@tiagojct/subsub 0.9.1 → 0.11.0

package/README.md

@@ -17,9 +17,7 @@ A Zotero librarian and research assistant that runs on your computer, in your br
 
 * **Code-Enforced Safety Gate**: The model **cannot** modify your library without your explicit approval. Every change is calculated first as a dry-run preview (+added in green, -removed in red) and executed only when you confirm.
 * **Full Journal & Reversible Edits**: Every library modification is recorded with a unique journal ID and can be reverted anytime using `/undo`.
-* **Two Dedicated Roles**:
-  * 📚 **The Librarian**: Cleans tags, imports references (DOI, PMID, ISBN), deduplicates, fixes metadata, retrieves open-access PDFs, creates collections, and checks retractions. *Has no web search tools; imports and PDFs come from Crossref, Unpaywall and Open Library.*
-  * 🔬 **The Researcher**: Reads items, searches PubMed and OpenAlex, finds gaps, checks citations in manuscripts, and writes structured Markdown notes in your Sub-Sub folder (it can sit inside an Obsidian vault). *Cannot modify library records.*
+* **Two modes**: The **Researcher** (the default) reads items, searches PubMed, Europe PMC and OpenAlex, finds gaps, checks citations in manuscripts and writes Markdown notes in your Sub-Sub folder (it can sit inside an Obsidian vault). It also does everything the Librarian does, so you can research a topic and import and tag what you find in the same conversation. The **Librarian** manages the library only: it tags, imports by DOI, PMID or ISBN, repairs metadata, finds duplicates, attaches open-access PDFs and checks retractions. It has no search tools; use it with a fast, cheap model for tagging sessions. Each mode can have its own model. In both modes, every library change needs your approval.
 * **Local & Private**: Listens strictly on `127.0.0.1` protected by an authenticated single-origin session cookie. Your library contents remain on your computer.
 * **Modern Web Interface & Terminal UI**: Use Sub-Sub in your browser with desktop shortcuts (`~/Applications/Sub-Sub.app`, Start menu, or `.desktop`), or launch it right in your terminal with adaptive color themes.
 
@@ -32,7 +30,7 @@ Profiles calibrate how much Sub-Sub drafts and executes for you. Choose your pro
 | Profile | Description |
 |---|---|
 | **Reader** | Guided and instructional. Reading notes are quotes with page numbers and reflective questions. Never generates unrequested summaries or syntheses. No bulk library edits. |
-| **Scholar** | Full research workflow: literature notes, conceptual syntheses, database searches, alerts, and the full librarian toolkit. |
+| **Scholar** | Full research workflow: literature notes, conceptual syntheses, database searches, alerts, and all library changes. |
 | **Author** | Scholar capabilities plus manuscript workflows: citation auditing against your library, automated Pandoc/Quarto bibliographies, and argument analysis. It critiques but does not write your paragraphs. |
 | **Editor** | Everything unlocked, including drafting full manuscript sections and abstracts upon request. |
 
@@ -106,7 +104,8 @@ Starting Sub-Sub with `subsub web` (or `subsub serve`) launches the browser appl
 
 * **Approval Previews**: Shows server-computed dry-run diffs directly in the UI. Changes are applied only after clicking **Yes** or pressing <kbd>Y</kbd>.
 * **Keyboard-First Review**: In approval dialogs, press <kbd>Y</kbd> or <kbd>Cmd</kbd>/<kbd>Ctrl</kbd>+<kbd>Enter</kbd> to approve, or <kbd>N</kbd>/<kbd>Esc</kbd> to decline.
-* **Theme Switcher**: Header toggle between **Light**, **Dark**, and **System** themes, with tailored styling for each mode (Try-Works amber for Librarian, Glauca blue for Researcher).
+* **Mode buttons**: Researcher and Librarian, in the header. The side panel shows the actions of the mode: Research and Library in the Researcher, Library only in the Librarian. The model button saves the model for the current mode, or for both modes.
+* **Theme Switcher**: Header toggle between **Light**, **Dark**, and **System** themes.
 * **One-Click Copy**: Built-in copy buttons on all code snippets, BibTeX records, and assistant literature notes.
 * **Session Search & Markdown Export**: Filter conversation history in real time and download full discussions as Markdown notes.
 * **Local Security**: Binds strictly to `127.0.0.1`. The initial link carries a random token that converts into an `HttpOnly`, `SameSite=Strict` cookie. Protected with strict Content Security Policy (CSP), origin validation, and sanitized output.
@@ -116,11 +115,11 @@ Starting Sub-Sub with `subsub web` (or `subsub serve`) launches the browser appl
 
 ## Usage
 
-### Interactive Modes & Commands
+### Starting and commands
 
 ```sh
-subsub             # Start in researcher mode
-subsub --librarian # Start in librarian mode
+subsub             # Start in the terminal (Researcher)
+subsub --librarian # Start in the Librarian
 subsub web         # Start browser view
 subsub -c          # Continue the previous session
 subsub -r          # Resume an older session
@@ -128,8 +127,8 @@ subsub -r          # Resume an older session
 
 Commands available within Sub-Sub:
 
-* `/librarian`: Switch to the librarian (modify library, tag, import; offline).
-* `/researcher`: Switch to the researcher (read library, search literature, write notes).
+* `/researcher`: Change to the Researcher (research and library). Each mode can have its own model.
+* `/librarian`: Change to the Librarian (library only).
 * `/profile [name]`: View or change active profile (`reader`, `scholar`, `author`, `editor`).
