact_failed
 
 Fired on compaction. See [compaction.md](compaction.md) for details.
 
@@ -547,6 +548,14 @@ pi.on("session_compact", async (event, ctx) => {
   // event.reason - "manual" (/compact), "threshold", or "overflow"
   // event.willRetry - whether the aborted turn is retried after compaction (overflow recovery)
 });
+
+pi.on("session_compact_failed", async (event, ctx) => {
+  // event.reason - "manual" (/compact), "threshold", or "overflow"
+  // event.errorMessage - present for non-abort failures
+  // event.aborted - true for cancelled/aborted compactions
+  // event.willRetry - whether the aborted turn would have retried after compaction
+  // event.fromExtension - whether extension-provided compaction content was being used
+});
 ```
 
 This accepted/rejected contract has been emitted since [#248](https://github.com/code-yeongyu/senpi/pull/248): accepted events carry `compactionEntry`, while rejected events carry `accepted: false` and `rejectionCause` with no compaction entry. Built-in handlers are required to guard on `event.accepted`; for example, the `claude-sdk-oauth` session registry now does so before recording a compaction fork.

package/docs/llama-cpp.md

@@ -53,6 +53,8 @@ Start Pi and configure the provider:
 
 Enter the router URL and optional API key. The default URL is `http://127.0.0.1:8080`.
 
+If you start the router with `--no-models-autoload`, `/login llama.cpp` only stores the connection. Run `/llama` to load a model, then `/model` to select the loaded model for the current session.
+
 Environment variables can configure the same values without `/login`:
 
 ```bash

package/docs/models.md

@@ -357,6 +357,8 @@ expected.
 
 Only OpenAI-compatible APIs apply it (`openai-completions`, `openai-responses`, `azure-openai-responses`); other APIs ignore it. Keys override senpi's named request fields (for example a `temperature` key here beats the request-level temperature), so prefer it as the single source of sampling truth for a model. In `modelOverrides`, `samplingParams` merges per key with the base model's value.
 
+A constant thinking-token cap can go here too, but it will not follow `thinkingBudgets` or leave room for the answer. Prefer `compat.thinkingTokenBudgetField` (or the `supportsThinkingTokenBudget` alias) for that.
+
 ### Thinking Level Map
 
 Use `thinkingLevelMap` on a model to describe model-specific thinking controls. Keys are senpi thinking levels: `off`, `minimal`, `low`, `medium`, `high`, `xhigh`, `max`. Maps may contain holes; for example, a model can expose `high` and `max` without exposing `xhigh`.
@@ -570,8 +572,10 @@ For providers with partial OpenAI compatibility, use the `compat` field.
 | `requiresThinkingAsText` | Convert thinking blocks to plain text |
 | `requiresReasoningContentOnAssistantMessages` | Include empty `reasoning_content` on all replayed assistant messages when reasoning is enabled |
 | `thinkingFormat` | Use `reasoning_effort`, `openrouter`, `deepseek`, `together`, `baseten`, `zai`, `qwen`, `chat-template`, or `qwen-chat-template` thinking parameters |
-| `chatTemplateKwargs` | `chat_template_kwargs` values for `thinkingFormat: "chat-template"`; use `{ "$var": "thinking.enabled" }` or `{ "$var": "thinking.effort" }` for senpi-controlled thinking values |
-| `chatTemplateArgs` | `chat_template_args` values for `thinkingFormat: "baseten"`; use `{ "$var": "thinking.enabled" }` or `{ "$var": "thinking.effort" }` for senpi-controlled thinking values |
+| `chatTemplateKwargs` | `chat_template_kwargs` values for `thinkingFormat: "chat-template"`; use `{ "$var": "thinking.enabled" }`, `{ "$var": "thinking.effort" }`, or `{ "$var": "thinking.budget" }` for senpi-controlled thinking values |
+| `chatTemplateArgs` | `chat_template_args` values for `thinkingFormat: "baseten"`; use `{ "$var": "thinking.enabled" }`, `{ "$var": "thinking.effort" }`, or `{ "$var": "thinking.budget" }` for senpi-controlled thinking values |
+| `thinkingTokenBudgetField` | Top-level request field used to cap reasoning tokens from `thinkingBudgets`, clamped so at least 1024 tokens remain for the answer. `"thinking_token_budget"` (vLLM), `"thinking_budget"` (Qwen/DashScope/SGLang), `"thinking_budget_tokens"` (llama.cpp). Off by default; not set on the generated catalog. |
+| `supportsThinkingTokenBudget` | Alias for `thinkingTokenBudgetField: "thinking_token_budget"` (vLLM). Prefer `thinkingTokenBudgetField`. Default: `false`. |
 | `cacheControlFormat` | Use Anthropic-style `cache_control` markers on the system prompt, last tool definition, and last user, assistant, or tool-result text content. Currently only `anthropic` is supported. |
 | `sendSessionAffinityHeaders` | For `openai-completions`, send session-affinity headers from the session id when caching is enabled. Default: `false`. |
 | `sessionAffinityFormat` | For `openai-completions` and `openai-responses`, the session-affinity header format: `openai` sends `session_id`/`x-client-request-id` (completions also `x-session-affinity`), `openai-nosession` omits the underscore-containing `session_id` header, `openrouter` sends `x-session-id`. Does not affect the `prompt_cache_key` body param. Default: auto-detected. |
@@ -585,6 +589,8 @@ For providers with partial OpenAI compatibility, use the `compat` field.
 
 `openrouter` uses `reasoning: { effort }`. `together` uses `reasoning: { enabled }` and also `reasoning_effort` when `supportsReasoningEffort` is enabled. `qwen` uses top-level `enable_thinking`. Use `qwen-chat-template` for local Qwen-compatible servers that require `chat_template_kwargs.enable_thinking` and `preserve_thinking`. Use `chat-template` for vLLM/Hugging Face chat templates that need configurable `chat_template_kwargs`, such as `chatTemplateKwargs: { "thinking": { "$var": "thinking.enabled" } }` for DeepSeek V3.x templates. Use `thinkingFormat: "baseten"` with `chatTemplateArgs` for providers that expose toggle controls through `chat_template_args` and optionally support top-level `reasoning_effort`.
 
+`thinkingTokenBudgetField` is independent of `thinkingFormat`. Do not enable it on the generated Qwen catalog: those models already send `reasoning_effort`, and DashScope rejects `thinking_budget` together with `reasoning_effort`.
+
 `cacheControlFormat: "anthropic"` is for OpenAI-compatible providers that expose Anthropic-style prompt caching through `cache_control` markers on text content and tool definitions.
 
 Example: