feat: per-model max output tokens and a conversation-time thinking level backed by agent settings (#28)

Co-authored-by: Alice <alice@prismshadow.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
2026-07-22 22:22:41 +08:00
parent 99f391f379
commit cabbab1c16
36 changed files with 797 additions and 42 deletions
@@ -0,0 +1,161 @@
+/**
+ * Session title generation: an **out-of-band, one-off request** that generates
+ * a short title from the first-turn conversation text (used by `Session.generateTitle`
+ * for assembly, not exported wholesale via the barrel).
+ *
+ * Called by `session.generateTitle()`: sends one request using the bare LLM for the session's
+ * Model (no tools, no system prompt, thinking off), without writing history or Trace. Material
+ * defaults to what the Session self-captures during run (see session.ts); this module is only
+ * responsible for the prompt format, driving the one-off request, and sanitizing the result —
+ * when to generate a title and where to store it is decided by the host (Web server / CLI).
+ * The narrow public surface — `SessionTitleResult` (part of `Session.generateTitle`'s
+ * signature) and the sanitation helpers the host's title fallback builds on — is re-exported
+ * by the barrel; the prompt/request internals are not.
+ */
+import { userText } from "../omnimessage/index.js";
+import type {
+  OmniMessage,
+  TextPayload,
+  TokenCounts,
+  TokenUsagePayload,
+} from "../omnimessage/index.js";
+import type { LLMInterface } from "../interfaces.js";
+
+/** Cap on conversation text spliced into the title request (user/model each truncated separately, to control cost). */
+const EXCERPT_MAX_CHARS = 2000;
+/** Cap on title length (fallback truncation for when the model occasionally ignores the constraint). */
+const TITLE_MAX_CHARS = 30;
+
+/**
+ * Special message markers that must never leak into a title. These are machine-inserted
+ * XML-ish blocks (a skill invocation wraps the body in `<use_skills>…</use_skills>`, a
+ * subagent handoff / scheduled task prepend their own blocks) — meaningful to the runtime,
+ * noise in a title. The list is a fixed allowlist so ordinary angle-bracket text (e.g. a
+ * user pasting `<div>`) is left untouched.
+ */
+const MARKER_TAGS = ["use_skills", "handoff_from", "scheduled_task"];
+
+/**
+ * Strips machine-inserted marker blocks (see MARKER_TAGS) from conversation text so titles are
+ * built from the human-meaningful body only — both the material sent to the model and the
+ * fallback derived from the raw first message. Removes the paired `<tag>…</tag>` block (any
+ * inner content, across lines) and any stray unpaired `<tag>` / `</tag>` left behind.
+ */
+export function stripConversationMarkers(text: string): string {
+  let out = text;
+  for (const tag of MARKER_TAGS) {
+    out = out
+      .replace(new RegExp(`<${tag}>[\\s\\S]*?</${tag}>`, "g"), "")
+      .replace(new RegExp(`</?${tag}>`, "g"), "");
+  }
+  return out.trim();
+}
+
+export interface SessionTitleResult {
+  /** The sanitized title; null when material is insufficient, the request fails, or the output is empty. */
+  title: string | null;
+  /** Token consumption for this request (accumulated token_usage.request); null if no request occurred or there's no usage. */
+  usage: TokenCounts | null;
+}
+
+/**
+ * Assembles the title-generation Prompt (exported for host/test assertion use). Uses English
+ * instructions to avoid polluting the title's language, and requires the **title to be in the
+ * same language as the conversation** (English conversation gets an English title, Chinese
+ * conversation gets a Chinese title); when assistant material is empty, it relies on the user
+ * request alone.