 * `/subsub`: Display Zotero connection health and library statistics.
 * `/history`: List recent library changes with journal IDs.
@@ -139,11 +138,11 @@ Commands available within Sub-Sub:
 
 Type the command followed by arguments:
 
-* **Librarian**:
+* **Library (both modes)**:
   * `/tag-batch [size]`: Tag untagged items using your tag vocabulary.
   * `/clean-tags`: Find irregular tags and reconcile with your vocabulary.
   * `/import-queue [path]`: Import pending references from an inbox note.
-* **Researcher**:
+* **Research (Researcher)**:
   * `/lit <question>`: Literature review: the library first, then PubMed, Europe PMC and OpenAlex in one merged search; reads before it summarises; an evidence table, agreement, disagreement and open questions; references checked; one note in `Research/` (Scholar, Author, Editor).
   * `/compare <question or citekeys>`: Source matrix (claim, evidence type, caveats, confidence), agreements and disagreements (Scholar, Author, Editor).
   * `/review <path>`: Critical review of your own manuscript: FATAL, MAJOR and MINOR issues against the reporting guideline that applies (CONSORT, STROBE, PRISMA, STARD, TRIPOD), with quotations and a revision plan. Comments only (Author, Editor).
@@ -177,12 +176,12 @@ subsub review apply 13-31
 | `about` | string | Brief background context about your research domain | `null` |
 | `language` | string | Response language, or `"auto"` to match your input | `"English"` |
 | `vault` | string | Path to your Sub-Sub folder | `$ZOTERO_VAULT` |
-| `models` | object | Model overrides per mode (e.g. `{"librarian": "opencode-go/glm-5.3-flash"}`) | Current model |
-| `defaultMode`| string | Default launch mode (`"researcher"` or `"librarian"`) | `"researcher"` |
+| `models` | object | The model for each mode, as provider/id: `{"librarian": ..., "researcher": ...}`; `{}` to choose them in Sub-Sub. The 0.10 key `model` (one for both modes) still works; `libraryChanges` is ignored. | `{"librarian": "opencode-go/mimo-v2.6-flash", "researcher": "opencode-go/mimo-v2.6-pro"}` |
+| `defaultMode` | string | `"researcher"` or `"librarian"`: the mode Sub-Sub starts in | `"researcher"` |
 | `serverDir` | string | Local development path for `zotero-local-mcp` | PyPI package |
 | `look` | boolean| Show header banners and status lines in terminal | `true` |
 | `quotes` | boolean| Display quotes from *Moby-Dick* in terminal banner | `true` |
-| `themes` | object | Mode-specific terminal themes | `true` |
+| `themes` | string or false | Terminal theme family (`subsub-glauca` or `subsub-try-works`); `false` keeps your pi theme | `"subsub-glauca"` |
 | `addons` | array | Optional servers; `["starbuck"]` adds reference checks | `[]` |
 | `starbuckDir` | string | Local Starbuck folder (used when it has a `pyproject.toml`) | `~/Projects/starbuck` |
 
@@ -203,7 +202,7 @@ Sub-Sub finds open-access copies (Unpaywall, Europe PMC). For other papers, use
 2. Optional: set `"starbuckDir"` to a local copy of Starbuck. Without it, Sub-Sub runs the pinned Starbuck release with uv.
 3. Start Sub-Sub again and run `subsub doctor`. The line `Starbuck` must show `ok`.
 
-The researcher gets four tools: `verify_check_manuscript`, `verify_check_references`, `verify_prepare_claims` and `verify_record_claims`. Starbuck writes its report to a `_starbuck` folder next to the manuscript. Inside the Sub-Sub folder or the working folder this needs no approval; elsewhere Sub-Sub asks first. Starbuck gets the contact email from the zotero-local-mcp env file. For the claim check, Sub-Sub's own model judges; Starbuck checks that every quoted passage is in the source.
+The Researcher gets four tools: `verify_check_manuscript`, `verify_check_references`, `verify_prepare_claims` and `verify_record_claims`. Starbuck writes its report to a `_starbuck` folder next to the manuscript. Inside the Sub-Sub folder or the working folder this needs no approval; elsewhere Sub-Sub asks first. Starbuck gets the contact email from the zotero-local-mcp env file. For the claim check, Sub-Sub's own model judges; Starbuck checks that every quoted passage is in the source.
 
 ---
 

package/dist/bench.js

@@ -3,7 +3,8 @@
  * with no changes applied. See docs/Subsub.md ("Model test").
  *
  *   subsub-bench prepare [--seed 7]        pick the tasks (read-only)
- *   subsub-bench run [--librarian a,b] [--researcher c,d] [--parallel 3] [--only task]
+ *   subsub-bench run [--models a,b] [--librarian a,b] [--researcher c,d] [--parallel 3] [--only task]
+ *   (--models sets the models for both task groups.)
  *   subsub-bench score                     objective checks -> results.md
  *
  * Results go to ~/Projects/subsub/bench-results/<date>/ (or --dir). Each model gets
@@ -19,8 +20,8 @@ import { expand, loadConfig, starbuckCommand, starbuckEnabled } from "./config.j
 import { findUv } from "./subsub.js";
 export const PACKAGE_DIR = resolve(dirname(fileURLToPath(import.meta.url)), "..");
 export const DEFAULT_MODELS = {
-    librarian: ["glm-5.3-flash", "deepseek-v4.1-flash", "qwen3.8-flash", "mimo-v2.6-flash", "minimax-m3", "mimo-v2.6-pro"],
-    researcher: ["mimo-v2.6-pro", "deepseek-v4-pro", "kimi-k2.6", "glm-5.3", "qwen3.7-plus", "minimax-m3"],
+    librarian: ["mimo-v2.6-flash", "glm-5.3-flash", "qwen3.8-flash", "gpt-5.6-luna", "hy3", "mimo-v2.6-pro"],
+    researcher: ["mimo-v2.6-pro", "qwen3.8-flash", "kimi-k3", "glm-5.3", "minimax-m3", "gpt-5.6-luna"],
 };
 export const TASKS = {
     librarian: ["tagging", "facet_fix", "import"],
@@ -530,9 +531,10 @@ export async function main(argv = process.argv.slice(2)) {
         if (!cfg.vault)
             throw new Error("No vault configured (ZOTERO_VAULT).");
         const t = JSON.parse(readFileSync(tasksFile, "utf8"));
+        const both = arg(argv, "models")?.split(",").filter(Boolean);
         const lists = {
-            librarian: arg(argv, "librarian")?.split(",").filter(Boolean) ?? DEFAULT_MODELS.librarian,
-            researcher: arg(argv, "researcher")?.split(",").filter(Boolean) ?? DEFAULT_MODELS.researcher,
+            librarian: arg(argv, "librarian")?.split(",").filter(Boolean) ?? both ?? DEFAULT_MODELS.librarian,
+            researcher: arg(argv, "researcher")?.split(",").filter(Boolean) ?? both ?? DEFAULT_MODELS.researcher,
         };
         const only = arg(argv, "only")?.split(",");
         const codes = codesFor([...new Set([...lists.librarian, ...lists.researcher])], dir);
@@ -556,7 +558,7 @@ export async function main(argv = process.argv.slice(2)) {
             if (!prompt)
                 return;
             const cf = join(cfgDir, `${j.code}.json`);
-            writeFileSync(cf, JSON.stringify({ ...userCfg, profile: "editor", models: { librarian: `opencode-go/${j.model}`, researcher: `opencode-go/${j.model}` } }));
+            writeFileSync(cf, JSON.stringify({ ...userCfg, profile: "editor", models: { librarian: `opencode-go/${j.model}`, researcher: `opencode-go/${j.model}` }, model: undefined, libraryChanges: undefined }));
             const s = await runOne({ ...j, prompt, out, cwd: cfg.vault, configFile: cf, timeoutSec });
             n++;
             console.log(`${n}/${jobs.length}  ${j.code}  ${j.task}  ${s.seconds}s  ${s.error ?? (s.settled ? "done" : "?")}`);
@@ -585,5 +587,5 @@ export async function main(argv = process.argv.slice(2)) {
         console.log(`Wrote ${join(dir, "results.md")}`);
         return;
     }
-    console.log("Usage: subsub-bench prepare | run [--librarian a,b] [--researcher c,d] [--only task] [--parallel 3] | score  [--dir DIR]");
+    console.log("Usage: subsub-bench prepare | run [--models a,b] [--librarian a,b] [--researcher c,d] [--only task] [--parallel 3] | score  [--dir DIR]");
 }

package/dist/bridge.js

@@ -1,6 +1,6 @@
 /**
- * Bridge to the Python MCP servers of zotero-local-mcp (librarian "zotero" and
- * researcher "scholar"). Starts them over stdio, lists their tools, and calls them.
+ * Bridge to the Python MCP servers of zotero-local-mcp ("zotero" for the library,
+ * "scholar" for searches and notes). Starts them over stdio, lists their tools, and calls them.
  * All safety logic (dry runs, version checks, journal, vocabulary, the researcher's
  * note-only client) stays in the Python servers.
  */

package/dist/config.js

@@ -6,10 +6,14 @@
  *   "profile": "scholar",
  *   "userName": "Ana", "about": "a master's student in health informatics", "language": "English",
  *   "vault": "~/Documents/Sub-Sub",
- *   "models": { "librarian": "opencode-go/glm-5.3-flash", "researcher": "opencode-go/mimo-v2.6-pro" },
+ *   "models": { "librarian": "opencode-go/mimo-v2.6-flash", "researcher": "opencode-go/mimo-v2.6-pro" },
  *   "defaultMode": "researcher"
  * }
  *
+ * Two modes: the Librarian manages the library; the Researcher does that too,
+ * and searches and writes notes. "model" (0.10) sets one model for both;
+ * "libraryChanges" (0.10) is no longer used.
+ *
  * `subsub init` writes this file. Without "serverDir" (or when that folder does
  * not exist), the Zotero servers run from PyPI with uv.
  *
@@ -22,6 +26,11 @@ import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
 import { homedir } from "node:os";
 import { dirname, join, resolve } from "node:path";
 import { isObsidianVault, settingsFile } from "./layout.js";
+/** The tested models, one per mode (model test, see bench-results). */
+export const DEFAULT_MODELS = {
+    librarian: "opencode-go/mimo-v2.6-flash",
+    researcher: "opencode-go/mimo-v2.6-pro",
+};
 export const PROFILES = ["reader", "scholar", "author", "editor"];
 /** The Zotero server release that this Sub-Sub version runs when there is no local server folder. */
 export const SERVER_VERSION = "0.5.1";
@@ -101,7 +110,14 @@ export function loadConfig(env = process.env) {
     const envValues = envFile && existsSync(envFile) ? parseEnvFile(envFile) : {};
     const vaultRaw = user.vault ?? env.ZOTERO_VAULT ?? envValues.ZOTERO_VAULT;
     const vault = vaultRaw ? expand(vaultRaw, home) : undefined;
-    const mode = user.defaultMode === "librarian" ? "librarian" : "researcher";
+    // "models" per mode; "model" (0.10) for both; {} or "" means: the user chooses in pi (/login, /model).