+ */
+export function buildTitlePrompt(userExcerpt: string, assistantExcerpt: string): string {
+  const clip = (s: string) => (s.length > EXCERPT_MAX_CHARS ? s.slice(0, EXCERPT_MAX_CHARS) : s);
+  const lines = [
+    "Generate a concise title for the conversation below.",
+    "Rules:",
+    "- Write the title in the SAME language the user is using.",
+    "- Keep it short: at most 6 words, or ~16 characters for CJK.",
+    "- Output ONLY the title text — no quotes, no trailing punctuation, no explanation.",
+    "- Answer immediately — do not think aloud or produce chain-of-thought.",
+    "",
+    "[User]",
+    clip(userExcerpt),
+  ];
+  if (assistantExcerpt.trim()) {
+    lines.push("", "[Assistant]", clip(assistantExcerpt));
+  }
+  // The trailing empty think block makes many reasoning models treat their thinking phase
+  // as already closed, so the one-off request spends its budget on the title itself.
+  lines.push("", "<think></think>");
+  return lines.join("\n");
+}
+
+/** Sanitizes model output into a title: strips any leaked marker blocks and leading/trailing quotes/brackets and trailing punctuation (until stable), collapses whitespace, and truncates if too long; returns null for an empty result. */
+export function sanitizeTitle(raw: string): string | null {
+  let t = stripConversationMarkers(raw).replace(/\s+/g, " ").trim();
+  // Stripping quotes can expose more punctuation underneath (or vice versa), so strip repeatedly until stable.
+  for (let prev = ""; prev !== t;) {
+    prev = t;
+    t = t
+      .replace(/^["'“”‘’「」『』《》〈〉【】()（）\s]+/, "")
+      .replace(/["'“”‘’「」『』《》〈〉【】()（）\s]+$/, "")
+      .replace(/[。.．!！?？;；,，、:：]+$/, "")
+      .trim();
+  }
+  if (!t) return null;
+  return t.length > TITLE_MAX_CHARS ? t.slice(0, TITLE_MAX_CHARS) : t;
+}
+
+/**
+ * Drives a single title-generation request: collects model text and token_usage, and resolves
+ * based on the outcome. Generation only requires user material (assistant material may be
+ * empty — a pure tool-only turn can still get a title); no request is sent if user material is
+ * empty; `title` is null if the request doesn't complete (any usage already produced is still
+ * returned).
+ */
+export async function generateTitleWithLLM(
+  llm: LLMInterface,
+  args: { userText: string; assistantText: string; signal?: AbortSignal },
+): Promise<SessionTitleResult> {
+  // Strip machine markers before the model ever sees the material, so a skill-invocation
+  // block (or handoff / scheduled-task marker) can't bleed into the generated title.
+  const userMaterial = stripConversationMarkers(args.userText);
+  const assistantMaterial = stripConversationMarkers(args.assistantText);
+  if (!userMaterial.trim()) {
+    return { title: null, usage: null };
+  }
+  const prompt = buildTitlePrompt(userMaterial, assistantMaterial);
+  const gen = llm.streamGenerate({
+    newMessages: [userText(prompt)],
+    ...(args.signal ? { signal: args.signal } : {}),
+  });
+  let collected = "";
+  let usage: TokenCounts | null = null;
+  for (;;) {
+    const step = await gen.next();
+    if (step.done) {
+      if (step.value.status !== "completed") return { title: null, usage };
+      break;
+    }
+    const msg = step.value;
+    if (isAssistantText(msg)) collected += msg.payload.text;
+    if (isTokenUsage(msg)) {
+      const r = msg.payload.request;
+      usage = usage
+        ? {
+            cache_read: usage.cache_read + r.cache_read,
+            cache_write: usage.cache_write + r.cache_write,
+            output: usage.output + r.output,
+            total: usage.total + r.total,
+          }
+        : { ...r };
+    }
+  }
+  return { title: sanitizeTitle(collected), usage };
+}
+
+function isAssistantText(msg: OmniMessage): msg is OmniMessage<TextPayload> {
+  const payload = msg.payload as { type?: string; role?: string };
+  return msg.type === "model_msg" && payload.type === "text" && payload.role === "assistant";
+}
+
+function isTokenUsage(msg: OmniMessage): msg is OmniMessage<TokenUsagePayload> {
+  return msg.type === "event_msg" && (msg.payload as { type?: string }).type === "token_usage";
+}