+    const one = typeof user.model === "string" ? user.model : undefined;
+    const given = user.models && typeof user.models === "object" ? user.models : undefined;
+    const models = given ?? (one !== undefined ? { librarian: one || undefined, researcher: one || undefined } : { ...DEFAULT_MODELS });
+    for (const k of Object.keys(models))
+        if (!models[k])
+            delete models[k];
+    const themes = user.themes === false ? false : typeof user.themes === "string" ? user.themes : user.themes?.researcher;
     return {
         configFile: file,
         profile: isProfile(user.profile) ? user.profile : "scholar",
@@ -112,11 +128,8 @@ export function loadConfig(env = process.env) {
         serverDir: expand(user.serverDir ?? "~/Projects/zotero-local-mcp", home),
         envFile,
         vault,
-        models: user.models ?? {
-            librarian: "opencode-go/glm-5.3-flash",
-            researcher: "opencode-go/mimo-v2.6-pro",
-        },
-        defaultMode: mode,
+        models,
+        defaultMode: user.defaultMode === "librarian" ? "librarian" : "researcher",
         // A Sub-Sub folder inside an Obsidian vault also reads the vault's context file.
         vaultContext: (user.vaultContext ??
             (vault ? [join(vault, ".claude", "CLAUDE.md"), ...(isObsidianVault(dirname(vault)) ? [join(dirname(vault), ".claude", "CLAUDE.md")] : [])] : [])).map((p) => expand(p, home)),
@@ -129,7 +142,7 @@ export function loadConfig(env = process.env) {
         uvPath: user.uvPath ? expand(user.uvPath, home) : undefined,
         look: user.look,
         quotes: user.quotes,
-        themes: user.themes,
+        themes,
         addons: Array.isArray(user.addons) ? user.addons.filter((a) => typeof a === "string") : [],
         starbuckDir: expand(user.starbuckDir ?? "~/Projects/starbuck", home),
     };

package/dist/doctor.js

@@ -92,7 +92,7 @@ export function runChecks(cfg, env, deps = {}) {
             fix: v.problems?.length
                 ? `Fix these lines in the tag list: ${v.problems.slice(0, 3).join("; ")}`
                 : noTopics
-                    ? "In Sub-Sub (librarian), ask: propose topics for my tag list."
+                    ? "In Sub-Sub, ask: propose topics for my tag list."
                     : undefined,
         });
     }
@@ -104,7 +104,7 @@ export function runChecks(cfg, env, deps = {}) {
         out.push(starbuckCheck(cfg, env, run));
     const agentDir = deps.agentDir ?? (env.SUBSUB_AGENT_DIR ? env.SUBSUB_AGENT_DIR : join(homedir(), ".subsub", "agent"));
     out.push(hasLogin(agentDir)
-        ? { name: "Model login", ok: true, detail: Object.keys(cfg.models).length ? Object.values(cfg.models).join(", ") : "pi's default model" }
+        ? { name: "Model login", ok: true, detail: Object.keys(cfg.models).length ? Object.entries(cfg.models).map(([m, id]) => `${m} ${id}`).join(", ") : "pi's default model" }
         : { name: "Model login", ok: false, detail: "no model provider yet", fix: "Type subsub, then /login." });
     return out;
 }

package/dist/init.js

@@ -4,7 +4,7 @@
  *
  * Writes:
  * - the settings file (~/.config/subsub/config.json): profile, name, language,
- *   notes folder, models;
+ *   Sub-Sub folder, models;
  * - the server settings file (~/.config/zotero-local-mcp/env, or the one that
  *   already exists): the Sub-Sub folder, tag list, alerts file, contact email.
  *   Other lines stay;
@@ -26,7 +26,7 @@ import { homedir } from "node:os";
 import { dirname, join, resolve } from "node:path";
 import { createInterface } from "node:readline/promises";
 import { fileURLToPath } from "node:url";
-import { configPath, expand, isProfile, loadConfig, parseEnvFile, PROFILES, readConfigFile, saveConfig, starbuckEnabled, withStarbuck, } from "./config.js";
+import { configPath, DEFAULT_MODELS, expand, isProfile, loadConfig, parseEnvFile, PROFILES, readConfigFile, saveConfig, starbuckEnabled, withStarbuck, } from "./config.js";
 import { defaultRunner, starbuckCheck } from "./doctor.js";
 import { applyMoves, FOLDER_NAME, leftBehind, NOTE_DIRS, planMoves, SETTINGS_DIR, SETTINGS_FILES, subsubFolder } from "./layout.js";
 import { PROFILE_SPECS } from "./profiles.js";
@@ -34,9 +34,9 @@ const TEMPLATES = resolve(dirname(fileURLToPath(import.meta.url)), "..", "templa
 export const STARTER_TAGS = {
     "health-sciences": "Health sciences: clinical areas, public health, methods and study designs",
     "health-informatics": "Health informatics: information systems, data, AI, digital health",
-    "any-field": "Any field: methods and document types only; the librarian proposes topics from your library",
+    "any-field": "Any field: methods and document types only; Sub-Sub proposes topics from your library",
 };
-export const TESTED_MODELS = { librarian: "opencode-go/glm-5.3-flash", researcher: "opencode-go/mimo-v2.6-pro" };
+export const TESTED_MODELS = DEFAULT_MODELS;
 /** Set KEY=VALUE lines in an env file text; other lines and comments stay. */
 export function upsertEnv(text, values) {
     const lines = text ? text.replace(/\n$/, "").split("\n") : [];
@@ -140,24 +140,26 @@ export async function runInit(io, flags, env = process.env, deps = {}) {
         io.say(`Tag list: keeping ${vocab}.`);
     }
     else {
-        explain("Tag list", `The librarian tags items only with the tags in this list. Pick a start; you edit the list later in ${SETTINGS_DIR}/${SETTINGS_FILES.tags}.`);
+        explain("Tag list", `Sub-Sub tags items only with the tags in this list. Pick a start; you edit the list later in ${SETTINGS_DIR}/${SETTINGS_FILES.tags}.`);
         tags = await io.choose("Starter tag list", Object.entries(STARTER_TAGS).map(([value, label]) => ({ value, label })), flags.tags ?? (fmup ? "health-sciences" : "any-field"));
     }
     explain("Contact email", "Unpaywall needs an email address to find open-access PDFs. Sub-Sub sends it only to Unpaywall, Crossref, OpenAlex and PubMed, as they ask.");
     const email = (await io.ask("Email (optional)", flags.email ?? envValues.ZOTERO_CONTACT_EMAIL ?? "")).trim();
-    explain("Reference checks (Starbuck)", "An optional add-on for manuscripts: it checks that each cited work exists, matches its citation and was not retracted. The researcher then has /verify.");
+    explain("Reference checks (Starbuck)", "An optional add-on for manuscripts: it checks that each cited work exists, matches its citation and was not retracted. Sub-Sub then has /verify.");
     const starbuck = await io.choose("Reference checks", [
         { value: "off", label: "Off" },
         { value: "on", label: "On (downloads Starbuck now; uses the email above)" },
     ], flags.starbuck ?? (starbuckEnabled(current) ? "on" : "off"));
-    explain("Models", "Sub-Sub needs an account with a model provider. You can connect one later in Sub-Sub with the model button (or /login in the terminal).");
+    explain("Models", "Sub-Sub needs an account with a model provider. Each mode can use its own model: the Librarian a fast, cheap one for tags and imports, the Researcher a stronger one for reading. You can connect a provider later in Sub-Sub with the model button (or /login in the terminal).");
+    const hadModel = "model" in existing || "models" in existing;
+    const show = (m) => (Object.keys(m).length ? Object.entries(m).map(([k, v]) => `${k} ${v}`).join(", ") : "choose in Sub-Sub");
     const modelOptions = [
-        ...(existing.models ? [{ value: "keep", label: `Keep: ${JSON.stringify(existing.models)}` }] : []),
+        ...(hadModel ? [{ value: "keep", label: `Keep: ${show(current.models)}` }] : []),
         { value: "later", label: "Choose later in Sub-Sub" },
-        { value: "opencode-go", label: `OpenCode Go, tested: ${TESTED_MODELS.librarian} for the librarian, ${TESTED_MODELS.researcher} for the researcher` },
+        { value: "opencode-go", label: `OpenCode Go, tested: ${TESTED_MODELS.librarian} for the Librarian, ${TESTED_MODELS.researcher} for the Researcher` },
     ];
     const agentDir = initEnv.SUBSUB_AGENT_DIR ? expand(initEnv.SUBSUB_AGENT_DIR, home) : join(home, ".subsub", "agent");
-    const modelDefault = existing.models ? "keep" : hasProvider(agentDir, "opencode") ? "opencode-go" : "later";
+    const modelDefault = hadModel ? "keep" : hasProvider(agentDir, "opencode") ? "opencode-go" : "later";
     const models = await io.choose("Models", modelOptions, flags.models ?? modelDefault);
     // ---- write
     if (move === "yes")
@@ -193,8 +195,13 @@ export async function runInit(io, flags, env = process.env, deps = {}) {
         setup: fmup ? "fmup" : undefined,
         vault: notes,
         envFile,
+        libraryChanges: undefined,
         addons: withStarbuck(current.addons, starbuck === "on"),
     };
+    // A model per mode; "model" (0.10, one for both) is replaced.
+    patch.model = undefined;
+    if (models === "keep")
+        patch.models = current.models;
     if (models === "later")
         patch.models = {};
     if (models === "opencode-go")

package/dist/look.js

@@ -20,7 +20,8 @@ export const QUOTES = [
     ["God keep me from ever completing anything.", "Moby-Dick, ch. 32"],
     ["Call me Ishmael.", "Moby-Dick, ch. 1"],
 ];
-export const MODE_THEMES = { librarian: "subsub-try-works", researcher: "subsub-glauca" };
+/** The terminal theme, without -dark/-light. subsub-try-works is the other one shipped. */
+export const DEFAULT_THEME = "subsub-glauca";
 /** Light or dark terminal, judged from the text colour of the current theme (dark text means a light background). */
 export function schemeFromAnsi(fgAnsi, themeName) {
     if (themeName?.endsWith("-light") || themeName === "light")
@@ -41,8 +42,7 @@ export function schemeFromAnsi(fgAnsi, themeName) {
     }
     return "dark";
 }
-export function themeName(mode, scheme, map = MODE_THEMES) {
-    const base = map[mode];
+export function themeName(scheme, base = DEFAULT_THEME) {
     return base ? `${base}-${scheme}` : undefined;
 }
 const plain = { mark: (s) => s, text: (s) => s, muted: (s) => s, dim: (s) => s, warn: (s) => s };

package/dist/profiles.js

@@ -5,7 +5,7 @@
  *
  * - reader:  explains each step; reading notes are quotes with page numbers and
  *            questions; no summaries or syntheses; no bulk library changes.
- * - scholar: literature notes, syntheses, searches, alerts; the full librarian.
+ * - scholar: literature notes, syntheses, searches, alerts; all library changes.
  * - author:  scholar plus manuscript work: citation check, bibliography, comments
  *            on the argument. Does not write the manuscript's paragraphs.
  * - editor:  everything, including drafting text for a manuscript on request.
@@ -26,13 +26,13 @@ export const PROFILE_SPECS = {
             "- A literature note is a reading note: the full reference, then direct quotes with page numbers (or section names), and under each quote one question for {{user}} to answer. Do not write summaries, syntheses, conclusions or paragraphs that {{user}} could hand in as their own text. When a prompt or the shared rules ask for a literature note or a synthesis, make a reading note instead and say so.",
             "- For searches, list works with one line each on what they study, not on what they conclude.",
             "- If {{user}} asks for a summary, a synthesis or text for an assignment, say that the Reader profile does not write these, and offer quotes and questions. {{User}} can change the profile with /profile.",
-            "- Librarian: at most 10 items per review note. Explain each proposed tag in the Reason column.",
+            "- Tag review notes: at most 10 items. Explain each proposed tag in the Reason column.",
         ].join("\n"),
     },
     scholar: {
         name: "scholar",
         label: "Scholar",
-        summary: "literature notes, syntheses, searches and alerts; the full librarian",
+        summary: "literature notes, syntheses, searches and alerts; all library changes",
         without: MANUSCRIPT,
         withoutPrompts: ["review"],
         batch: 25,
@@ -40,7 +40,7 @@ export const PROFILE_SPECS = {
             "Profile: Scholar.",
             "- Write literature notes and syntheses from what you read, with a source for every claim.",
             "- Do not draft text for {{user}}'s manuscripts, theses or assignments. If asked, say that the Author and Editor profiles help with manuscripts (/profile).",
-            "- Librarian: 25 items per review note.",
+            "- Tag review notes: 25 items.",
         ].join("\n"),
     },
     author: {
@@ -55,7 +55,7 @@ export const PROFILE_SPECS = {
             "- Write literature notes and syntheses from what you read, with a source for every claim.",
             "- Help with {{user}}'s manuscripts: check citations against the library, export the bibliography, and comment on the argument, the structure and the evidence that is missing, with the sources that could support each point.",
             "- Do not write the manuscript's paragraphs. Say what is missing and where the support is; {{user}} writes the text. The Editor profile drafts text (/profile).",
-            "- Librarian: 25 items per review note.",
+            "- Tag review notes: 25 items.",
         ].join("\n"),
     },
     editor: {
@@ -69,7 +69,7 @@ export const PROFILE_SPECS = {
             "Profile: Editor. All tools.",
             "- Write literature notes and syntheses from what you read, with a source for every claim.",
             "- When {{user}} asks, draft text for a manuscript: cite library items as [@citekey], use only claims from sources you read, and mark drafted text clearly so that {{user}} can check it.",
-            "- Librarian: 25 items per review note unless {{user}} asks for more.",
+            "- Tag review notes: 25 items unless {{user}} asks for more.",
         ].join("\n"),
     },
 };

package/dist/prompt.js

@@ -48,12 +48,12 @@ export function fill(text, cfg) {
         .replace(/[ \t]+\n/g, "\n");
 }
 export function roleText(mode, cfg) {
-    const raw = readIf(join(ROLE_DIR, `${mode}.md`)) ?? `You are the ${mode}.`;
+    const raw = readIf(join(ROLE_DIR, `${mode}.md`)) ?? `You are Sub-Sub, the ${mode}.`;
     return cfg ? fill(raw, cfg) : raw;
 }
 export function systemAddition(mode, cfg, cwd) {
     const user = who(cfg);
-    const parts = [`# Sub-Sub: ${mode} mode`, roleText(mode, cfg)];
+    const parts = [`# Sub-Sub: ${mode === "librarian" ? "Librarian" : "Researcher"} mode`, roleText(mode, cfg)];
     const shared = readIf(cfg.sharedRules);
     if (shared) {
         parts.push("# Shared Zotero rules (from the Sub-Sub folder)", shared, `Sub-Sub note: in Sub-Sub you do not need a separate dry run before a change. Call the write tool with dry_run=false; Sub-Sub shows ${user} the server's preview and applies it only on approval. This replaces the dry-run steps above.`);

package/dist/roles.js

@@ -3,7 +3,7 @@
  */
 import { profileSpec } from "./profiles.js";
 export const FILE_TOOLS = ["read", "edit", "write", "grep", "find", "ls"];
-/** Prefixes of the tools that Sub-Sub's servers provide (the gate checks the mode for these). */
+/** Prefixes of the tools that Sub-Sub's servers provide (the gate checks the mode and the profile for these). */
 export const SERVER_PREFIXES = ["zotero_", "scholar_", "verify_"];
 export const ZOTERO_READ = [
     "status",
@@ -40,26 +40,38 @@ export const PREVIEW_WRITES = new Set([
 ]);
 /** Writes without a dry run that still need a yes. */
 export const CONFIRM_WRITES = new Set(["scholar_export_bibliography"]);
-export function toolsFor(mode, registered, profile = "editor") {
+/** Zotero tools that look at the library or write only a note in the Sub-Sub folder. */
+export const ZOTERO_CHECKS = ["tag_audit", "audit_metadata", "find_duplicates", "check_retractions", "missing_pdfs", "write_tag_review"].map((n) => `zotero_${n}`);
+/**
+ * The tools of a mode. The Librarian manages the library: Zotero tools and files.
+ * The Researcher has the same, plus searches, notes and reference checks. The
+ * profile then removes tools. Every library change is still previewed (gateKind).
+ */
+export function toolsFor(mode, registered, opts = {}) {
     const has = new Set(registered);
-    const without = new Set(profileSpec(profile).without);
-    const wanted = mode === "librarian"
-        ? [...registered.filter((n) => n.startsWith("zotero_")), ...FILE_TOOLS]
-        : [
-            ...ZOTERO_READ,
-            ...registered.filter((n) => n.startsWith("scholar_") || n.startsWith("verify_")),
-            ...FILE_TOOLS,
-        ];
+    const without = new Set(profileSpec(opts.profile ?? "editor").without);
+    const zotero = registered.filter((n) => n.startsWith("zotero_") && (opts.bench || !n.startsWith("zotero_bakeoff_")));
+    const wanted = [
+        ...ZOTERO_READ,
+        ...ZOTERO_CHECKS,
+        ...zotero,
+        ...(mode === "researcher" ? registered.filter((n) => n.startsWith("scholar_") || n.startsWith("verify_")) : []),
+        ...FILE_TOOLS,
+    ];
     return [...new Set(wanted)].filter((n) => has.has(n) && !without.has(n));
 }
 /** Why a Sub-Sub tool is not available now, or undefined when it is. */
-export function unavailableReason(name, mode, registered, profile) {
-    if (toolsFor(mode, registered, profile).includes(name))
+export function unavailableReason(name, mode, registered, opts = {}) {
+    if (toolsFor(mode, registered, opts).includes(name))
         return undefined;
+    const profile = opts.profile ?? "editor";
     if (profileSpec(profile).without.includes(name)) {
         return `${name} is not part of the ${profileSpec(profile).label} profile. The user can change the profile with /profile.`;
     }
-    return `${name} is not available in ${mode} mode. The user can switch with /${otherMode(mode)}.`;
+    if (mode === "librarian" && toolsFor("researcher", registered, opts).includes(name)) {
+        return `${name} is part of the Researcher, which also does everything the Librarian does. The user can switch with /researcher.`;
+    }
+    return `${name} is not available.`;
 }
 /** A dry run only when dry_run is absent (server default true) or clearly true. */
 export function isDryRun(value) {
@@ -74,6 +86,3 @@ export function gateKind(toolName, input) {
         return "path";
     return null;
 }
-export function otherMode(mode) {
-    return mode === "librarian" ? "researcher" : "librarian";
-}

package/dist/subsub.js

@@ -1,15 +1,16 @@
 /**
  * Sub-Sub: a Zotero librarian and research assistant for pi.
  *
- * - Two modes with different tool sets: librarian (changes the library, no web)
- *   and researcher (reads the library, searches PubMed/OpenAlex, writes vault notes).
+ * - Two modes: the Librarian manages the library (Zotero tools only); the
+ *   Researcher does the same and also searches PubMed, Europe PMC and OpenAlex
+ *   and writes notes. Each mode can have its own model.
  * - An approval gate: every library change is previewed by the server (dry run)
  *   and shown to the user; nothing is applied without a yes.
  * - Profiles (reader, scholar, author, editor) set what Sub-Sub writes and which
  *   tools it offers; /profile changes it.
  * - Commands: /librarian, /researcher, /profile, /subsub, /history, /undo.
  * - Add-on: with "addons": ["starbuck"], Starbuck runs as a third server
- *   ("verify") for reference checks; its tools belong to researcher mode.
+ *   ("verify") for reference checks.
  *
  * The Python servers start in session_start (not in the factory) and stop in
  * session_shutdown. Tools are registered once and use the current bridge.
@@ -27,8 +28,8 @@ import { describeArgs, formatPreview } from "./preview.js";
 import { headerLines, paintPreview, QUOTES, schemeFromAnsi, themeName } from "./look.js";
 import { systemAddition } from "./prompt.js";
 import { gateKind, SERVER_PREFIXES, toolsFor, unavailableReason } from "./roles.js";
-const PACKAGE_DIR = join(dirname(fileURLToPath(import.meta.url)), "..");
 const MODE_ENTRY = "subsub-mode";
+const PACKAGE_DIR = join(dirname(fileURLToPath(import.meta.url)), "..");
 const BENCH_ENTRY = "subsub-bench";
 function ownVersion() {
     try {
@@ -102,8 +103,10 @@ export async function createSubsub(pi, deps = {}) {
     }
     if (deps.bridge)
         registerTools(deps.bridge);
-    pi.registerFlag("librarian", { description: "Start Sub-Sub in librarian mode", type: "boolean", default: false });
+    pi.registerFlag("librarian", { description: "Start Sub-Sub in the Librarian mode", type: "boolean", default: false });
     let mode = cfg.defaultMode;
+    const benchRun = Boolean(process.env.SUBSUB_BENCH_OUT);
+    const toolOpts = () => ({ profile, bench: benchRun });
     let profile = cfg.profile;
     const user = who(cfg);
     /** True when started with the `subsub` command (not as a package inside plain pi). */
@@ -147,7 +150,7 @@ export async function createSubsub(pi, deps = {}) {
         if (!lookOn(ctx) || cfg.themes === false)
             return;
         scheme ??= schemeFromAnsi(ctx.ui.theme.getFgAnsi("text"), ctx.ui.theme.name);
-        const name = themeName(mode, scheme, cfg.themes);
+        const name = themeName(scheme, cfg.themes || undefined);
         const theme = name ? ctx.ui.getTheme(name) : undefined;
         // A Theme object (not a name) changes the theme for this run only; the saved setting stays.
         if (theme)
@@ -219,7 +222,7 @@ export async function createSubsub(pi, deps = {}) {
     }
     async function applyMode(ctx, next, remember) {
         mode = next;
-        pi.setActiveTools(toolsFor(mode, registered(), profile));
+        pi.setActiveTools(toolsFor(mode, registered(), toolOpts()));
         const spec = cfg.models[mode];
         if (spec) {
             const i = spec.indexOf("/");
@@ -230,7 +233,6 @@ export async function createSubsub(pi, deps = {}) {
         }
         if (remember)
             pi.appendEntry(MODE_ENTRY, { mode });
-        applyTheme(ctx);
         status(ctx);
     }
     pi.on("session_start", async (_event, ctx) => {
@@ -249,6 +251,7 @@ export async function createSubsub(pi, deps = {}) {
         // Judge light or dark before Sub-Sub changes the theme.
         if (lookOn(ctx))
             scheme ??= schemeFromAnsi(ctx.ui.theme.getFgAnsi("text"), ctx.ui.theme.name);
+        applyTheme(ctx);
         await applyMode(ctx, restored ?? (pi.getFlag("librarian") ? "librarian" : cfg.defaultMode), false);
         if (lookOn(ctx)) {
             setHeader(ctx);
@@ -293,7 +296,7 @@ export async function createSubsub(pi, deps = {}) {
             coerceArgs(schema, input);
         resolvePathArgs(name, input, (p) => normalizeToolPath(p, ctx.cwd));
         if (SERVER_PREFIXES.some((p) => name.startsWith(p))) {
-            const why = unavailableReason(name, mode, registered(), profile);
+            const why = unavailableReason(name, mode, registered(), toolOpts());
             if (why) {
                 if (benchOut)
                     record({ kind: "wrong_mode", tool: name, input });
@@ -417,17 +420,17 @@ export async function createSubsub(pi, deps = {}) {
         };
     }
     pi.registerCommand("librarian", {
-        description: "Sub-Sub: switch to the librarian (changes the library, no web)",
+        description: "Sub-Sub: switch to the Librarian (manages the library: tags, imports, metadata, PDFs)",
         handler: async (_args, ctx) => {
             await applyMode(ctx, "librarian", true);
-            ctx.ui.notify("Sub-Sub: librarian mode", "info");
+            ctx.ui.notify("Sub-Sub: Librarian mode. For searches and notes, switch to /researcher.", "info");
         },
     });
     pi.registerCommand("researcher", {
-        description: "Sub-Sub: switch to the researcher (search, notes; library read-only)",
+        description: "Sub-Sub: switch to the Researcher (searches, notes, reviews, and everything the Librarian does)",
         handler: async (_args, ctx) => {
             await applyMode(ctx, "researcher", true);
-            ctx.ui.notify("Sub-Sub: researcher mode", "info");
+            ctx.ui.notify("Sub-Sub: Researcher mode. It searches, writes notes and also changes the library, with a preview.", "info");
         },
     });
     pi.registerCommand("profile", {

package/dist/web.js

@@ -549,13 +549,12 @@ export async function startWeb(opts, env = process.env, log = () => { }) {
                         return send(res, 409, { error: "busy" });
                     const body = await readJson(req);
                     const spec = `${String(body.provider)}/${String(body.id)}`;
-                    // Sub-Sub keeps one model per mode: remember it for the current mode, then restart pi so the
-                    // extension reads the new setting.
+                    // Remember the model for the current mode (or both), then restart pi so the extension reads it.
                     const mode = subsubState?.mode === "librarian" ? "librarian" : "researcher";
                     const models = { ...(loadConfig(env).models ?? {}), [mode]: spec };
                     if (body.both)
                         models[mode === "librarian" ? "researcher" : "librarian"] = spec;
-                    saveConfig({ models }, env);
+                    saveConfig({ models, model: undefined }, env);
                     await agent.request({ type: "set_model", provider: String(body.provider), modelId: String(body.id) });
                     await restartAgent();
                     return send(res, 200, { ok: true });
@@ -583,8 +582,8 @@ export async function startWeb(opts, env = process.env, log = () => { }) {
                     const body = await readJson(req);
                     saveApiKey(agentDir, String(body.provider ?? ""), String(body.key ?? ""));
                     const current = loadConfig(env);
-                    if (body.provider === "opencode-go" && Object.keys(current.models ?? {}).length === 0)
-                        saveConfig({ models: TESTED_MODELS }, env);
+                    if (body.provider === "opencode-go" && Object.keys(current.models).length === 0)
+                        saveConfig({ models: TESTED_MODELS, model: undefined }, env);
                     await restartAgent();
                     return send(res, 200, { ok: true });
                 }